Ecosystem overview

The Delphi ecosystem is three products that form one pipeline but also work apart:

Planning (DPE) Delphi Planning Ecosystem (DPE) — the research design: orientation, phenomenon, panel and question architecture.

Facilitation (Pronoia) Pronoia, descriptive facilitation — running the panel: rounds, dialogue, the pause, and descriptive reports.

Analysis (DAE) Delphi Analysis Ecosystem (DAE), interpretive analysis — in-depth analysis of the collected data under quality gates and human review.

The pipeline runs as handoffs: Planning → Facilitation (Pronoia) → Interpretive analysis (DAE). Each part can also be used on its own.

🗺️ Interactive ecosystem map: open the visual, clickable overview — tabs, phases, quality gates and foundations. Offline, projectable.

Getting started

Type metodix (in Claude Code /metodix) — it opens the Metodix start menu and routes you to the right part of the ecosystem (planning · pre-analysis · DAE analysis · single method · resume). Everything begins there.

What do you want to do?

Three levels: operate → delegate → system

You don't move up a level by gathering more knowledge — you move up by ceasing to do something by hand. That is the through-line of these guides: each part makes visible the one next line whose crossing actually raises your level — not "here is every feature".

The methodological guardrail stays. Delegation does not mean losing control: AI suggestions are drafts (they require an audit), interpretation happens in DAE under quality gates, and a panelist's speech/dictation is edited before sending. The arc runs operate → delegate → system, not human out of the loop.

Descriptive vs. interpretive — a key boundary

The most methodologically important boundary runs between Pronoia and DAE:

A Pronoia report is the source material for analysis, not the analysis itself. "What it means" belongs to DAE, under quality gates (κ / QVG) and human review.

Where to go next

• As a designer → Planning (DPE) (orientation, panel, questions)
• As a facilitator → Facilitation (Pronoia) (rounds, dialogue, reports)
• As an analyst → Analysis (DAE) (CP-0…CP-20, quality gates)
• For the whole picture → Integration guide (handoffs end-to-end)

Which way to Pronoia?

Facilitation (Pronoia)

A Delphi study gets into Pronoia by one of two ways. Both end in the same place: a ready study whose panel, theses and rounds you can still edit after it has been created. The difference is where you start from.

Where to find them

Both are side by side in the dashboard sidebar: ➕ New study (wizard) and 📥 Import plan file. Hovering over a button shows a short hint about who it is for.

The ways cross

Choosing one does not close the other:

• In the last step of the wizard you can save the plan as a file and import it later – for example after a colleague has checked it.
• From the import page you can open any plan or template in the wizard and go through it step by step before creating the study.

Next

• Your first Delphi with the wizard – a picture of every step.
• From a plan to Pronoia – the plan file, special templates and example studies.
• Finishing your plan in Pronoia – after either way: from a draft to ready to invite.

Your first Delphi with the wizard

Facilitation (Pronoia)

The wizard is the way to your first Delphi. It asks for the study's details in seven steps, and a ready template fills in every step for you. Your task is to change what is yours: the topic, the phenomenon, the panel and the theses.

Open the wizard from the dashboard button ➕ New study (wizard). Set aside half an hour to an hour.

Before you start

• Every step has guidance. Open it from What happens in this step? The ⓘ next to a field explains a single choice.
• An unfinished plan is kept. The wizard saves it in your browser after every change, so you can continue later on the same computer.
• Nothing is locked. Everything can still be edited in Pronoia after the study has been created.

Start from a template

The templates are at the bottom of the wizard's first step, below the navigation buttons. There are four, one for each Delphi variation:

• Classic Delphi – scale theses with reasoning and an expert panel, two rounds. The safest first choice.
• Argument Delphi – the reasoning and the panel's discussion are the core of the data.
• Real-Time Delphi – one round that is open from start to finish; panelists see the group's position and update their own at any time.
• Barometer Delphi – the same theses in recurring measurement waves for long-term monitoring.

If you are not sure where to start, choose Classic Delphi or Real-Time Delphi. Press Use this template on the template's card.

Once a template is loaded, an orange box at the top of every step tells you what is worth changing in the template at that step. In an empty wizard the template loads straight away. If steps 2–7 already have content, the wizard first asks whether to replace it with the template. The title, subtitle and purpose are always kept.

Step 1: Study

Write the Title and Purpose / research question for your own topic. On a first study, keep the template's orientation and variation as they are – they were chosen to fit the template. Each option has a short explanation underneath, and the link Which variation fits? Read the guide: the four variations → takes you to a comparison.

Locale is the language in which the panelists see the study. It can differ from the language you use the wizard in.

Step 2: Phenomenon

The phenomenon description is the core of the study: what is studied and what is left out. Replace the template's FILL IN text with your own description – at least 100 characters. Sub-questions and blind spots are optional; empty rows can be removed with the × button.

Step 3: Panel

The panel is described as a matrix with two axes: Expertise (what the panelist knows about the phenomenon) and Stake (whom the phenomenon affects). Each axis has categories, and each category has a title and a description. Change the template's categories into the groups of your own field.

Panelists are written as roles, not names – for example "municipal director of education". Names and email addresses are added later in Pronoia's panel view. The panel must include at least one human; AI panelists are optional.

An AI panelist's perspective profile is shown as a chart: the four axes with their names and values, and the other AI panelists as faint dashed outlines, so you can see how far apart the profiles are. Download PNG and Download SVG save the chart with its headings and the poles of each axis, for example for a report or a slide.

Step 4: Theses

The theses are the statements and questions the panel answers. Keep the template's structure and question types, and write the thesis texts in your own words (at least 20 characters). You add new theses with the buttons at the bottom of the page, for example + Scale or + Open question.

Step 5: R0 Intake

Round 0 is an optional orientation conversation with a few key informants before the first round. It is off by default, and on a first study it is best left off. In the Barometer template, this step also sets the schedule of the measurement waves.

Step 6: Plan review

Review your plan against three criteria: relevance, clarity and diversity. Tick a criterion once you have satisfied yourself that it is met – Pronoia does not check them for you. The templates bring the criteria already ticked, so read them through anyway and untick any criterion that is not yet met.

Step 7: Summary and creation

The summary shows the whole plan at a glance. When everything looks right, press Create study in Pronoia →. The study is created at once, and you can open it in the facilitator view.

If you like, you can also save the plan as a file with Save the plan file (Bootstrap Package, JSON) – for an archive or for a colleague to check, for example. You do not need the file if you create the study directly.

If you stop halfway

When you open the wizard again on the same computer and browser, it continues where you left off and tells you when the draft was saved. Start over gives you an empty wizard – the wizard asks you to confirm before the draft is removed.

What next

The study is now in Pronoia as a draft. Next you finish it: confirm the AI mode, fill in the controller, make the theses your own, add the real panelists and create the invite links. This is shown with pictures in Finishing your plan in Pronoia. From opening the round onwards: The facilitator's process, from planning to handover.

From a plan to Pronoia

Planning (DPE) → plan file → Facilitation (Pronoia)

This is the way when the plan already exists – or when you want to start from a special template or a ready example study. Open the page from the dashboard button 📥 Import plan file.

Where a plan file comes from

A plan file (JSON) is the plan in machine-readable form. It comes about in two ways:

• With the planning plugins. Type metodix in Claude and choose planning. When planning asks for the target platform, choose Pronoia. After the quality gate, planning produces a plan file that you save on your computer.
• From the wizard. The wizard's last step has the button Save the plan file (Bootstrap Package, JSON).

The planning plugins are for Metodix's licensed partners. If you do not have them yet, Metodix will give you the installation instructions.

Import your own file

Drop the file on Your own plan file, or click to choose it. Pronoia checks the plan at once and shows a preview. Nothing is created yet.

Or start from a template

At the bottom of the page there are two sets of templates:

• Special templates – ready plans for particular needs: a participatory civic Delphi, a corporate strategy Delphi and a perspective panel made up entirely of AI panelists.
• Ready example studies – filled-in examples on familiar topics with AI panelists. Create an example, run the AI panelists and see what a finished Delphi looks like.

Each template card has two buttons: Preview leads to the same preview as your own file, and Open in the wizard opens the template in the wizard for editing.

Read the preview

The preview tells you whether the plan is in order and shows it with the same names as the wizard: study, phenomenon, panel, theses, rounds and Plan review. Notes do not prevent creation; if the plan needs fixing, the create button stays grey and the reasons are listed.

The Technical details are folded below the preview. Open them when you check a file from the planning plugins: they hold the file's schema identifier, producer, route and language code, and the perspective layer figures if the plan has one.

Create the study – or open it in the wizard

Below the preview there are three choices:

• Create study → creates the study directly from the plan. Everything in the file goes into Pronoia.
• Open in the wizard opens the plan in the wizard, where you can go through it step by step and edit it before creating the study. The wizard does not carry all the extra details from the planning plugins (for example the perspective layer) – if they matter, create the study directly.
• Cancel goes back to the start.

If the wizard has an unfinished draft, the page first asks whether to replace it. The wizard shows at the top where the plan came from.

The same plan twice

If the same plan file has already been imported, the page says so and asks whether to create a new parallel study anyway: Create a new study or Don't create. A plan is sometimes run again on purpose, so the import does not prevent it.

After creation

Once the study has been created, press Open the study →. If the file had fields that have no counterpart in Pronoia, they are listed under Fields with no receiver – the import still succeeded. The study is a draft that you finish before the panel can be invited: Finishing your plan in Pronoia. From there the work goes on as with any study: The facilitator's process, from planning to handover. The whole arc from planning to analysis: Integration: Plan → Pronoia → DAE.

Finishing your plan in Pronoia

Facilitation (Pronoia)

Once the study has been created with the wizard or imported as a plan, it is a draft in Pronoia. The template has filled in a frame: the description, the rounds, the panel roles and the theses. Some things, however, are always left to you, because they are facts about your study that a template cannot know.

This guide takes the draft to the point where the panel can be invited to the first round. Set aside half an hour to an hour – most of it goes into the theses and the panelists.

Where you start

On the wizard's last screen, press Open in facilitator view →. On the import page the same button is Open the study →. Both screens also link to this guide.

Later you reach the study from the left side of the dashboard, under Studies.

What now? and the four areas

At the top of the facilitator view is the What now? card. It tells you which stage the study is in and lists what blocks progress (BLOCKED) and what is still missing (MISSING). The link on each row takes you straight to the place where the item is fixed.

Below the card are the study's four areas: Planning, Collection, Analysis and Reporting. In this guide you work almost entirely in Planning. Its tools are Overview, Settings, Round plan, Panelists, Theses and AI panelists.

Next to each view's heading is a label saying what a change applies to, for example Applies to: the study or Applies to: round 1. It tells you whether you are changing the whole study or one round.

Overview: what is still missing

Overview is a list headed Building the study. Pronoia computes it from what the study contains – you do not tick it yourself.

• ✓ means the item is in place.
• ○ means something is missing from the item.
• · means a decision, not a gap (for example whether Round 0 is run).
• The label recommended means the item does not block opening, but it is worth taking care of.

Below the list it says how many blocking items are open: Opening is waiting on 1 item(s). The → button on a row takes you to the view where the item is fixed.

What came from the template, what is left to you

The order is up to you. Below, the items are covered in the order that is usually most natural.

1. Make the basics your own

Open Settings. At the top is Study basics:

• Edit title opens a window where you change the study's name. A template's name begins with Template: – change it, because the panelists see the name in the invitation.
• Description is a one-sentence introduction.
• Phenomenon is the core of the study: what is studied and why now. Replace the template's FILL IN text.
• Attachments for panelists is optional: for example a one-page introduction to the panel. Panelists see the attachments on the card that introduces the study.

Changes are saved automatically: after a moment the end of the form reads Saved at hh:mm.

2. Confirm the AI mode

The first round cannot be opened until you have consciously chosen what AI may do in the study. Under AI use in this study, press Confirm the AI mode. Choose one of three:

• No AI — nothing is sent to an AI provider
• Help only — method help, no panel content sent out (default)
• Full — also AI panelists, R0 chat, deepening, coding

Press Confirm and continue. Afterwards the section says The AI mode is confirmed. If you later change the mode, you confirm it again.

3. Controller and legal basis

In the template, Require informed consent before answering is already switched on: the panelist accepts the consent text before seeing any panel content. That is why invite links cannot be created until Controller and legal basis has been filled in:

• Contact person's name and Contact person's email – you, or whoever answers the panelists' messages. Left empty, your own account details are used.
• Controller's name and Contact details – the organisation that carries out the study.
• Legal basis for processing – choose from the list. Under each basis is a note on what it means.

The Consent text is ready in three languages. You do not need to write it unless you want your own wording.

There is no need to save. Once the fields are in place, the Settings row in the overview turns ✓.

4. Rounds

Open Round plan and press Edit phasing. Here you decide two things:

• Round 0 – whether an orientation is run before the first round, in which key informants discuss the phenomenon. In the template it is off (No). Choose Yes if you want to hear the panel before the theses are fixed.
• Actual rounds – usually two or three. Reserve one too many rather than too few: an unused round costs nothing.

The path shows the whole arc from planning to the end. Press Save.

If you switch Round 0 on, its orientation questions are set on the Collection tab Round 0. More on that in The facilitator's process, from planning to handover.

5. Make the theses your own

Open Theses. At the top is a form for a new thesis, and below it the theses of round 1. The template's theses are written in general terms and include an instruction in brackets, for example (Edit for your own phenomenon.)

Press Edit on a thesis and write your own. In the same form you can add an explanation and adjust a scale thesis's criteria, such as probability and desirability. Preview: how a panelist sees this question shows the thesis through a panelist's eyes. Press Save.

You add a new thesis with the form at the top and the + Add button; there you also choose the thesis type, for example Scale or Open. A template thesis you do not need can be deleted (Delete) or hidden (Hide).

6. Panel and panelists

Open Panelists. At the top is the panel matrix: the template's roles (P-01, P-02 …) are placed in cells by expertise and stake. Below the matrix, PANEL COVERAGE shows how well the viewpoints are covered: Coverage score, Missing viewpoints (empty cells) and, when needed, the warning Panel supplement recommended before the round.

The template's roles are places, not people. Press Edit on each one and enter the real person's name and email. On the same row you can check the cell and mark the person as a decision-maker (Decision-maker). Press Save.

Add new panelists with the form in the view and the + Add button – especially to the cells where a viewpoint is missing. A panelist without an email address cannot be invited, so the Panelist contact details row in the overview turns ✓ only when everyone has contact details.

Try as panelist shows the study as a panelist sees it.

7. AI panelists (optional)

If you chose Full as the AI mode, you can add AI panelists on the AI panelists tab: Create 5 default profiles, Generate profile suggestions from the phenomenon, or Import Planning Handoff. AI panelists are always marked in the panel. In a first study they are not needed.

8. Invite links

When everything above is in place, go to the Collection area and open Invitations. Press Write and send the invitation letter →. Pronoia creates a personal link for each panelist and opens a ready invitation in Communication, addressed to the panel's human panelists. Nothing is sent until you press Send there – sending is covered in the next guide. If you only want to create the links, Generate invite links does that without opening a letter.

You do not see the links themselves. A new study is in closed link mode: each link travels only in the letter to its own panelist, and the screen just says that the links have been created. The button next to the letter shows how many there are, for example ✓ Invitation links created (3) – create missing. Closed mode protects the panelists' anonymity, and it is the recommended setting.

If you would rather send the invitations from your own email, choose Open in Settings under Facilitator access to panelists’ response links. Invitations then shows each panelist's link, Copy copies it to the clipboard, and EMAIL TEMPLATE gives a message in which you replace {name} and {link} with each panelist's details. In blind mode the links are always closed.

Creating the links does not yet send anything or open the round.

Ready to invite

Go back to Overview in the Planning area. When the list is in place, it says Everything that blocks opening is in place. below it, and the What now? card suggests Move to the response phase next.

The plan is now ready. Opening the first round, sending the invitations and following the collection are the subject of the next guide: Your first round in Pronoia.

Remember

• Nothing is locked. Theses, panelists and settings can still be edited between rounds – the guide simply follows the order in which the first round comes together.
• The list is computed. If a row stays ○, press its → button: it takes you to the place where the item is fixed.
• Practise safely. You can rehearse on an example study before your own: The facilitator trial in five stages.

Your first round in Pronoia

Facilitation (Pronoia)

The previous guide ended with the plan ready and the What now? card suggesting the response phase next. This guide carries on from there. In it you open round 1, invite the panel, follow the responses and the dialogue, close the round and prepare round 2.

The main path is Argument Delphi, Pronoia's signature variation: the panelists rate the theses and justify their ratings, then read each other's justifications and discuss them. The differences in the other variations are at the end of the guide.

A round usually lasts 1–3 weeks. Your own work in it is an hour or two: opening and invitations at the start, a check every few days, and closing at the end.

How a round runs

Everything in this guide is done in the Collection area. Its tabs are Situation, Round management, Invitations, Facilitation, Communication and Responses.

In Argument Delphi, round 1 runs like this:

The path also shows the phase 1c Revision marked ↺ during the dialogue. In Argument Delphi the panelists revise their ratings during the dialogue, and there is no separate revision phase.

All transitions are made with the same orange button on the Round management tab. Pronoia does not move the round on by itself, and the deadline does not close the round.

1. Before opening: deadline and visibility

Open Collection › Facilitation. The top line says that round 1 is not open yet and that the settings take effect when you open the round.

• Deadline is a date the panelists see. It does not close the round; it reminds you and the panel. +7 days moves it a week on, Clear removes it.
• Visibility & interaction sets, separately for each phase, what a panelist sees. The template's default suits most studies: in the response phase a panelist sees the others' responses only after giving their own (After you answer), and in the dialogue phase everything is open (Open). The checkboxes (Edit own, Commenting, Threads, Graph) allow or block single actions.

2. Open round 1

Open Round management. The path shows the whole course from planning to closure, and the You are here label tells you where you are. The orange button is always the next step. Now it is Open round 1.

On the card's INFORMING AND INTERACTION row, the buttons Write to the panel and Invites and links lead to the same actions as the Communication and Invitations tabs.

Press Open round 1. The round is now open and the phase is Response. The card shows the number of responses (at first 0 / 9 – three panelists times three theses), and the orange button changes to the next step, Move to the dialogue phase.

Opening does not yet send anything to anyone, and it does not use credits. The panelists hear about it only when the invitations go out. That is why the reminder button is still dimmed (Remind 0 not yet answered), and the note beside it says that three panelists have not been invited yet.

3. Invite the panel

There are two ways to send the invitation.

Through Communication (recommended). Open Communication. In the COMMUNICATION CENTRE the recipient is Whole panel. Choose the Invitation to the study template. If you come to Communication from the button on the Invitations tab, the template is already selected and email is switched on. The letter comes with ready text and placeholders: [nimi] is replaced by the panelist's first name, [link] by their own link and [study] by the study's name. The short Round open template is meant for later rounds, when the panelists already know the study.

• Also send by email takes the message to the panelists' inboxes. Without it the message appears only in Pronoia.
• Include »What we ask of you« adds a short description of what is expected of the panelist to the letter.
• Preview the letter shows the letter exactly as it will go out, and sends nothing.

Finally press Send. The delivery report tells you how many inboxes the message reached and who had no address.

The links yourself. This needs open link mode, which you choose in Settings (see Finishing your plan in Pronoia); in the default closed mode the links are not shown. In open mode, open Invitations. There each panelist has a personal link, which Copy copies to the clipboard. The link is personal, so do not send the same link to two people.

4. What the panelist sees

The panelist opens their link and first accepts the consent text (I have read this and I take part). Then they press Start answering → and see the theses one at a time.

In Argument Delphi a scale thesis is rated on two criteria, for example probability and desirability, on a scale of 1–7. A justification is required. In open theses the panelist writes an answer, for example the strongest counter-argument. Skip question leaves the thesis out.

The panelist sees everything under a handle (Aristonike in the picture), and the other panelists see them only by it. As the facilitator you also see the code (P-01).

5. Follow the responses

Situation shows at a glance where the round stands: the number of responses, justifications, panelists who have answered, comments and visits. What do the figures mean? explains each figure.

If responses are slow, the button Remind 3 not yet answered on the Round management card sends a reminder to those who have not answered yet (the figure is the number of people who have not answered). The same works on the Communication tab with the Reminder template.

The responses themselves are on the Responses tab, thesis by thesis: who answered, with which values and with which justification.

6. The dialogue phase

When there are enough responses, press Move to the dialogue phase on the Round management tab. If many responses are missing, Pronoia first asks whether you really want to move on.

Dialogue does not close answering: a late panelist can still respond. Now each panelist sees, for every thesis, their own response, the panel's distribution and the others' justifications anonymously. They can comment on the justifications, reply to comments and change their own rating with Edit. In Argument Delphi this is the most important phase of the round: changing a rating in the light of another's justification is the point of the method.

As the facilitator you stay outside the discussion. If you want to tell the panel something, use the Communication tab rather than commenting on justifications yourself.

If you moved the round to dialogue too early, Back to the previous phase restores the response phase. The step back is recorded in the study's log.

7. Pick material for round 2

The Responses tab gathers the round: at the top the numbers of panelists and responses, below that every thesis with its responses.

Further down is a timeline, where justifications and comments appear from newest to oldest. When you notice a justification or a comment that could become a round 2 thesis, mark it with a star (☆). The pick goes to MATERIAL FOR THE NEXT ROUND, and in the field Why did you pick this? Visible only to you — in the basket and in the thesis sources, never to the panel. you can write yourself a note.

A pick does not change the comment and is not visible to the panel. It is your memory between rounds.

8. Close round 1

When the discussion has quietened down or the deadline has passed, open Round management. In the dialogue phase the card also shows a Disagreement row: on which theses the panel shares a view, where it is scattered and where it splits into two camps. To the summary → takes you to the detailed results.

Press Close round 1 (pause). In Argument Delphi the revisions are made during the dialogue, so the round moves from the dialogue straight to the pause. Closing does not end the study. Answering ends, and the material waits for the next round. The study is ended separately with End the study when the time comes.

The What now? card now tells you that you are in the pause and names the next task: an instrument has to be chosen for round 2.

9. Prepare round 2

During the pause there is a card below the tab row: BUILD ROUND 2 — PREPARATION PAUSE. It offers three ways forward:

• Copy round 1 as a starting point brings round 1's theses in as a draft for round 2. You then edit them in the light of what the dialogue taught.
• Start from scratch → takes you to round 2's definition, where you write the new theses. Your picks wait in the Planning area › Theses under SOURCE MATERIAL FOR THE NEW THESIS: select picks, write the thesis, and the selected picks are recorded as its sources.
• Open with the same theses (re-rating) opens round 2 straight away, and the panel rates the same theses again.

PIRE — method assistant (Run analysis) reads round 1's results and suggests, thesis by thesis, how the next round could be built. It suggests; it does not decide. Notify panelists sends the panel a message in Pronoia that round 1 has been closed and round 2 is being prepared.

When round 2's theses are ready, the round is opened in the same way as round 1, and the steps of this guide repeat.

Other variations

Classic Delphi runs through the same phases as Argument Delphi. The difference is in the aim: Classic Delphi seeks a shared view, so in the pause you read above all on which theses the positions converged and on which they did not. If the round's dialogue phase is switched off in the plan, the orange button after the response phase is directly Close round 1 (pause).

Real-Time Delphi runs a single round in which answering and dialogue are open at the same time. Opening is Open Real-Time Delphi (publish), and a panelist can change their response at any time. Round management has no phases; you close the window with Close the window (pause) once the picture is stable. No round 2 is built.

Barometer Delphi measures the same theses again wave by wave. The round is always open (Always open — a wave is saved, not closed) and is never closed. A wave is saved with Save wave 1 (the wave number changes), and between waves a thesis can be retired and a new one brought in.

Remember

• The round does not move on by itself. The deadline does not close the round, and every transition is your own decision with the orange button.
• Opening invites no one. Invitations go out only from the Communication tab or from your own email.
• Pick at once. Star a promising justification when you notice it. In the pause it is waiting for you.
• The whole process in more detail, including Round 0 and reporting: The facilitator's process, from planning to handover.
• Next: The second round and closing.

The second round and closing

Facilitation (Pronoia)

The previous guide ended in the pause after round 1. This guide carries on from there. In it you build the round 2 survey from round 1's results, open the round, close it and end the study. Reading the results and exporting the material for analysis are in the next guide.

The main path is still Argument Delphi. The differences in the other variations are at the end of the guide.

What is different in round 2

Round 2 runs through the same phases as round 1: response, dialogue and pause. The difference is in the questions. The second round of a Delphi usually asks new, more focused questions derived from round 1's results and discussion. Therefore:

• Round 2's theses are theses of their own. They are named R2-T1, R2-T2 … and round 1's theses R1-T1, R1-T2 …
• The panelist sees round 1's result next to the thesis from which the round 2 thesis was derived.
• Development between rounds is followed at panel level. When the questions are different questions, their figures are not compared one by one. The analysis shows how the shares of the classes (shared view, scattered, two camps) change from round to round. Qualitative development is read in the DAE analysis.

1. Build the round 2 survey

In the pause after round 1 there is a card below the tab row: BUILD ROUND 2 — PREPARATION PAUSE. The previous guide introduced its three ways forward. This guide takes the most common one: Copy round 1 as a starting point.

After copying, round 2's theses are drafts below the card. The copy inherits the thesis settings, for example the criteria, the scale and the required justification. Open a thesis with Edit and reword it in the light of what the dialogue taught. The thesis card's SOURCE MATERIAL tells you where the thesis was derived from (Derived from thesis R1-T1). If you starred comments in the dialogue, they wait under SOURCE MATERIAL FOR THE NEW THESIS: select a pick and write a new thesis from it, and the pick is recorded as the thesis's source.

If you want to start over, Clear Round 2 removes round 2's drafts.

AI aids. The card also has Deepening thesis suggestions (AI). It picks drafts for new theses from the justifications of contested theses, and every draft refers to the justifications it came from. The suggestions are drafts that you word and own. The more precise argument analysis is done in DAE. PIRE — method assistant, on the other hand, is not AI: it computes a suggestion, thesis by thesis, from round 1's figures.

2. Open round 2

When the theses are ready, press the card's Open Round 2 (new theses) →. Round 2 is now open in the response phase, and the round card counts the round's own responses: three panelists times three theses gives 0/9.

If you notice after opening that the survey is not ready, Return Round 2 to preparation returns the study to the preparation pause. Do it before the panel starts answering.

Invite the panel the same way as in round 1: from the Communication tab with the Round open template, or from your own email. A panelist's link stays the same throughout the study.

3. What the panelist sees in round 2

The top of the panelist's page says that this is a new survey: round 2's theses have been derived from the previous round's results, and they are answered afresh, not by editing earlier responses.

Next to a derived thesis there is a folded Round 1 result (R1-T1). When the panelist opens it, they see how the panel answered the thesis this one was derived from: the distribution and median per criterion. It shows no justifications and not the panelist's own earlier response. The result is folded by default so that it supports judgement rather than steering the answer. A new thesis that you wrote yourself has no such result.

If you do not want to show the previous round's distribution, clear Previous round distribution in the response phase settings on the Facilitation tab.

4. Responses, dialogue and closing

Round 2 runs like round 1:

1. Follow the responses on the Situation tab and remind when needed.
2. Move the round to dialogue with Move to the dialogue phase.
3. Star material in the dialogue if you plan to build a round 3.
4. Close the round with Close round 2 (pause).

The round card's Disagreement row tells you, for round 2's own theses, on which of them the panel shares a view, where it is scattered and where it splits into two camps.

5. The last pause

When the last round of the plan has been closed, the path in Round management shows the step Pause · end the study or add a round. The What now? card says that the last round has been closed and the study is waiting for your decision. The study does not end by itself.

There are two options:

• Add a round. If the dialogue opened questions worth taking to the panel, Edit the round plan → adds a round 3 to the plan. You then build its survey the same way as round 2's.
• End the study. When the panel has said what it has to say, end the study.

Before ending, it is worth reading the results. The instructions are in the next guide.

6. End the study

In the round box in Round management, press End the study. Ending a Delphi is your decision that the panel has said what it has to say. It is not the same thing as the last round closing. Ending:

• closes the open round, if there is one
• opens the Closing phase to the panel
• makes the next DAE handover final.

The window Closing — end the study opens, with the form Closing letter to the panel. The letter is shown on the panelist's home page, and it can also be sent by email. The form contains a template letter that is a checklist of what the letter should cover: thanks, what the panel did, what was learned, what follows from the result, and a contact. There are two kinds of square brackets. Fields such as [name], [rounds] and [facilitator] are filled from the data for each reader: the form shows their values, and pressing a field inserts it at the cursor. Prompts are written in your own words or removed. There are two: [2–3 things the panel's answers showed – including where views differed] and [who receives the results, what they are used for, when]. In the picture the whole letter has already been written in the facilitator's own words, so no prompts are left. As long as the letter has a prompt, a misspelled field or a field without a value, the study cannot be ended. The form lists the points, and pressing a point takes you to it.

Under Attachments on the panelist's page you choose what is shown with the letter. The option Also show the panel the same descriptive report you see attaches below the letter the same descriptive results report you see in Reporting: the spread map, the per-thesis results and the highlights. The same report can also be offered as a Word file to download, and the Basic analysis figures per thesis can be attached. No attachments are selected by default. Attachments contain only aggregates, no justifications and no panelist codes. They are not an analysis.

The AI disclosure section shows how many AI panelists, AI responses and AI comments the data contains. Tick Simulated arguments were used in planning the study if you used them. You confirm the disclosure when you end the study, and it is frozen in the archive.

The option Also send the letter by email (through Communication) is on by default. It mails the letter after the study is ended, in the same way as other messages sent from Communication, and the form tells you how many it goes to. Preview the letter shows the letter as it will be sent. If you skip the email now, you can send the letter later from Communication with the template Closing letter.

When the letter is ready, press the form's End the study.

7. What the panelist sees at the end

The panelist's page now says that the study has ended. Below that are A LETTER FROM THE FACILITATOR and the card THE NAME YOU WORE, which tells the panelist the handle the others knew them by (Aristonike in the picture) and that name's story in the history of Delphi. Then come the attachments you chose, and every thesis's final result: the panel's distribution and the discussion. The discussion and the panelist's own responses remain readable, but they can no longer be changed.

What can be undone

• A phase transition can be undone with Back to the previous phase. For example, a round closed too early can be returned to dialogue. The step back is recorded in the study's log.
• Ending can be undone by adding a round. The panel may already have read your closing letter, however, so tell it why the study continues.

Other variations

Classic Delphi often asks the same theses again in the second round. Then you choose Open with same theses (re-rate) → on the card. The panelist sees their own earlier response and can change their rating. Because the theses are the same, the analysis follows each thesis's development from round to round.

Real-Time Delphi is a single round. When you close the window, you are directly in the last pause, and the study is ended with the same closing letter.

Barometer Delphi does not close rounds but saves waves. A barometer is ended with the same End the study button once the measurement series is complete.

Remember

• Round 2 asks new questions. Build it from round 1's results and dialogue, and do not repeat the same survey unless it is a deliberate re-rating.
• The study does not end by itself. Closing the last round takes you to a pause, and ending is your own decision.
• The letter is yours. The template is a checklist: fields are filled from the data, and prompts are written in your own words.
• Next: Results and reporting.

Results and reporting

Facilitation (Pronoia)

This guide carries on where the previous one left off: the rounds are done and the study has been ended or is about to be. In this guide you read the results in Pronoia, make a descriptive results report for the panel and export the material to DAE analysis.

Keep the division of labour in mind. Pronoia describes what the panel answered: distributions, consensus and dissensus, changes of position and the amount of discussion. The figures are computed by rules, without AI. DAE interprets: theming, argument and discourse analysis, scenarios and synthesis are done in the DAE analysis pipeline under quality gates.

Where the results are

The results are in the Analysis area. Its tabs are Summary and, as tools, Statistics & change and Quantitative analysis. At the top of the area you can choose which round's material you look at (Round being viewed R1, R2). The report and the export are in the Reporting area.

1. Summary

The Summary is the overall picture of the study. Analysis — state of the material at its top tells you whether the material is readable and how much dialogue and how many changes of position it contains. With the choices on the CONTENTS row you show or hide sections.

• Overall picture: the numbers of theses, panelists and responses, and the response rate.
• Highlights: the thesis with the widest spread and the thesis that was discussed most. They are often the most interesting points for the analysis.
• Round movement: the change between rounds. If round 2 has its own new questions, thesis-level movement is not computed, because the questions are different questions. Then development is read from the panel-level figures of Quantitative analysis and from the DAE analysis.
• Per-thesis results: every thesis with its distribution and central values.

AI aids. The Summary's EPISTEMIC 7×3 HEATMAP classifies the justifications by their knowledge base and function. The classification can be done by hand or suggested with AI (AI-suggest coding). AI suggestions are suggestions that you accept or correct. The actual argument analysis is done in DAE.

2. Statistics & change

Statistics & change gathers the rounds' figures thesis by thesis.

• Per-thesis panel stats: respondents, comments, comments on comments, reactions, changes of position and the thesis's class (shared view, scattered or two camps). Theses are named with their round: R1-T1, R2-T1.
• Convergence: each thesis's agreement (A) and spread (IQR) from round 1 to the latest, with a label for the direction of change, for example stable. Rising agreement and narrowing spread mean that the panel is converging. A stable dissensus is an equally valid result.
• Dialogicity: how much the panelists commented on each other's rationales and on those comments.
• Shift directions and revisions: who changed their rating, in which direction and on what grounds.

In Argument Delphi the changes of position are made during the dialogue. That makes them the most important part of the material: they show which justification made someone change their position.

3. Quantitative analysis

Quantitative analysis is the basic analysis of the panel's responses. It is computed with the same code as the quantitative stage of the DAE analysis (CP-2), so Pronoia's and DAE's figures are the same for the same material.

• Whose responses: People, AI panelists or Everyone. By default the humans' responses are counted, and humans and AI are not mixed.
• Round: in a multi-round study the analysis shows one round's questions, by default the latest. You can switch the round with the buttons.
• Composition of the survey: how large a share of the round's questions is a shared view, how large a share scattered and how large a share two camps.
• Table: for every question the number of respondents, agreement A, the location on the scale and the class.
• Development by round: each round's own class distribution. The questions of different rounds are usually different questions, so development is followed at panel level: does the share of shared view or of two camps grow. A question-level series is shown only if the same thesis has been rated again. In Argument Delphi the stopping rule is also shown, telling you whether the dissensus is argued and settled.
• Group comparison and Criterion pairs: whether the panel's groups differ from each other and whether two criteria, for example probability and desirability, go together.

The figures tell you what is in the material. What they mean for the research question is a matter of interpretation.

4. The descriptive results report

Open Reporting › Reports. Generate results report compiles a descriptive report of the panel's responses: a summary and the spread by thesis. The report can be viewed in the browser and printed to PDF, and Download Word (.docx) gives it as an editable file.

The report is source material, not an analysis. The same report can be shown to the panel together with the closing letter (previous guide).

5. Export to DAE analysis

Open Reporting › Export. The list Reporting — what remains at the top of the page tells you what is still missing for completing the study: readable material, the handover to DAE analysis, the results report, the closing letter and ending the study.

The export gives three files:

1. Download CSV: the quantitative material as a table, one row per panelist.
2. Download Handoff JSON: the whole material with justifications, dialogues and changes of position. This is the actual input of the DAE analysis.
3. Download DAE_STATE.json: the control data of the analysis pipeline, such as the study's metadata, the panel profile and the orientation.

If the study has been ended, the handover is final. An export made before ending is an interim handover, and you can make one, for example, between rounds.

The bridge to DAE analysis

DAE analysis is an analysis pipeline that uses AI and is run in Claude. Save the three files in the same folder. Then open a new Claude conversation, type metodix and choose DAE analysis from the start menu. Attach the Handoff package. The pipeline proceeds checkpoint by checkpoint: the quantitative foundation, content analyses (dialogue, arguments, discourse), system analyses, the futures phase, synthesis, scenarios and reports. The qualitative development between rounds, that is, how the questions and justifications changed from round to round, is read right there.

Other variations

Classic Delphi seeks a shared view. Read above all Convergence and Quantitative analysis's lead indicator (the consensus indicator) if the same theses have been rated in several rounds.

Real-Time Delphi is a single round. No round comparison is made; instead Quantitative analysis reads from the timeline whether the responses have settled.

Barometer Delphi compares waves. Quantitative analysis shows what has changed from wave to wave and what has stayed.

Remember

• Pronoia describes, DAE interprets. Pronoia's figures are the result of rules, not interpretation.
• The questions of different rounds are usually different questions. Development is followed as shares of the classes (shared view, scattered, two camps) and qualitatively in DAE, not thesis by thesis.
• Export the material after ending the study, so that DAE gets the final handover.
• The whole process in more detail: The facilitator's process, from planning to handover.

Integration: Plan → Pronoia → DAE

A panelist experiences one study — its rounds, its feedback, its Home view. The facilitator runs something larger. Their process spans three stages — design, collection and analysis — and reaches across two connected ecosystems. Pronoia is the bridge between them.

Planning (DPE) → Planning Handoff Package → Facilitation (Pronoia) → Export contract → Analysis (DAE)

Upstream sits the Delphi Planning Ecosystem (DPE), where the study is designed. Downstream sits the Delphi Analysis Ecosystem (DAE), where the collected material is interpreted. Pronoia is the data-collection layer in the middle: it receives a finished design, runs the study, validates the result, and hands a clean dataset on to analysis.

Why this matters for the facilitator. The quality of the analysis is decided long before analysis begins — in how the study is designed and how cleanly it is collected. Owning the whole arc, and the two handoffs that join it, is the facilitator's core responsibility.

Integration is the top of the capability arc

Assembling a single study by hand is operating. When the handoffs carry the design from stage to stage without rebuilding, the whole chain becomes a system — infrastructure the work runs on. The through-line of this guide is exactly that shift: you stop re-assembling the study in every tool, and let the handoffs flow.

Stage 1 — Design (DPE)

Before Pronoia collects anything, the facilitator designs the study in the Planning ecosystem. Each step is a deliberate methodological choice that shapes everything downstream.

Planning ends at a quality gate: only a coherent design — orientation, phenomenon, panel and theses aligned — is packaged for collection.

Handoff in — the Planning Handoff Package

The design crosses into Pronoia as a structured Planning Handoff Package: the orientation, the panel matrix and its type, the theses, and the perspective/PHS context. This seeds Round 0 — so the facilitator does not rebuild the study in Pronoia, but continues it.

Stage 2 — Collection (Pronoia)

This is the facilitator's operational core. In brief, the facilitator opens Round 0 as an AI-assisted intake (motivating participants and mapping the phenomenon, optionally with key informants), then runs each round through its phases — response → dialogue → revision → close — setting the iteration stance with policy presets (classic / RTD / open / barometer), protecting anonymity and identity, and reading the panel through the pulse, the group comparison and the process metrics.

What distinguishes the facilitator is an active role: collection is not a passive wait for answers but a continuous effort to spur the panel along — and the levers work at three levels.

Spurring is targeted, not generic. The most effective prompt is specific: show one group the gap between its position and another's, or show an individual how far the panel has moved since they last looked. Real-time data is what makes that precision possible.

The validation gate — earning the handoff

Collection does not flow into analysis automatically. Before any handoff, Pronoia validates the result, and a failing check blocks the export until it is fixed:

• Panel-composition gates — the orientation-specific minimums (matrix size, stakeholder groups or observation distances, domains or outside roles, decision-maker rules).
• Epistemic coverage — the argument space read as a coverage map; thin coverage is a prompt to supplement the panel before closing.

This gate is the facilitator's quality promise to the analysis stage: what crosses over is complete and well-formed, not raw.

Handoff out — the export contract

On a clean validation, Pronoia emits a structured payload for the DAE ecosystem. At the facilitator's level it carries:

• the orientation and panel matrix type;
• the argument space — theses × panel members and their arguments;
• the epistemic coverage — how much of the argument map is filled;
• the perspective summary — health score, dominant perspectives and named blind spots;
• the validation status — confirming the dataset passed its gates.

Downstream consumers verify this contract before they begin, so a malformed dataset cannot silently enter analysis.

Stage 3 — Analysis (DAE)

In the DAE ecosystem the collected material is interpreted through a gated pipeline. The facilitator commissions and oversees it rather than running every method by hand.

Foundation → Content → System → Futures → Synthesis → Report

Quality gates and human judgement sit throughout: where automated confidence is not enough, the analyst is brought into the loop — and the facilitator decides how far interpretation should reach.

Closing the loop — learning across studies

The arc is not a straight line but a cycle. Both ecosystems keep a memory: the planning side records each design and its lessons, the analysis side keeps a case library of completed runs. The facilitator feeds these lessons back into the next study's design — so each Delphi run makes the next one better.

The facilitator's through-line. Design well, collect cleanly, analyse rigorously, and learn from the result. Pronoia is where that arc is held together — the point where a designed study becomes trustworthy data, ready for interpretation.

The facilitator's end-to-end playbook

1. Design with the end in mind. Choose orientation, scope the phenomenon (pre-analysis), design the panel and write the theses — knowing they define the whole argument space.
2. Cross the first handoff intact. Carry the Planning Handoff Package into Pronoia and continue the study rather than rebuilding it.
3. Run collection deliberately. Open the intake, set the iteration stance, advance the phases, protect anonymity, and watch the pulse and metrics.
4. Earn the second handoff. Clear the validation gate — panel composition and epistemic coverage — before closing; supplement a thin panel if needed.
5. Commission analysis. Export the validated dataset, oversee the DAE pipeline, and decide how far interpretation and reporting should go.
6. Close the loop. Capture the lessons and carry them into the next study's design.

Source: Pronoia — The Facilitator's Process (Metodix, June 2026). This block reframes the same arc along the capability arc.

Delphi variations

A study's variation decides how responses are collected and how movement is tracked over time. It sits alongside the orientation (N/S/R/E/D/D+) chosen in planning, and it sets the collection key and the lead indicator the analysis uses. The names are proper names and the same in every language: Argument Delphi, Classic Delphi, Real-Time Delphi, Barometer Delphi. The value stored in the contract is a different thing from the display name (the value for Classic Delphi is still Classical).

The variation is an editable contract block; if absent, the default is Argument / round / CDI (backward compatible). Common to all four: Round 0 (orientation) belongs to every one, and dialogue exists in every one — the variations differ in where in the round the dialogue sits, not in whether it exists.

Templates for different purposes

Ready-to-run templates pair a variation with a panel and starter theses, so you can instantiate a complete study in one step — from the new-study wizard (which pre-fills every step) or by importing the package. Each is a contract-conformant starting point you then edit.

Next: choose orientation and questions in Planning, and design the matrix in Panel design.

Real-Time Delphi — how to run one

Real-Time Delphi is the variation with one round that is open from the start — for answering and for dialogue at the same time. A panellist waits for nothing: they see the theses and the panel's discussion immediately, answer at their own pace, and can come back and change their position for as long as the study is open.

In the contract the variation is Real-Time, the collection key is window and the lead indicator is RCI. The three travel together: the collection key says the collection period is a window rather than a round, and the indicator says what the analysis reads.

When to choose RTD

RTD suits a study where time is short and participants are scattered. It is asynchronous: each person takes part when they can, and nobody waits for a round to close. It also suits a fast-moving phenomenon, where the gap between two rounds would already have dated the question.

RTD does not fix the aim. It can aim at consensus or at dissensus — the aim is resolved per thesis, not by the variation. This differs from Classic and Argument Delphi, where the aim is a property of the whole study. In practice: if you want the panel to seek common ground on this thesis but map disagreement on that one, RTD gives you that without splitting the study in two.

What RTD does not replace: if you want to see how the panel moved — that someone changed position after reading others' reasoning — you need two rounds. Movement is the difference between two measurements, and a single measurement has no difference to compute. That is the price of RTD, and it is worth knowing before the variation is chosen rather than after.

What the software does for you

The single round is built in, so there is nothing to remember:

• The wizard does not ask for a number of rounds. The round count is not a question but a property of the variation, and in RTD it is one. (Before 5 September 2026 the wizard asked every variation, defaulting to 2 — a choice that does not exist.)
• The server refuses a second round. POST /rounds answers 409 single_round_mode. You cannot drift into a multi-round setting by accident.
• All phases are open. Response, dialogue and revision are all open at once in RTD. That is why the phase bar carries no round number: it says open round, not "Round 1".
• The movement panel says outright that movement is not computed. It is not an empty box but an explicit message — in a single-round study there is no between-round movement.

Planning

Planning follows the same route as in any other variation: orientation, phenomenon, panel, question architecture.

Round 0 belongs to RTD too. It is the orientation and key-informant stage, and it is the only numbered step before that one round. Round 0 material is intake — it is not panel data and does not travel into the analysis as responses.

Keep the number of theses moderate. Because everything is open at once, the panellist sees the whole workload at first glance. In a multi-round setting the work is spread across rounds; here it is not.

Require reasoning. The value of RTD comes from panellists reading each other's reasoning and responding to it within the same window. Without reasoning what remains is a survey with a comment box.

While the study runs

With no round comparison, you follow three things:

1. The build-up of participation. Who has answered, who has started, who has not opened the link. In a single window this is the only measure of progress.
2. The density of dialogue. Is discussion emerging, or are people only answering? In RTD the discussion is what separates it from a survey, so its absence is a finding rather than silence.
3. The panel's pulse per thesis. Where views sit close together and where they spread. This is read from one measurement, not from the difference between two.

The facilitator's role is lighter than in a multi-round Delphi: no rounds are opened or closed, no theses are rewritten between rounds. What remains is what matters most in RTD — keeping the discussion alive. If dialogue fades, you address it with a message to the panel, not with a new round.

Closing

The study ends when the facilitator ends it. The window does not close by itself and the round does not "fill up". Closing is therefore a real decision, and it is worth telling the panel in advance — otherwise a participant does not know whether there is still time to come back.

The discussion of a closed study remains readable.

What an RTD study does not have

These belong to multi-round variations and are not worth looking for in an RTD study:

• round comparison and movement analysis
• a pause between rounds
• a notification that a new round has opened
• a round number in the phase bar
• retiring a thesis — that belongs to Barometer Delphi only, and the server forbids it elsewhere

If you find yourself needing one of these, the variation is the wrong one. That is not a fault but a choice, and it can be made differently in the next study.

Orientation, variation and question architecture

Planning (DPE)

Planning determines what Pronoia collects and what DAE can interpret. The quality of the study is decided here — two locked starting choices (orientation and variation) plus the scoping of the phenomenon and theses shape everything downstream.

Orientation — from which standpoint the future is studied

Orientation is the study's locked starting choice. It affects how theses are framed, how the panel is composed, and how the facilitator acts between rounds (in Pronoia it also drives the PIRE round-gap assistant).

Orientation constrains methodological choices. An R study does not use convergent "central-tendency only" feedback, because that would erase the very deviance it aims to protect. In an N study consensus is a natural target. Orientation also sets the panel type: N/S/R/D → expert matrix, E → observer matrix (see Building the panel).

Variation — the cadence of rounds

Alongside orientation, a variation is chosen that sets how rounds and feedback are paced. In Pronoia this is delphi_mode:

Variation and orientation are different axes: orientation gives the standpoint, variation the cadence. The same normative study can be run as any of the four. Round 0 (orientation) belongs to every variation.

Scoping the phenomenon — what is actually studied

Before the theses, the phenomenon is defined: the boundary of the inquiry. This is planning's most critical scoping step — it decides what is in and what is out. An optional pre-analysis sharpens it:

• Futures Wheel — an impact network: first-, second- and third-order consequences (STEEP+). Maps what follows from the phenomenon.
• Relevance Tree — a goals–means hierarchy: priorities and success criteria. Maps what the phenomenon aims at.

Pre-analysis produces a scoped, structured view of the phenomenon that feeds directly into panel and thesis design.

Question architecture — the theses

The theses are the collection instrument: every argument is one panel member's response to one thesis, so the set of arguments is theses × panel members. Thesis structure follows the orientation:

• Dialectical theses (N / S / R / D) — a claim taken a position on by probability and desirability; dialogue produces for/against arguments.
• Abductive theses (E) — more open and exploratory: what is emerging, what it looks like.

A single thesis's logic can be tuned with a P/D tag: probability (P) leans to consensus, desirability (D) allows dissensus — overriding the orientation default for that thesis. There are at least three theses.

Planning ends at a quality gate. Only a coherent design — orientation, phenomenon, panel and theses aligned — is packaged as the Planning Handoff Package and crosses into Pronoia for collection. The design is not rebuilt in Pronoia but continued (see Integration guide).

When there is no real panel — simulation

If no real panel exists yet, the Argument Simulator (Route B) generates synthetic, orientation-aware arguments for the theses — a test or seed dataset that can be carried into DAE. Simulation tests the theses before a real panel; it does not replace one.

Sources: help/02-orientaatiot, planning-core (M1a). Variations correspond to Pronoia's delphi_mode states.

Building the panel: epistemic space, perspective and gates

Planning (DPE)

A Pronoia panel is not a random set of experts. It is a designed epistemic space whose reach and gaps are deliberate and documented. Panel composition is a methodological decision, not a recruitment convenience — because the panel defines the entire argument space, its coverage and blind spots are known before any data exists.

Three working principles:

• A homogeneous panel (same education, sector, culture) is an epistemic risk and is recorded as such — even if every member is individually excellent.
• The target is commensurability — enough shared ground for genuine dialogue — not full consensus. Disagreement that is well covered is a feature, not a defect.
• Watch for lock-in: a single perspective claiming universality. It can masquerade as dialogue simply because it is loud.

Orientation sets the panel type

Emergent studies are about where you stand to watch, not credentials — which is why decision-makers are excluded and observation distance replaces expertise as the selection basis.

The perspective layer: position + view

The optional but recommended perspective layer rests on a simple idea: perspective = position + view. The same phenomenon genuinely looks different from different positions — not because anyone is wrong, but because the phenomenon is multidimensional. This is not relativism: the aim is to identify what each position can see and what it necessarily cannot.

When the layer is active, each role carries three attributes:

Blind-spot analysis. The facilitator checks which positions identified in the phenomenon map are systematically absent from the panel matrix. The key distinction: a blind spot is a systematic absence, not a random gap. Naming blind spots before round 1 is part of designing the space, and they become candidates for supplementation later.

Perspective Health Score (PHS)

After the panel is designed, Pronoia summarises its diversity as a Perspective Health Score on a 0–12 scale, built from four dimensions (position diversity, epistemic diversity, value-basis variation, marginal voices — each 0–3):

Epistemic coverage: 7 × 3 = 21 cells

Arguments are read on two dimensions — where the claim draws from (knowledge base) and what it does to the thesis (function):

Pure uncertainty ("perhaps", "it depends") is not a knowledge base but a modality of the claim, and it is marked with the flag MOD (formerly EPV). Why the classes are drawn as they are, and where their boundaries lie: the method library article Argument analysis in Pronoia: the epistemic 7×3+COND.

Seven knowledge bases × three functions = a 21-cell map of the argument space. Empty cells are the panel's blind spots — viewpoints that never appeared. The Epistemic Coverage Index (ECI) is simply how many of the 21 cells are filled. Earlier runs typically filled roughly 9–14/21, but that figure dates from the previous code set and is only indicative: coder agreement on the knowledge-base axis is still low (κ = 0.12), so the ECI is read as a whole, not cell by cell; very low coverage is a signal to add roles that bring the missing viewpoints.

Quality gates before collection

Before a panel is accepted, it must pass orientation-specific gates:

If a gate fails, the panel is not yet ready for collection — the facilitator adjusts composition or theses and re-checks. The gates are guardrails that keep coverage and balance honest before any arguments are gathered.

Building the panel in practice

1. Start from orientation. Confirm the orientation; it decides whether you build an expert or an observer matrix.
2. Build the matrix. Lay out the two axes (e.g. expertise × stakeholder, or observation position × context) and place members so the hard minimums are met.
3. Activate the perspective layer when standpoint matters: give each role its observation position, epistemic type and value basis.
4. Map blind spots against the phenomenon map — name the systematically missing positions rather than treating gaps as random.
5. Read the PHS and act on it: supplement a thin panel before round 2, or add perspective theses where recommended.
6. Mind the coverage. Watch which of the 21 epistemic cells stay empty; low coverage prompts adding roles.
7. Guard against lock-in. Keep checking that one loud perspective is not crowding out the rest; the aim is commensurable dialogue.
8. Clear the gates. Confirm the orientation-specific quality gates before collection — and re-confirm after any panel change.

Source: Pronoia — Building the Panel (Metodix, June 2026).

Simulation: the Argument Simulator (Route B)

Planning (DPE)

When no real panel exists yet, the Argument Simulator (Route B) generates realistic expert and observer arguments for the designed theses. It is orientation-aware (N/S/R/E/D/D+) and produces structured arguments that the DAE pipeline reads (CP-1…CP-20). The purpose is to test the theses and panel logic before a real panel — or to provide material to analyse when no panel is available.

This does not replace a real panel. The arguments are synthetic, GenAI-disclosed drafts that require an audit. The actual interpretation happens in DAE under quality gates — simulation provides source material, not conclusions.

How it proceeds — six phases

The simulator activates once the planning context has passed the QG-SIM-TRANSFER gate (orientation, question structure, panel logic, theses). Then:

1. Individual arguments — one argument per role per thesis. Dialectical (N/S/R/D): thesis + antithesis + synthesis view. Abductive (E/D+): observation → explanation → anticipation.
2. Quick analysis — distributions, the 7×3 epistemic profile (knowledge base × function), conditional asymmetry.
3. Dialogues — 5–8 thesis tensions or observation clusters: position → concessions → synthesis insight.
4. Analysis dimensions — tension architecture, CLA pre-profile (causal layers), the 7×3 epistemic cross-tab, MLP pre-profile (multi-level).
5. Anomaly check & supplementation — an automatic blind-spot check (see below). Where needed, the panel is supplemented with missing viewpoints (marked source: supplement, ≤25 % of the panel).
6. Summary & handoff — the overall result, a keystone candidate and readiness for the DAE pipeline (→ CP-1 panel validation).

Each transition produces a light Learning Note that carries lessons from planning through simulation into DAE.

Anomaly check — surfacing blind spots

After phase 4 the simulator automatically checks whether the material leans unhealthily:

A trigger leads to a supplementation protocol: roles are designed to bring the missing viewpoints and a supplementary simulation is run — the same logic as supplementing a real panel (see Building the panel).

Worked example — Food Paradigm 2050

The Documentation Sprint's worked example shows Route B in practice:

The phenomenon: five competing food-system paradigms (techno-capitalist, distributed bio-economic, fragmentation, multipath cultural-pluralism, post-smallholder), entry tension technological autonomy ↔ environmental sustainability. The example shows how a D orientation propagates coherently from the goal hierarchy to the theses, and how the perspective layer integrates without friction.

Sources: Argument Simulator v2.5 (metodix-planning-simulation); worked example CASE_202605_FOOD-2050-WE-001 (Documentation Sprint).

Building the Delphi panel

The panel defines the entire argument space of a study: every argument is one panel member's response to one thesis, so the arguments a study can ever produce are theses × panel members. Design the panel well and its coverage — and its blind spots — are known before any data exists. This guide is the methodological centre of planning a Delphi study in the ecosystem.

Planning (DPE) chooses orientation and drafts the panel. Facilitation (Pronoia) is where the panel is finalised, validated and gated before collection. The panel is a designed epistemic space, not a recruitment convenience.

The panel as a designed epistemic space

A panel is not a random set of experts; its reach and gaps are deliberate and documented. The guiding principles the facilitator works with:

A homogeneous panel (same education, sector, culture) is an epistemic risk and is recorded as such — even if every member is individually excellent. The target is commensurability — enough shared ground for genuine dialogue — not full consensus. Disagreement that is well covered is a feature, not a defect. And watch for lock-in: a single perspective claiming universality, which can masquerade as dialogue simply because it is loud.

The two-axis matrix is primary

The panel's primary structure is a two-axis matrix. Its cells hold the roles that panelist behavior is grounded in. The facilitator ultimately defines the axes and their classification — the matrix is the design surface, and the templates below are only starting points.

The axes are context-dependent. The canonical choices:

Two properties matter in practice. A cell may hold 0..n panelists — several voices in one role, or none. Empty cells are expected, and a systematically empty cell is a structural blind spot (not a random gap). Naming blind spots before round 1 is part of designing the space; they become candidates for supplementation later.

Orientation sets the panel type

The study orientation chosen in planning (N, S, R, E, D or D+) decides whether the panel is built as an expert matrix or an observer matrix, and which axes, selection logic and minimums apply.

Emergent studies are about where you stand to watch, not credentials — which is why decision-makers are excluded and observation distance replaces expertise as the selection basis.

Perspective: position + view (Hautamäki)

The optional but recommended perspective layer rests on a simple idea: perspective = position + view. The same phenomenon genuinely looks different from different positions — not because anyone is wrong, but because the phenomenon is multidimensional. The aim is to identify what each position can see and what it necessarily cannot.

Following Hautamäki's theory of viewpoints, the deeper point is the multiplicity of perspectives in general — a panel should span the perspective space, not just tick boxes. The familiar four archetypes (constructivist optimist, empirical realist, critical-conservative, eco-social holist) are one useful special case, not the whole space; add or vary perspectives as the phenomenon demands.

When the layer is active, each role carries three attributes that make its standpoint explicit:

The perspective layer complements the matrix cell: the cell sets what a member speaks to (its role); the perspective sets how it reasons.

AI panels (Route B / C)

A study can run wholly or partly with AI panelists — a full-AI perspective panel makes an end-to-end run possible without recruiting humans, which is valuable for testing the whole process and for human-validated AI-analysis development (AI proposes → human validates → κ/calibration → improve).

An AI panelist sits in a matrix cell (its role) and carries an explicit perspective profile that complements it. Because the AI acts on that description, it must be precise and sufficient — this is enforced by the AI Panelist Description Standard (docs/ai_panelist_description_standard.md): each AI panelist needs a filled cell (expertise × stakeholder), a distinct perspective label, the four Hautamäki dimensions (ontological / epistemological / axiological / temporal) as real descriptions, perspective_axes in [0,1], a bias direction and the model used; across the panel the labels are unique and the perspectives diverse. Route B/C also requires a GenAI disclosure, and humans stay in the loop: every AI response is reviewed before a round closes.

Reading the arguments: the epistemic lens

As arguments arrive, Pronoia reads each one on two dimensions — where the claim draws from (its knowledge base) and what it does to the thesis (its function).

Knowledge base — seven categories: EMP empirical (observed data, measurement or a systematic research result) · KOK experiential (first-hand professional or lived experience) · NOR normative (a value, right or norm) · TEO theoretical (a model, theory or conceptual causal mechanism) · INT institutional (a law, policy or institution) · IMA imaginative (a scenario, thought experiment or imagined construct) · ANA analogical (an analogy, precedent or cross-country comparison). Pure uncertainty ("perhaps", "it depends") is not a knowledge base but a modality of the claim, and it is marked with the flag MOD (formerly EPV).

Function — three categories: TUK supporting (reinforces the thesis) · HAS challenging (questions, weakens, refutes) · AVA opening (introduces a new viewpoint, widens the frame).

Seven knowledge bases × three functions give a 21-cell map of the argument space. Empty cells are the panel's blind spots — viewpoints that never appeared. The Epistemic Coverage Index is how many of the 21 cells are filled; earlier runs typically filled roughly 9–14 of 21, a figure from the previous code set and only indicative (coder agreement on the knowledge-base axis is still low, κ = 0.12), and very low coverage signals a need to add roles that bring the missing viewpoints.

Why the classes are drawn as they are, and where their boundaries lie: the method library article Argument analysis in Pronoia: the epistemic 7×3+COND.

Panel diagnostics (coverage, PDI, lock-in, PHS)

Description quality is per panelist; coverage is a property of the whole panel. The ecosystem computes it deterministically — the computable core of MaxPerspective (M5), aligned with Hautamäki's viewpoint relativism, available live in the new-study wizard and as a tool over any panel:

• Matrix coverage — filled vs empty cells (Absent cell = structural blind spot) and fill rate.
• Dominance & diversity per axis — aspect dominance AD% per category, the dominance class (DOMINANT >50% / STRONG / MODERATE / WEAK / TRACE / ABSENT), and the Perspective Diversity Index PDI = 1 − Σ(ADᵢ/100)² (Herfindahl). Lock-in is flagged when one category exceeds 50%.
• Structural PHS (0–50) — PDI×25 + (1 − maxAD/100)×25. The full Perspective Health Score also weighs awareness and openness, which are qualitative (the six-phase MaxPerspective analysis) and not inferable from a panel spec.
• AI perspective spread — over perspective_axes, the spread of the AI panel across the perspective space; low spread is perspective lock-in even when the matrix looks covered.
• Supplement recommendations — which categories to fill (Absent blind spots), where to reduce concentration, and how to broaden perspective spread.

As a fast human-facing summary the facilitator may also record a PHS quick score (0–12) from four diversity dimensions — position diversity, epistemic diversity, value-basis variation and marginal voices, each 0–3: 0–5 diversity too thin (supplementation required) · 6–8 adequate with gaps (Hautamäki perspective theses recommended) · 9–12 excellent diversity.

Quality gates before collection begins

Before a panel is accepted it must pass orientation-specific gates that turn the principles above into checkable minimums:

If a gate fails, the panel is not yet ready — the facilitator adjusts composition or theses and re-checks. The gates keep coverage and balance honest before any arguments are gathered.

Building the panel in practice

Ready-to-run templates give a validated starting point for each panel type — Classic Delphi, Argument Delphi, Barometer Delphi, Participatory civic, Real-Time, Corporate strategy and a full-AI perspective panel — each a contract-conformant package you instantiate in one step (from the study wizard or by importing a planning package) and then edit.

A working sequence for the facilitator: pick a template (or start blank) → set the two axes for your phenomenon → place roles in the cells (leaving deliberate blanks) → for AI panels, write each perspective profile to the standard → read the live diagnostics (PDI, lock-in, blind spots, structural PHS, AI spread) and act on the recommendations → pass the quality gates → begin collection.

The panel is where a Delphi study's epistemic reach is decided. Everything downstream — the arguments, the analysis, the scenarios — is bounded by the space you design here.

The iterative process: rounds, phases and convergence

Facilitation (Pronoia)

Why Delphi iterates

Delphi is not a one-shot survey. Its defining mechanism is iteration with controlled feedback: experts answer independently, then receive an anonymized picture of the whole panel's answers and reasoning, and reconsider in that light. Repeating this loop is what separates Delphi from a poll.

The iteration does specific work: it lets reasoning surface and be exchanged rather than just tallied; it gives people a structured chance to revise in response to arguments they had not considered; and across rounds it either moves the panel toward consensus or reveals a stable, well-understood dissensus — which is an equally valuable result.

The core sequence is: independent judgement → feedback → reconsideration. Anonymity and the staged release of feedback exist to protect that sequence from social pressure — bandwagon effects, dominance and anchoring — so that movement between rounds reflects better reasoning, not the loudest voice.

The Pronoia process at a glance

A study advances through a small state machine that the facilitator drives one step at a time:

Draft → Round 0 → Round 1 → Round 2… → Close → Analysis

Round 0 is the orientation, and it belongs to every variation; every numbered round runs the same internal sequence of phases; and the loop repeats until the facilitator closes the study and hands a clean dataset to analysis. The variation's own character begins when Round 1 opens: Real-Time Delphi runs one round in which responding and dialogue are open at the same time, and Barometer Delphi rates the same theses again from one wave to the next. Classic Delphi and Argument Delphi run the same process; they differ in the aim (consensus / dissensus) and in the indicators, not in the phases.

Round 0 — mapping and motivating (AI-assisted intake)

Round 0 is more than a setup screen. In Pronoia it is run as an AI-assisted chat dialogue that does two jobs at once:

• Motivating. The dialogue meets people where they are, explains what the study asks of them and why their view matters — so they arrive at Round 1 invested rather than cold.
• Mapping. It elicits how each participant frames the phenomenon — the issues, positions and tensions they see — which scopes the argument space and feeds directly into thesis design and the panel matrix.

Selective / key-informant variant. The dialogue does not have to be run with the whole panel. The facilitator can run it selectively with a handful of key informants — people who know the terrain — use their input to map the phenomenon and seed the theses, and then invite the wider panel straight into Round 1.

Whichever way it is run, Round 0 ends by fixing the two things that define the argument space: the panel (built as the orientation-appropriate matrix, cleared through its quality gates) and the theses (at least three, in the structure the orientation calls for). While in this setup state the facilitator can still edit and, where used, regenerate questions; once Round 1 opens, the instrument is locked so everyone answers the same questions.

The phases inside a round

Every numbered round runs the same four phases. Separating them is deliberate: it keeps "form your own view" apart from "engage with others", which is the heart of the Delphi method.

The response and dialogue phases are also called 1a and 1b in the facilitator controls — sub-rounds of the same round.

A phase adds a capability, it never removes one. Opening the dialogue phase does not close answering: a panelist who did not get to answer can still answer during the dialogue, and may take part with the full repertoire for the whole round. The idea of Delphi is that a panelist can change their answer when someone else's reasoning gives a good reason — and that requires answering to stay open.

Controlled feedback: the visibility ladder

What a panelist can see is set per phase, and it is the main lever Pronoia gives the facilitator over anchoring. Three settings climb from fully protected to fully transparent:

Alongside visibility, each phase independently controls whether commenting, replies and the result graph are available — which is how the response phase stays quiet and focused while the dialogue phase opens up the argument.

Policy presets — choosing an iteration stance

Rather than set every phase by hand, the facilitator picks a preset that encodes a methodological stance; individual phases can still be overridden per round.

Preset and variation are different things. The Classic preset is the shared stance of Classic Delphi and Argument Delphi — the same process, a different aim. Real-Time Delphi is a different cadence rather than a different method: instead of discrete rounds, the panel answers and sees feedback continuously for the length of one round, and the facilitator closes when the picture is stable. Barometer Delphi is a monitoring stance: the same instrument is measured again wave by wave, a thesis may be retired between waves and replaced by a new one, and change over time is itself a finding.

Revision and movement

The point of feedback is to make informed movement possible. When a panelist changes an answer during the dialogue or revision phases, Pronoia records it as a revision rather than overwriting silently — so the facilitator can see how much and where the panel moved between rounds.

That movement is read against the panel pulse — the consensus / dissensus picture per thesis. Convergence shows where the panel is settling; persistent dissensus shows where a real, structured disagreement lives. Both are signals for the next decision.

Crucially, the learning movement is not uniform across the panel. Different groups — expertise areas, stakeholder positions or observation distances — typically behave differently: some converge quickly, others hold their ground, others shift only late. Reading movement by group is central to describing how the panel actually learns — and it is also where the most effective nudging happens: simply surfacing that one group has moved while another has not is itself a prompt for the lagging group to reconsider. Pronoia's group comparison along the panel-matrix axes makes these differential movements visible.

The panelist Home — real-time data that fuels iteration

Iteration only works if panelists come back and engage. The panelist Home view is designed to make that happen: it turns the panel's live data into something informative and absorbing, so that seeing the panel move makes the panelist want to move too.

The view adapts to where the panelist is. On the first visit, before they have answered, it stays deliberately minimal — a short "About this study" and a direct route into the questions — so nothing stands between them and their first independent answer. Once they have answered, the layout puts the whole panel first:

See the pulse → Notice movement & tension → Revise → See your effect & learn

Why real-time data fuels iteration. Seeing live consensus form, or a tension open up around one's own position, is what moves a panelist to reconsider and revise. Each revision then changes the picture others see, which prompts further movement — and through that loop the panel, and each panelist, learns.

Tracking the panel — process metrics

Throughout the whole process Pronoia follows the panel with a small set of activity metrics. Together they show, at a glance, whether the panel is alive and engaging:

Revisions and replies are the most telling: they are the traces of panelists actually reconsidering in light of one another — the iteration working as intended.

Iterating across rounds — when to continue, when to close

After a round closes, the facilitator makes one decision: run another round, or stop.

• Run another round when positions are still moving, when dialogue raised new considerations that deserve a fresh answer, or when a thin panel has just been supplemented.
• Close when the panel has converged, or when dissensus is stable and well-understood — further rounds would add cost without insight.

Between rounds the facilitator can also act on panel health: if diversity was too thin, supplement the panel (or add perspective theses) before the next response phase, so the new round genuinely widens coverage rather than repeating the same voices.

Closing hands off to analysis. On close, the collected and validated material becomes the input to the analysis ecosystem (the DAE pipeline and its methods). Iteration in Pronoia ends exactly where structured analysis begins.

The facilitator's iteration playbook

1. Set the stance first. Choose a preset (classic / RTD / open) that matches the stakes, and adjust individual phases only if you have a reason.
2. Open with the intake dialogue. Run the AI-assisted Round 0 chat — with the full panel or a few key informants — to motivate participants and map the phenomenon before the theses are locked.
3. Open Round 0, lock the instrument. Finalise panel and theses, then open Round 1 so everyone answers the same questions.
4. Advance one phase at a time. Move response → dialogue → revision deliberately; let the response phase do its un-anchored work before opening the panel.
5. Keep the panel engaged. Use reminders and messages, and point panelists to their live Home view — iteration only works if people return.
6. Read the pulse before deciding. Look at convergence and dissensus per thesis, and how much positions moved, before choosing to iterate or close.
7. Tend panel health between rounds. Supplement diversity where it was thin so the next round adds coverage.
8. Close cleanly and hand off. Stop when consensus or stable dissensus is reached, lock the study, and pass the validated dataset to analysis.

Source: Pronoia — The Iterative Process (Metodix, June 2026).

Anonymity: identity modes and group reporting

Facilitation (Pronoia)

Principle

The core of the Delphi method is to reduce social-conformity pressure and dominance through (semi-)anonymity. In Pronoia, panelists are strictly anonymous to one another yet identifiable through a pseudonym, so that in dialogue a participant recognises the same conversation partner as the same person. Anonymity protects honest judgement; it must not break accidentally through small groups, AI participants, or comparisons.

Group-level reporting runs along the panel matrix — the expertise (primary) and stakeholder (secondary) axes — rather than a single legacy grouping. The same k≥3 protection applies to every axis and to the panel-pulse group dots.

Three identity modes (per-panel setting)

A. Pseudonymous (default)

A stable word handle per panelist (e.g. Aurora, Vega) — memorable and distinctive, giving no hint of identity or status, stable for the whole study. Real identity sits only behind the invite link (invite token). The link is stored in the database only as a hash, so it cannot be read from the database or from backups. Full functionality: Home, dialogue, revisions, recognition, arena.

B. Role play (optional layer on top of A)

The panelist writes a role or perspective they answer from (e.g. Municipal finance director in 2035) and places it in a box of the panel matrix. Off by default. The division of work is clear: the facilitator owns the structure – the axes and categories of the panel matrix – and the panelist chooses the box their role fits best.

• The role is a name, the box is a group. Group comparison and analysis use the box the panelist chose, not the free role text. Freely written roles therefore do not break the panel into groups of one, and k≥3 protection stays in force.
• Others see the role next to the panelist's reasoning and comments, after the handle, marked self-selected. It is a claimed perspective, not verified expertise.
• AI role. The facilitator can commission a role for an AI panelist in the same matrix. It is shown marked AI role, and the AI answers from the role's perspective. A role a human chose and a role an AI plays are never confused.
• When role play is switched off, roles are shown to no one; the saved role and box are kept.

C. Fully anonymous one-shot poll

No identity, a single response – a seminar or mobile poll the facilitator shares as a QR code or link. No sign-in, own page, arena, revisions, recognition or persistent handle. Voters answer the theses and may write a short reason; the writer is never named. Each submission is stored as an anonymous participant who does not appear on the panel roster or in group comparison, and the answers go into the round's results and distribution like any other answers.

The duplicate-vote guard is per browser: the same browser answers once. It stops accidental repeats but not deliberate ones – another browser or a private window can answer again. The one-shot poll therefore suits taking the temperature of a room, not a vote whose outcome is binding.

Cross-cutting rules (always on, in every mode)

AI is always disclosed. Every AI panelist is marked with "AI" plus a short role description, 2–3 words (e.g. AI · Economist, AI · Labour Researcher). Shown everywhere an AI panelist appears: dialogue arguments and comments, panel lists, pulse group labels, reports. Never hidden, never presented as human.

k-anonymity (k = 3) across all group-level views. No sub-group with fewer than three people is reported at group level — comparisons, sentiment and panel-pulse group dots included — and the rule applies independently on each matrix axis. Complementary protection: when a group falls below the threshold it is folded (into an "Other" group) until every cell has at least three; if "Other" is still below three, the breakdown is not shown at all. Panelists with no value on an axis are excluded from that axis — they never form a residual bucket. The same threshold applies to cross-tabulations (group × thesis): a numeric cell is hidden when n < 3. On small panels the facilitator is steered to a 2–3 group split; anonymity wins. In the extreme, the pulse is shown at whole-panel level only.

Visibility & the after-submit reveal. Per-round phase policy controls when a panelist sees others. The default classic preset uses after-submit for the response phase: no anchoring before you commit your own answer, but once submitted the panelist sees others' anonymized responses and the aggregate graph. Strict blind remains available in the facilitator policy editor. All revealed material stays pseudonymous and k≥3-protected; the reveal changes timing, not the identity rules.

What is available in each mode

*Unless blind-facilitator mode is enabled.

By default the facilitator sees real identities (for reminders and quality control); a per-panel blind-facilitator mode hides them, so the facilitator also sees only pseudonyms (the invite mapping still exists in the database). Blind mode applies to the whole team, including the owner who switched it on. It can be lifted only before any human panelist has been invited or has answered; after that it is locked, because panelists take part under its promise. Switching it on and lifting it are both logged.

A response belongs to the panelist. Only the panelist can change their own response. The study team cannot write a response on a human panelist's behalf, and resetting responses is an owner-only action that removes human responses only in a test copy. Every reset is logged. What is recorded in the log, what is not, and what the service administrator has committed to is described on the Logging policy page.

Group comparison — current behaviour

• Two axes, stacked. Both the panelist Home and the facilitator Communication view show the expertise (primary) comparison and, beneath it, the stakeholder (secondary) comparison — each with its own roster, numeric matrix and sentiment.
• Numeric comparison where the question type allows. Group means and their deviation from the panel mean (plus a divergence figure) are computed for scale questions only; other question types do not produce a numeric matrix.
• Comment-based sentiment for every question type. Each group's tone (positive / cautious / critical / neutral, with a net figure) is derived from arguments and comments across all theses, so open, ranking and multiple-choice questions are covered too.
• Pulse group dots follow the primary (expertise) axis, matching the group-comparison default.
• All of the above remain under k≥3 with complementary suppression; no "Out" / "Other" buckets appear for a fully classified matrix panel.

In practice: role play and the one-shot poll

Each has its own working guide:

• Role play: a role in a box of the panel matrix: when role play suits a study, the facilitator's steps, a template for the panel, AI roles, and reading results by box.
• One-shot poll at a seminar: QR code and anonymous answers: preparation, the QR code and presentation view, what to tell the room, disruptions, results and limits.

Data model (implemented) — for the developer-partner

• Study: facilitator_blind, role_play_enabled, handle_theme, one-shot fields (oneshot_comment_window, oneshot_moderated), policy_preset + per-round phase_policy (visibility = blind | after_submit | open), panel_matrix (JSON: primary/secondary axis names and categories), matrix_type (expert | observer), language.
• Panelist: display_handle (themed, panel-unique, stable), self_selected_role, participant_type (human | ai_agent), matrix_primary / matrix_secondary, description, is_decision_maker, language.
• Axis grouping (_panelist_axis): primary → matrix_primary (AI perspective as fallback; in role play the box the panelist chose – the role is a name, not a group); secondary → matrix_secondary; returns empty when the axis is empty, so the panelist is excluded rather than bucketed.
• Shared group reporting (grouped_display): k = 3 + complementary suppression ("Other") + None-exclusion + small-panel fallback to whole-panel level. Used by /groups (Home) and /community (facilitator) with both axes stacked, and by the pulse group dots.
• AI label derived from participant_type = ai_agent plus the ai_profile role, rendered as "AI · {role}", role standardised to 2–3 words.
• Blind facilitator: when facilitator_blind = true the UI shows only handles (names/emails hidden), though the invite mapping still exists in the database. Lifting it is refused (409 blind_locked) once any human panelist has been invited or has answered.
• Invite link: invite_tokens.token = sha256(link); the link is sealed with a server key into the token_sealed column, and the key is in neither the database nor the backups.
• Resetting responses: owner only; human responses only in a test copy (otherwise 409 human_responses_protected); every reset is logged (responses_reset).

Source: Pronoia — Anonymity Design (Metodix, June 2026, status: implemented).

How Delphi Pronoia's two interfaces are built

Pronoia has two faces. A panelist sees one; a facilitator sees the other. They are not two applications and not two versions of the same screen. They are two questions asked of one running study, and the whole design follows from the fact that the two questions are different.

This guide explains the idea and the structure. The two guides beside it walk through the process — what actually happens, in order, with what on the screen.

1. One study, one truth

Both interfaces read the same two facts:

Everything visible is derived from those two. There is a third field, Study.status, and it is a presentation — a label derived from the round and the phase. Nothing decides anything from it.

This sounds like an implementation detail. It is the reason the interfaces are trustworthy, and it was learned the hard way. Before August 2026 the question "which round are we on" was answered by five surfaces reading three different sources, one of which recognised rounds by matching the strings round1, round2, round3 — so in a four-round study, round 4 never highlighted anywhere. Two surfaces could disagree for fifteen seconds because one polled and the other did not.

The rule that replaced it: round and phase are the truth; anything else is a label. When you see the same fact in two places in Pronoia now — the round selector and a banner, say — they are two readings of one computation, not two computations.

2. The panelist interface

2.1 The idea

The panelist's question is "what is open for me now?" — not "what is the state of the study". The view answers that question directly and organises itself around the answer.

Five principles govern it:

P1 · The phase organises the view, not the menu. What you land on is what is open for you. The facilitator's switches decide exceptions, not the normal case.

P2 · The panel's process is always visible; the panel's content only after you have answered. "27 of 40 have answered", "two new comments since your last visit" — these say nothing about what the panel thinks, so they cannot anchor your judgement. The distribution, the average and other people's arguments do, so they wait until you have answered. The condition is per thesis, not per page: answer this thesis, see this thesis's distribution.

P3 · The frame never changes; the main column changes. Same heading, same menu button in the same place, the form always in the same spot. Not different pages — the same page in different states. A returning panelist needs recognisability more than a per-phase optimum.

P4 · An empty service is not in the menu. Every menu row is a promise that something is there. A "Group comparison" that opens empty is worse than no row. Emptiness is computed from the data, not switched on by hand.

P5 · The description and the notices are not extras. The study description sits above the form on first answering; the privacy and AI notices are always reachable. These are part of what the panelist consents to — not content that has to earn its place by being interesting.

2.2 The structure: six states

The panelist view has six states, defined by the panelist's own question:

S4 is the richest moment of the whole process, and it is easy to waste. It is where a panelist sees what the round produced and forms the view they will bring to the next one.

2.3 What is on the screen

The frame is constant: title, Panel pulse (how far the round has got), the menu, and the main column that changes with the state.

• Panel pulse shows process, never content: how many have answered, what has changed since your last visit, whether the round is live.
• More services is an index, not a drawer of extras — and by P4 it only lists what actually exists for this study.
• AI panelists are marked wherever they have an effect. If the panel includes AI participants, the panelist is told so, and AI contributions are marked in the distribution as well as in the text. This is not a footnote in the terms; it is a mark on every surface where an AI answer counts.

2.4 Two ways to name the journey

The same structure can be shown in two vocabularies, chosen per study:

This is naming, not structure: the states, the rules and the surfaces are identical. A study that wants the Delphic imagery gets it; a municipal panel that would find it strange gets plain words.

3. The facilitator interface

3.1 The idea

It is tempting to mirror the panelist view: the study is in the dialogue phase, so show the facilitator the dialogue phase. That is wrong, and the reason is the whole design.

The facilitator's phase is not the study's phase. The panelist asks "what is open for me"; the facilitator asks "what is my work right now", and the two do not coincide. While the response phase is open, the study's phase is response — but the facilitator's work is monitoring, and preparing the next round. A straight mirror would give the facilitator a screen that says "wait".

So the unit is not the phase. It is (round, phase) → the facilitator's task, and there are four kinds of task:

Plus one thing that is not a place at all: intervention — advance, open, close, handle an exception. It is never its own screen; it is always present where it applies.

3.2 The structure: two levels and a bar

Both levels are at the top of the screen, in one navigation that stays in place while you scroll. A choice always brings the work directly beneath it, so there is nothing to look for. Only "What now?" sits above the navigation, and it scrolls away once the work begins (3.7).

Level 1 — the kind of work. Four buttons, always the same four:

Planning · Collection · Analysis · Reporting

Level 2 — where the work happens. Tabs that belong to the current kind of work:

Three newer blocks live inside these: Try as panelist (the preview) is on the response-link card in Collection, the Timeline — reasoning and dialogue is in Collection's Responses, and the pick basket Thesis material for the next round is in Analysis above the thesis list — and opens beside the thesis editor in Design when a new thesis is written.

The bottom bar — always available. Nine things that belong to no round and no kind of work:

Settings · Responses · Panelists · Theses · Communication · Team · Results · Export · Observation

Team leads to Team & access and Observation to Share an observation; the full name also shows when you point at the button. Your role and balance are at the right end of the bar in short form; the full text shows when you point at them.

Method library is in the left sidebar (📖 when collapsed), and Summary under Analysis.

A door carries its destination's name: Responses leads to the tab of that name in Collection. One place, one name — even when a place has two doors (decision K2).

Two of these deserve a note.

• Export appears twice on purpose. Reporting is a phase — what you do at the end. Export is a function — something you can do at any moment. Same view, two different questions, two different moments.
• Results is the only item that is not a tab. It opens a Results so far window over your current work. Glancing at an earlier round must not take you away from what you are doing, so it is a window and not a destination. It contains no editing surface at all; there is exactly one way onward — Open in Analysis → — and all editing happens there.

Collection's level 2 is a process. The order is the order in which the work is done: Situation tells you where things stand, Invitations is the first act, Facilitation gathers the round's settings (deadline, dialogue phase, visibility and interaction), Communication holds the messages to the panel — the same view as Communication in the bottom bar — and Responses is the per-thesis result. Round 0 comes first when the study runs it, and Waves last in the Barometer.

Collection shows the open round. There is no way back into a closed round: its material is read in Analysis, and the server refuses to change a closed round's dialogue, phase rules or schedule. Settings decided before a round begins — opening time and adding a round — are in Planning's Rounds section, and exceptions — changing the phase by hand and resetting an old round — in Reporting's Maintenance.

The bottom bar also shows your role and AI balance. They are information, not doors: they lead nowhere, but they are always visible.

3.3 Default, not compulsion

The phase decides where you land and what gets weight — it does not decide where you may go. The rail stays free, because in a Delphi study you return to planning in the middle of collection: round 2's theses are built while round 1 is still running.

When you are looking somewhere other than where the study is, the view says so. It does not stop you and it does not hide anything; it just refuses to let you believe you are looking at the current round when you are not.

3.4 "Applies to:" — the marking that prevents the worst mistake

Planning contains five things, and they are not all the same size:

So when you are preparing round 3, you must know whether you are editing round 3's theses or also round 1's. Every planning view carries a marking that says which: Applies to: the study / Applies to: round 3. Round-scoped editors take the round as an explicit parameter — never "whatever is current".

3.5 AI actions are in the background, behind a control

If the panel includes AI panelists, the Collection view shows a card that says how many there are. The actions — generating responses, comments, reactions, revisions — are behind AI actions ▾.

The reason is not tidiness. A block of AI operations rendered straight into the main column grows the very column the whole design is trying to shorten, and it puts machine operations in front of the work that is actually the facilitator's at that moment: watching a human panel answer.

If the actions are not available, the card says why — the first round has not been opened, round 0's AI sessions are generated from the Round 0 view, or the study has ended. An absent surface that gives no reason is indistinguishable from a broken one.

3.6 One panel, one list

The panel is a single list in which AI participants are marked — participant_type is a column on a panelist, not a different kind of object. But a default view listing every panelist would, in several existing studies, be data rather than a panel: measured on 25.8.2026, the database held 80 AI panelists and 33 people.

So the panel view has a filter — SHOW: All · People · AI — which is view state and changes nothing. The row always shows all three counts, so what is hidden is readable rather than inferred.

3.7 "What now?" — the card that also says why

The status card's top row names where you are in the arc — Planning, Round 1 running, Pause after round 1 — prepare round 2, Ending. Below it are the next 2–3 steps, and the order is itself information:

blocker before gap · gap before transition · transition before this phase's work

Each step carries three things. The reason says where the prompt comes from (blocked · missing · next phase · this phase's work); a blocking reason is amber. The door is a button that takes you there and scrolls the control into view. The rationale is a small ? that opens the method library at the right article — the rationale is a door, not a sentence written into the card.

The card blocks nothing. When nothing is required, it shows this phase's work: empty is an answer, not an empty card.

A blocker is visible before you hit it. The card asks the gates in advance, so it never prompts you past what would stop you — "Choose the instrument for round 2" appears before you press forward, not afterwards as an error.

The card is at the very top of the screen, above the navigation. You can collapse it when the space is needed for work — but even collapsed, the card shows the first step if that step is a blocker.

3.8 The phase narrows what you see

The Collection view has a row, Informing and interaction, below the AI runs: reminder, message to the panel, invitations. It is not a new view but doors into existing actions, and its contents follow the phase — in the response phase reminder · message · invitations, in dialogue only message, in the pause message · pause notice.

For the same reason some sections fold by phase: what does not belong to the work at hand is a chip rather than a block. Folding is a default, not a rule — your own choice wins, and a folded section leaves the chip that opens it.

3.9 The contents row — a view's sections on one line

Views with several sections (Responses, Communication, Summary, Statistics & change, Waves) have a CONTENTS row at the top. Its chips are the view's sections, and a chip's name is a door: it takes you to the section and opens it if the phase has folded it (3.8). The small × after a chip hides the section. A dashed, faded chip means the section has no content in this study yet — it leads nowhere, because there is nothing there.

For example, Analysis's Statistics & change view holds seven analyses — thesis stats, convergence, key figures, dialogicity, shift directions, revisions and refresh — and Summary holds, among others, highlights and the epistemic 7×3. The contents row is their index.

3.10 Mail goes out as a letter

Every Pronoia mail goes out in two versions: as plain text and as a letter. The mail program shows the letter; a program that does not display HTML shows the same text as before. You still write only the text — the letter is built around it: the study's name at the top, [link] turns into a button where you wrote it, and at the end the contact person and a note that the recipient can reply directly. The letter has no images and no tracking: you will not learn when a panelist opened the message, and it was never promised to them.

4. The same round, from both sides

Read that table downwards and the two designs are one design: the panelist gets content when it can no longer anchor them; the facilitator gets it immediately, because their job is to see it.

5. What the interfaces deliberately do not do

Five rules that explain absences you might otherwise read as gaps:

1. No empty service. A menu row that opens onto nothing is removed, not left as a promise.
2. No second computation of the same fact. There may be several buttons that advance a round; there is one computation of what "advance" means. Two buttons are fine, two answers are not.
3. AI is marked on every surface where it has an effect — not disclosed once and then forgotten.
4. A closed round stays closed. Its participants and its results are history. Changing them afterwards would silently change a response rate that has already been shown and possibly analysed, and comparison between rounds is the core of the method.
5. Pronoia describes; it does not interpret. Distributions, movement between rounds, who said what — those are description. Interpretation is the analysis ecosystem's job, and the boundary is deliberate.

Written 25.8.2026, against the interfaces as rebuilt on 23.–25.8.2026. Button and tab names are taken from the application's own English strings. The specifications behind this text are docs/UI_Panelistin_nakyma_IA_v2.md and docs/UI_Vaihearkkitehtuuri_IA_v2.md; where they and this guide disagree, they are right and this is stale.

The panelist's process, round by round

This guide walks the panel from the invitation to the final results. It is written for the person who has been invited — you is the panelist — and the facilitator who prepares the panel can read it as a description of what their panelists will actually see.

The companion guide, How Delphi Pronoia's two interfaces are built, explains the design. This one is the walk: what is on the screen at each point, in order, and why some things are not there yet.

1. What you are joining

A Delphi panel is not a survey. A survey asks you once and adds you up. A Delphi asks you, shows you what the rest of the panel said and why, and asks again — and the interesting part is what happens in between.

Three things follow from that, and they are worth knowing before the first screen:

• Your reasoning matters more than your number. The rating locates you on a scale; the reasoning is what other panelists read and what the analysis finally works from. A rating without reasoning is half an answer.
• You are anonymous to the other panelists. They see you as a code, not a name: "You respond anonymously — other panelists see you only as {code}". The facilitator's visibility into who you are depends on how the study was set up, and the privacy notice says which arrangement applies.
• Disagreement is not a failure of the panel. It is often the finding. Nothing in the interface pushes you toward the middle, and the results screens show spread as a fact rather than a problem.

If the panel includes AI panelists, you are told so before you start. "This panel includes AI panelists" is stated up front, and every AI contribution is marked wherever it has an effect: beside the argument, and as a hatched share inside the distribution.

2. Before anything opens: the invitation and the consent screen

You arrive by a personal link. It is yours; it identifies you to the study without a password, which is why it should not be forwarded.

The first screen is Before you begin. It states what the study is, who is running it, how your data is handled, and — if the panel includes them — that some panelists are AI. You continue with I have read this and I take part, and Read the privacy notice stays reachable from the menu for the whole study.

This screen is not a formality that the design tries to get past. It is the thing you are agreeing to, so it comes before the content rather than under it.

3. Round 0: the orientation conversation

Most studies begin with Round 0 — a conversation rather than a form. An AI assistant called Kastalia opens the phenomenon with you in your own words before any thesis is rated.

Kastalia says what it is, in its own words, at the start:

"I am an AI, not a person. I help you open up the questions in your own words before the round itself — I don't assess your answers and I won't tell you which view is right. I carry the data; the seeing is yours."

What is on the screen: the phenomenon under study, what the panel design already knows about your angle ("From the panel design, Pronoia knows about you" — correct it in the chat if it is wrong), and a progress line, "Answered {answered}/{total} orientation questions". You write freely; there is no scale.

You end it yourself with I'm done →. Then you land on your own home view, and the theses open later.

Round 0 is skipped in some studies. If it is, your first screen is the one in §5.

4. The frame that never changes

From here on, one thing is worth internalising because it makes everything else easy: the page does not change, its middle does. The heading is where it was, the menu button is where it was, the form is in the same place every round.

Three parts of the frame are always there:

Panel pulse shows process, never content. It tells you that the panel is moving — how many have answered, how many new comments there are — and it deliberately does not tell you what the panel thinks until you have answered. There is no target number in it: it says "27 have answered", not "27 of 40". A target turns a sign of life into a scoreboard, and a scoreboard tells a late respondent that they are late.

5. Your turn to answer

This is the state you were invited for. The main column holds the theses of the current round, and for each one: the thesis, the scale, and a field for your reasoning — "Explain your reasoning:".

Answering, in practice

• The rating and the reasoning travel together. The placeholder says it plainly: "Why do you rate this way?" Some theses are open-ended and have no scale at all; some have two criteria (for example probability and desirability), and then you answer both — they are different questions about the same thesis.
• Some theses are not scales. Depending on the study you may be asked to rank options, choose one or several, place a point on a plane, distribute points, mark a time window ("Earliest" / "Latest", or "Will never happen"), or continue a time series. The pattern is the same in all of them: your position, then your reasoning.
• One at a time, or all at once. One at a time shows a single thesis with "Thesis {i}/{n}" and Previous / Next; Show all returns the full list. This is a reading preference, and it changes nothing else.
• Your work is kept while you write. A local draft is saved as you type and the form says "Draft saved on this device". If you leave and come back, "Unsaved draft found" offers Restore draft or Discard. The draft lives in your own browser, not in the study.
• Saving. Each thesis has Save; Save all ({n}/{total}) submits the ones you have completed. When every thesis has an answer, the view says "Thank you for your answers!"

What you do not see yet, and why. During the response phase the panel's content — the distribution, the average, other people's arguments — is not shown before you have answered. Not because it is secret, but because seeing it first changes what you write. A first answer that has already been anchored to the panel's average is worth less to the study than a first answer that is yours. The condition is per thesis, not per page: answer this thesis, see this thesis's distribution.

Studies can be stricter or looser about this. The round's visibility setting is one of Blind, After you answer, or Open, and the usual setting for a response phase is After you answer.

6. Answered, and the round is still open

Your answer is in. Two things are now true.

You can still change it. "Your answers to all {total} theses have been saved to the study. There is nothing more to send, and you can close this page. You can come back to edit your answers while the round is open." — Edit answers reopens the form. Nothing is locked until the round closes — not even when the study moves into the dialogue phase: a thesis you did not get to answer can still be answered during the dialogue.

The theses you answered begin to show the panel. Under a thesis you have answered, the distribution appears: bars for the responses, a line for the mean, a dashed line for the median, a band for the interquartile range, and your own answer marked with ★ and an orange bar. Beside it, the numbers the panel uses — Md, SD, IQR, n.

If the panel includes AI panelists, the distribution says so: the AI share is drawn in a second colour and hatched, the legend names both groups — People ({n}) and AI panelists ({n}) — and n is given split, "{h} human, {a} AI". Colour alone is never the carrier of that fact, because colour alone fails for part of the panel.

7. Dialogue: the round where the panel talks to itself

The dialogue phase is what makes this a Delphi. "You now see other panelists' anonymous answers and arguments. Explore perspectives and consider if you want to refine your own answer."

Dialogue is open. Here the panel's content does not wait for your answer — during dialogue it is the thing being participated in, and a gate would shut out exactly the panelist who most needs to be let in. (A study can still choose a blind dialogue round; then the interface says so.)

Each thesis card is in one order, and the order is the argument:

1. the thesis
2. your own answer — so you read the others against what you actually said
3. the distribution — where the panel landed
4. the discussion, behind a link with the numbers in its name — "Discussion · {a} arguments · {c} comments"

The count is in the link on purpose: you can decide whether it is worth opening without opening it.

Inside the discussion

• OTHERS' RATIONALES are the rationales given with the ratings; comments are what the panel said about each other's rationales. They are different things, which is why the link counts them separately.
• Contributions from AI panelists carry the AI marker — the same fact as the hatched share in the chart, from the same field.
• If the study uses role play, the writer's role is shown next to the handle. A role a human wrote is marked self-selected, and a role the facilitator commissioned from an AI AI role. You write your own role and place it in a box on your own page, on the Role play card.
• On your own rationale you do not comment but clarify: the button reads Add a clarification. Others see the clarification under your rationale, marked own clarification, but it is not counted as dialogue – dialogue is what panelists say about each other's rationales.
• You can Comment on a comment too, and react with Insightful, Agree or Challenge. A comment that another panelist has commented on is locked from editing: "Locked — the comment has been commented on". Editing it would silently rewrite what someone else commented on. Refining your own comment yourself does not lock it. If your position changes after the lock, you say so in a new comment — then the change shows as dialogue, not as history edited after the fact.
• Translate renders a contribution in your language when translation is enabled for the panel; Show original puts it back.

Changing your mind is the point. If the round allows revision, you edit your answer and say what moved you — the field asks "Why did you change your mind? (What argument or information influenced you?)" — and you can classify the change: More agreement, More disagreement, Nuance / clarification, Reframing. Then Save revision.

Not changing your mind is equally a result. A held position with a reason that has survived the counter-arguments is stronger evidence than a position nobody challenged.

8. Between rounds — the richest screen in the study

When a round closes, the screen does not say "please wait". It says Round {n} has closed, and then it shows what the round produced:

"Below is where the panel landed this round: distributions, other people's reasoning, and what was left open. This is the round's interim result — not a waiting screen."

This is the moment the whole method is built around, and it is easy to waste. The round's result is in front of you with no obligation attached: nothing to fill in, nothing due. Whatever view you bring into the next round is formed here.

If the next round has been scheduled you are told when — "Round {n} opens on {date}" — and if it has not: "The next round has not been scheduled yet." The interim result does not depend on the schedule; time is an addition, not a condition.

In a Barometer study this is where a wave ends and the next begins; the interim result is the result, and there is no final screen.

9. When the study ends

"The study has ended" — all rounds are closed. The final distributions, the panel material, the discussion and your own path through it stay readable. If you revised your answers along the way, the interface says how often: "You made {count} revisions to your answers based on the dialogue."

What happens after this is outside the panel: interpretation — theming, scenarios, argument and discourse analysis — is done in the analysis ecosystem, and the results reach you as a message rather than as a new screen. Pronoia describes; it does not interpret.

10. More services: the index

More services is an index, not a drawer. It opens a list with one row per service: an icon, a name, one line about what it is, and a mark on the right.

Four things about this list are deliberate:

1. A row only exists if the service exists. Group comparison appears when the study has comparison axes; the signals channel appears when the study collects signals. A row that opens onto nothing is worse than no row.
2. The mark is a fact, not a nudge. An unread message is highlighted because it is waiting for you. Everything else is a count, and an empty service says nothing yet rather than pretending to be full.
3. One service at a time, and one step back — Back to services. Your work stays open behind it: "Pick a service — your work stays open behind this."
4. The privacy and AI notice is always reachable. It is not a service that has to earn its place by being interesting.

Two of these have a note worth reading.

Panel pulse draws one track per measure: a light band for mean ± standard deviation, a mark for the mean coloured by where it sits on the scale (the colour shows position, not whether the result is good), the group means as dots when the study has groups, and the ends of the scale underneath. Consensus is not drawn as geometry anywhere: a consensus figure is a classification that needs a calibrated threshold, and that classification belongs to the analysis, not to a bar.

Group comparison is hidden when the groups are too small: "there are not enough reportable sub-groups of at least 3 people. Results are shown at whole-panel level only." Small groups identify people. The suppression is the promise working, not a fault.

11. Two ways the same journey is named

A study chooses one of two vocabularies, and the switch reads View: with Story and Plain:

This is naming, not structure. The states, the rules and the surfaces are identical in both; a panel that would find the Delphic imagery strange gets plain words. In the Story view, tapping a station explains what that stage means — "Tap a station — I'll tell you what it means".

Wherever you are, You are here marks the current station, and Round {n} / {total} says how far along the panel is.

12. If the study is real-time

In a Real-Time Delphi there is one round, and answering and dialogue are open at once in it from the start. The orientation conversation (§3) may be part of it too. "Other panelists' answers appear continuously. You can change your answer at any time."

The rule from §5 still does the work, thesis by thesis: a thesis shows you the panel as soon as you have answered that thesis. Updates arrive live where the connection allows, and the interface says which mode it is in rather than pretending — "Updated just now", and a tooltip that distinguishes a live connection from fallback polling.

There is no "between rounds" screen in a real-time study, because there is no between. In a Barometer Delphi, by contrast, the same theses are returned to one measurement wave at a time: the gap between rounds is the gap between waves, and a thesis may go out of use or be replaced by a new one during it.

13. Your answers, your data

• Anonymity toward the panel is structural, not a setting you have to find. Other panelists see a code.
• Leaving is possible without erasing the discussion. You ask the study's contact person, and they carry it out — the promise is kept by someone who is able to keep it. Withdrawal deletes your name, email address and personal link permanently; the answers you have already given stay in the material under the pseudonym, because other panelists have read and argued with them. That asymmetry is deliberate, and it is stated here rather than discovered afterwards.
• A closed round stays closed. Answers to it are not edited afterwards — a response rate that has already been shown, and the comparison between rounds, would otherwise change quietly underneath the study.
• The privacy and AI notice is the authority on how your data is handled in this particular study, and it is one click away from every screen.

14. For the facilitator: what is worth saying out loud

Three things panelists ask, that the interface answers but the invitation letter usually does not:

1. "Do I have to finish in one sitting?" No. Drafts survive, answers can be edited while the round is open, and the study waits.
2. "Why can't I see the results yet?" Because you have not answered yet — and the moment you answer that thesis, its distribution opens. It is a rule about anchoring, not about access.
3. "What happens to what I wrote?" It is read by the panel under a pseudonym during dialogue, and by the analysis afterwards. The reasoning is the material; the number is the index to it.

Written 26.8.2026, against the interfaces as rebuilt on 23.–25.8.2026. Button and tab names are taken from the application's own English strings, and a guard test (scripts/test_guides_coverage.py) fails if this guide names a control differently from the program. The specification behind this text is docs/UI_Panelistin_nakyma_IA_v2.md; where it and this guide disagree, it is right and this is stale.

The facilitator's process, from planning to handover

This guide runs a study from the empty draft to the handed-over result. It follows the order the work actually happens in, and it says at each point what is on the screen, what the panel sees at the same moment, and which actions cannot be taken back.

The companion guide, How Delphi Pronoia's two interfaces are built, explains why the facilitator view is organised the way it is. The one-sentence version, because everything below follows from it: your phase is not the study's phase. The panelist asks "what is open for me"; you ask "what is my work right now", and while the response phase is open your work is monitoring — and preparing the next round.

0. The map you are working on

Four kinds of work, always the same four buttons:

Planning · Collection · Analysis · Reporting

Nine things that belong to no round and no phase are in the bottom bar, which is always available:

Settings · Responses · Panelists · Theses · Communication · Team · Results · Export · Observation

Method library is in the left sidebar, and Summary under Analysis.

The names are the screen's own, not the guide's: a bottom-bar item is a door, and a door is named after the place it leads to. Two are shortened so the bar fits on one line: Team leads to Team & access and Observation to Share an observation.

The phase decides where you land and what gets weight. It does not decide where you may go: in a Delphi you return to planning in the middle of collection, because round 2's theses are built while round 1 is still running. The button marked default is where the study says you belong; when you are somewhere else, the view tells you so — "You are looking at {a}. The study is at {b}." — with a Back that returns you. It never hides anything; it only refuses to let you believe you are looking at the current round when you are not.

The navigation is at the top of the screen. The four buttons and the tabs of the chosen kind of work sit in one block that stays in place; a choice brings the work directly beneath it. Collection shows the open round — closed rounds are read in Analysis.

1. Planning: a completeness list, not a wizard

A step-by-step guide with pictures for the same work: Finishing your plan in Pronoia.

Planning is both a sequence (the first time through) and a place (round 3's theses, built mid-study). Pronoia resolves that by showing the first pass as a list of what is missing rather than as numbered steps — Building the study:

"The first round is made of these {n} things. The order is for reading; readiness is computed from what the study contains."

There are now eight, each of which is also the way in. The settings, rounds, panel and theses rows block opening; the rest are recommendations or decisions:

Role play and the anonymous one-shot poll are settings, and each has its own working guide: Role play: a role in a box of the panel matrix and One-shot poll at a seminar: QR code and anonymous answers. Role play needs the panel matrix to be built first.

The list is computed, not ticked off. Readiness comes from what the study contains, and the header states the situation without arithmetic on your part: "Opening is waiting on {n} item(s)", or "Everything that blocks opening is in place." Items that are advice rather than obstacles are marked recommended — a thesis without a scale is open-ended, which is information, not a fault.

Two of the gates are enforced by the server, not by the screen. A study does not leave draft while the AI mode is unconfirmed, and invitations do not exist until the data-controller declaration is complete. The gaps are named as they are: "AI mode not confirmed", "controller name", "controller contact", "legal basis for processing", "consent text still has placeholders". The list reads the server's answer instead of guessing it, so you never meet a gate you were not warned about.

And the gap is closed in the view the list leads to. In Settings, under the AI-mode choice, it says which of the two holds: "The AI mode is confirmed" or "The AI mode is not confirmed — a round cannot be opened until it is". Confirm the AI mode opens the same choice dialog that otherwise appears only after you hit the gate. The confirmation applies to the selected mode: change the mode and it has to be confirmed again, because a change does not carry the old confirmation with it.

Every planning view says what it applies to. "Applies to: the study" or "Applies to: round {n}", and the panel row says the hybrid out loud: "Applies to: the study · participation is per round". When you are preparing round 3 this is the difference between editing round 3's theses and quietly editing round 1's. Round- scoped editors take the round as an explicit parameter — never "whatever is current".

Rounds are not fixed at creation. "The number of rounds is not locked at creation: the length of a Delphi can be a result rather than a precondition." Dialogue is on by default for an added round; a pure feedback round is made by switching it off, deliberately:

"This round has NO dialogue phase — it collects positions but the panel never argues about them."

Building the panel. The panel is one list in which AI participants are marked — SHOW: All · People · AI filters the view and changes nothing. Under it, PANEL COVERAGE reports the Coverage score, the primary and secondary coverage, the cells that are empty, and — when they are — "Panel supplement recommended before the round".

Reserve enough rounds — an unused round costs nothing. The number of rounds is set when the study is created, and it is not a promise you have to keep: the program never requires a reserved round to be run. You may close the study after round two even if you reserved four; the rounds you did not open simply never existed for the panel. The reverse is harder — a round can be added mid-process, but that is the exception (see the round-management card), not the path. So when the number is in doubt, reserve the larger one.

2. Round 0: orientation in seven steps

Round 0 is a single box in Round management, below the path, and the Round 0 tab opens the same box. The box shows seven steps and where you are: Questions, Respondents, Invitation and opening, Conversations, Results, AI analysis and Selection. The steps are a checklist, not a lock: you can open any of them at any time, and a step whose prerequisites are missing says what it is missing. Below the box is a line about the next round.

The AI budget has its own line at the top of the box. Round 0 uses AI in three places – the AI panelists' conversations, Kastalia's replies to people and the analyses – and the price difference between models is large. Before the first AI run you give round 0 a credit cap, and the window shows an estimate for the whole round with three models (Haiku, Sonnet, Opus), itemised. The suggestion is the highest-quality model whose estimate fits the cap; you can choose another and confirm. All round 0 AI calls use the confirmed model. When the cap is reached, AI conversations and analyses stop, but human conversations continue on the cheapest model so that no respondent is left mid-conversation. The line shows what has been spent and what is left, and Change opens the window again.

• Questions sets the orientation questions and the purpose of each one. The purpose is shown to the panelist in the conversation under "Why do we ask this?". The default questions are in the study's language, and they are visible before you take them into use: Use these questions. The conversation settings – the opening text shown to the panelist, the AI assistant's deepening task, the number of follow-ups and the word limit – are under the same step.
• Respondents is a choice of its own, made before opening. Whole panel is the default; Selected key informants narrows the round to the ones you tick. The scope covers AI panelists as well as humans, and the same choice drives the AI run, human access to the conversation and the cost estimate. The scope is an explicit choice, not derived from the ticks: one tick must not quietly shut the rest of the panel out.

From then on the scoped number appears everywhere round 0 is discussed — readiness, the steps of the box and the invitation card — beside the panel size. In the collect view the scope has a line of its own, and Who round 0 is for → leads back to the choice.

• Invitation and opening: Open Round 0 publishes the round. From this moment the panelists' links lead somewhere. A human panelist needs a personal link, and links are never created automatically, so the step names the people taking part who have no link. Links are created in Invitations — Generate invite links — and the invitation message is sent from Communication. In a blind study the interface says what it cannot do rather than failing quietly: "Blind mode hides the addresses: the links are created, but invitations cannot be mailed by hand." AI panelists need no invitation.
• Conversations gives the state of each session — Started, In progress, Completed, Not opened — and the average number of turns. The AI panelists' orientation chats are started here separately; they do not start by themselves. Regenerate AI chats deletes the existing ones and cannot be undone; the confirmation says so.
• Results shows the conversations to read one panelist at a time, and a digest of the answers.
• AI analysis: Draft plan suggestions reads the conversations and suggests additions in three sections – phenomenon description, panel composition and questionnaire. Every suggestion names its sources as panelist codes, and one that rests only on AI panelists' answers is marked. The result is a suggestion to the facilitator, not a decision. Behind the expandable line Themes and perspectives are two further analyses: Themes shows what was said in the conversations, and Perspectives who looks at the phenomenon from where and what nobody said (blind spots). Items are shown as cards with their sources and quotes; a quote is checked against the answers, and if it is not found it is removed and the card says so. Any item can be added to your picks – phenomenon description, panel or questionnaire – and it appears in Selection as accepted and marked as picked from an analysis. If the AI response is cut off or cannot be read, nothing is saved and the message says why; run the analysis again.
• Selection: you handle the suggestions one at a time – Accept or Reject, and Undo decision returns a suggestion to unhandled. The decisions are stored beside the suggestion, so the selection is documented. This step does not change the plan yet: all picks are made first. Save picks and close Round 0 ends the orientation conversations, but Round 1 does not open yet. If suggestions are still unhandled, the step says how many; they are not picked.

The Round 1 box: completing the plan before opening

Once the picks are saved, Round management shows a Round 1 box in the same form, and the Round 0 box shrinks to a single line you can expand. The steps are Phenomenon and questionnaire, Respondents, Invitation and opening, Answering, Results, Analysis and Picks for the next round. Real-Time Delphi and Barometer Delphi have six steps, because they have no next round.

• Phenomenon and questionnaire shows the accepted picks. Add to the phenomenon description opens the phenomenon editor with the suggestion added, and Use as a draft thesis fills the form in Theses with the AI's suggestion, which you edit and save. Mark as used checks off a pick you handled another way. Nothing is written into the plan by itself: you save the changes.
• Respondents shows the panel picks and a door to the panel. In Classical Delphi the whole panel answers on every round.
• Invitation and opening names the human panelists who lack a link, and Open Round 1 starts the study proper.
• Answering, Results and Analysis show the round's phases and lead to the existing views. The AI light analysis for the next round will come to the box later.

During the pause, the Round management card and the What now? card say what is still unfinished in preparing Round 1. You can still open the round – it is said, not blocked.

When not everyone has answered but you have already moved on to the analysis or closed the round, Conversations is shown as incomplete with its numbers — it is not marked done.

A study can skip Round 0 — "Advance straight to Round 1. Sensible e.g. for a pure AI panel."

Round 0 belongs to every variation. It comes before the panel opens, and the variation's own character begins only when Round 1 opens — Real-Time Delphi and Barometer Delphi get the orientation round too, and the only way to skip it is the same in all. Before 13.9.2026 the state machine jumped straight to Round 1 in those two, and the card was hidden.

3. Collection: the round-management card

Collection's tabs are in the order of the work: Situation · Invitations · Facilitation · Communication · Responses (with Round 0 first when it is run). The round-management card is in Responses. The settings mentioned below live in their own homes: visibility & interaction and the deadline in Collection's Facilitation, the opening time and adding a round in Planning's Rounds section, and the phase exception and resetting an old round in Reporting's Maintenance.

Everything about running a round happens in Collection, and the card marked Round management is where the round's state and its controls live: the round, the Phase, the accumulation of Responses, and a button whose label is the act — "Move to the dialogue phase", "Open Round 0 (intake)", "Close round 2 (pause)".

Closing a round is a pause, not an ending. Closing even the last round leaves the study open: rounds can be added. Ending the study has its own control, End the study (⏹), and it is a decision — it asks for a closing letter to the panel and marks the handoff final. Adding a round afterwards undoes it. Before 31 Aug 2026 finality was a side effect of a round, and the facilitator had no way to say the study is complete.

There may be several buttons that advance a round; there is exactly one computation of what advancing means, and it happens on the server. That is why the label reads as a destination rather than a verb.

A step says that it was taken. After a successful transition a notice names the destination — "Round 0 opened · 4 key informants". Without it, opening looked like it had failed: the only thing that changed was the button's own label, which became the next destination, and a second click in the same spot carried the study past a round.

And a step that leaves something unrun is asked first. If round 0 is unopened or unfinished, moving to round 1 asks for confirmation that names both the destination and what is being left out: "Opening round 1. Round 0 (intake) will not be run." Round 0 is one-shot — orientation gathered afterwards is not orientation but post-hoc justification. Other transitions are not asked: a confirmation asked every time is learned as something to click through.

Before you open a round, two settings are worth a deliberate look, because both change what the panel experiences and neither can be repaired afterwards:

• Visibility & interaction — per phase. Round 1a — response is blind by default; Round 1b — dialogue is where the panel reads each other. The choices are Blind, After you answer, Open.
• Deadline — not set until you set it, then "{n} days left", +7 days, Clear, and deadline passed when it has. The deadline is your countdown; the panelist learns of it through a message, not through a clock on their screen.

Response links show as a state, not as addresses. A panelist’s personal link is a bearer credential: whoever holds it can answer in the panelist’s name. New studies therefore default to closed link mode (as in eDelphi): the link is created when you send the invitation and travels only to the panelist. The response-link card shows each panelist as not invited yet, invited dd.mm. with its validity, or revoked and expired. Each row has Invite or Renew and send (the new link goes in the letter, and a confirmation shows the recipient first) and Revoke. If a panelist’s email does not work, Show once reveals the link a single time: you write a reason, and the reveal is recorded in the study’s audit log. Only the study owner can switch to open mode in the settings (links visible and copyable as before); blind mode always keeps links closed. Studies created before 20 October 2026 stay open.

Try as panelist. The response-link card has a button that opens the survey in a new tab exactly as a panelist sees it at this stage — "The preview identity is not a panel member and nothing is saved." The previewer is an identity of its own (an ESIK code), the consent card and the Round 0 conversation are skipped, and every attempt to write returns "Not saved (preview)". It does not appear in the response rate or in the reciprocity figures. End preview closes the link and removes the previewer; a new one can be opened at any time. This is where the question what does this look like to a panelist gets its answer before opening, not after.

4. The response phase: your work is monitoring — and preparing

The study's phase is response. Yours is monitoring, and the screen is built for watching rather than doing: accumulation per thesis and per panelist, and one intervention that matters — Remind inactive ({n}), with a confirmation that names the number, and a result that says what happened: "Reminder sent to {n} panelists".

Read the response rate carefully. If the panel is mixed, an AI panelist answers every thesis on every round; a combined rate is therefore partly a property of the panel's construction rather than of its engagement. Pronoia computes the human rate, the AI rate and the combined rate separately and treats the human panel as primary when there are humans — and where there is nobody to divide by, the figure is absent rather than zero, because nobody was asked and nobody answered are different facts.

The AI panel's actions live behind a control. AI actions ▾ holds four: Generate thesis responses, Generate dialogue comments, Generate reactions, Generate revisions. They are behind a button because a block of machine operations rendered into the main column grows exactly the column this design is shortening, and puts machine work in front of the work that is actually yours: watching a human panel answer.

When the actions are not available the card says why rather than disappearing:

• "Actions become available once the first round is open."
• "Round 0 is open — the orientation conversations are started here."
• "The study has ended — no new responses are generated."

The middle one is not only an explanation. While round 0 is open the card carries the command itself — 🤖 AI panellists’ chat, and Regenerate AI chats when there are sessions to redo — next to Open the Round 0 view. It used to say the run was started from another view, and there was no way there from that sentence: a sentence pointing at another view is not a path to it.

Every generation passes a cost gate first. AI run cost estimate states the scope, the estimate in credits, the budget remaining, who pays and which limit binds — wallet balance, monthly cap, personal cap or this study's own cap — before Continue and generate. If the estimate is over: "The estimate exceeds the remaining budget — some responses may be produced rule-based."

Meanwhile, the other half of your work. This is when round N+1 gets built. The rail is free, Applies to: keeps the two rounds apart, and Results lets you glance at an earlier round without leaving what you are doing (§7).

5. The dialogue phase: monitoring and moderation

The panel now reads each other. Your view gains the comment stream, the dialogue threads — "Dialogue threads (comment-on-comment)" — and the AI panelists' activity in the same stream, marked.

Dialogue does not close answering. The phase adds a capability — commenting — and removes none: a panelist who did not get to answer can still answer during the dialogue, and changing an answer in the light of someone else's reasoning is the method's idea, not an exception. You need not open anything for it. A comment is locked against editing only once another panelist has replied to it; a panelist's own refinement of their own comment does not lock it.

Moderation here is mostly restraint. The dialogue phase is the study's learning phase, and a facilitator who answers arguments becomes a panelist with special powers. What the interface supports is the facilitator's own channel — the COMMUNICATION CENTRE — addressed to the Whole panel, a Group, or an Individual, with templates for Reminder, Round open and Results and placeholders ([link], [handle], [study]) resolved per recipient. Group reaches a panelist through both the home cell and any additional membership on either axis — a person who belongs to two groups receives the group message from both.

Two options there are worth understanding:

• Also send by email delivers the message to inboxes as well as the panel.
• Notice only, no text sends a mail that says a message is waiting and links to the panelist's own page, leaving the text inside the panel: "Choose this when the message contains panel content that is not meant for inboxes."

When a message goes to email it goes out as a letter: the study's name at the top, [link] as a button and the contact person at the end — the text version travels along for mail programs that do not display the letter. Two more choices:

• Include »What we ask of you« adds, after the button, a short three-point description of what the panelist is asked to do. The description is the same in every variation. It is on by default.
• Preview the letter shows the letter exactly as it will be sent — the same builder makes both. Placeholders are filled with sample values, and the preview sends nothing.

The delivery report is explicit about who did not get it — recipients without an address, withdrawn panelists skipped, and whether mail was sent at all.

6. The revision phase

Panelists revise their answers in the light of the dialogue, and say what moved them. Your view is the accumulation of revisions and the movement between rounds: Revisions, Revised, Revision rate, under the subtitle "Panelist changes to R1 responses after dialogue".

In Pronoia the revision is made during the dialogue. The panelist changes their rating in the dialogue phase with Edit, and the round moves from the dialogue straight to the pause (Close round 1 (pause)). That is why the round's path marks the phase "1c Revision" with ↺ during the dialogue instead of leaving it waiting for its turn. A separate revision phase can be opened only as an exception (chapter 10), and then the path shows it as the current phase.

Revisions is a count, not a share: there is no maximum number of revisions, and the raw figure tells you something about the panel's diligence that a percentage would flatten. Revision rate, by contrast, is one share, computed in one place: the people who changed their position divided by everyone who responded, humans and AI panelists together. Before 14.9.2026 it was computed in four places in four ways, and the community view showed 1600 %.

7. Glancing at results without leaving your work

Results in the bottom bar is the only item there that is not a tab. It opens Results so far over what you are doing: round pills for the rounds that have results, and a row per thesis with n, median and IQR.

The window contains no editing surface at all, and there is exactly one way onward — Open in Analysis → — which carries the round you were looking at with it: "Viewing only — editing happens in Analysis." When there is nothing to show it says which nothing: "No results yet — the first round has not been opened. Once it opens, results gather in this window as answers come in.", or "Round 0 is the orientation; its results appear in the Round 0 view."

Glancing is by definition something that does not interrupt. That is why it is a window and not a destination.

8. Between rounds: your richest screen too

The round closes. For the panel this is the interim result; for you it is preparing, and it is the only point in the process where the main column sits on the boundary between two kinds of work — analysis produces what planning consumes.

What Analysis gives you here

• Summary — distributions, central tendency and spread per thesis.
• Statistics & change (thesis stats, convergence, key figures, dialogicity, revisions) and ROUND-OVER-ROUND MOVEMENT — Converged, Diverged, Stable, and "Attention — split theses": "Theses where views diverge the most — often the most interesting part of the analysis."
• A readability verdict that is about your data rather than your patience: Not readable yet ("Stability cannot be measured before the round is closed and there is something to compare against"), Needs action ("Coverage is thin. Remind the non-responders before reading the result"), or Readable.

What builds the next round

• PIRE — the inter-round engine — proposes, one thesis at a time, how to continue: convergence, stable dissensus, noisy spread, an emergent theme, a bandwagon flag, and an operation to match. You accept, change the operation, flip the mode, or lock the thesis out of the next round. Its own help states the limit plainly: "PIRE proposes — you confirm. It does not replace deep analysis."
• Build the round. "Build the next round's survey from the previous round's results — copy as a starting point and edit, or open with the same theses (re-rating)." The first of these is the default in a round-based Delphi: the next survey is built from the previous results by editing, removing and adding. Plain re-rating is an exception worth justifying — in a barometer it is, conversely, the whole idea. Proposals that involve interpretation are flagged for you to reformulate before opening; you always edit the theses before the round opens. A thesis can also be copied as the base of a new one and hidden or shown mid-round behind the same gates; hiding does not delete answers.
• Reading the discussion and picking. The pause is the moment to go through the previous round's discussion question by question and branch by branch. Open the whole discussion shows a thread in full; the Timeline shows reasoning and dialogue in one stream, newest first. Any reasoning or comment can be starred (☆) — "Mark as thesis material for the next round" — noting at the same time the reason for the pick. Picks collect in the basket Thesis material for the next round: "A pick does not change the comment and is not visible to the panel. It is your memory between rounds."
• The basket is source material, not a thesis. When you write a new thesis, the basket opens beside the editor under Source material for the new thesis: "Write the thesis yourself. The contributions you select are recorded as its sources." No button turns a pick into a thesis. The selected ones are recorded on the thesis card as Source material, which later tells where the question came from. Basket and source material are different things: the basket is your tool between rounds, the source material is bookkeeping. That is why they behave differently when a contribution is deleted — the basket skips it, the source material keeps the record and the reason but the content is emptied.
• The panel's readiness for the next round is a different question from the panel's composition: who answered last round, who withdrew, where the coverage gaps are. Composition is edited in Planning; readiness is read here.
• If a round was opened past the pause and you need its survey built after all, Return Round {n} to preparation closes it back to the pause and keeps the answers already given.

When you want real interpretation, it happens outside Pronoia: Send round to DAE analysis opens the export. Theming, scenarios, argument and discourse analysis and synthesis are the analysis ecosystem's work, under its own quality gates. Pronoia describes; it does not interpret, and this button is where that boundary is crossed on purpose.

9. Reporting and handover

Generate results report produces a descriptive report — summary, per-thesis result cards, the consensus–dissensus map, round comparison R1→R2, and highlights (greatest disagreement, most discussed, most revised). Print / Save as PDF and Download Word (.docx) take it out of the browser.

The report says what it is, in its own footer, and the sentence is worth quoting to whoever receives it:

"This report is not an analysis but its source material."

Export produces three files for the analysis pipeline: the CSV (quantitative data), the Handoff Package (responses, arguments, dialogue, revision history) and DAE_STATE.json (pipeline control data). Export sits in two places on purpose: Reporting is a phase — what you do at the end; Export is a function — something you can do at any moment.

Anonymize study is the last step and it is irreversible: panelist names and email addresses are cleared and invite links revoked, leaving anonymous research data — pseudonyms without a key. Do it when the study is over and the material has been exported.

10. Exceptions, and what they cost

Three things sit outside the normal flow. Each is available, each is marked as an exception, and each is recorded.

11. What the panel sees while you work

Read that column downwards and the design is one design: the panelist gets content when it can no longer anchor them; you get it immediately, because your job is to see it.

12. Five things that cannot be undone

1. Regenerating Round 0 AI chats — the existing ones are deleted.
2. Resetting a thesis, a round or a study's responses — responses, revisions and comments for that scope are deleted. A round reset asks you to type the study name.
3. Withdrawing a panelist — name, email and link are permanently deleted.
4. Anonymizing the study — identities are removed and links revoked.
5. Closing a round — a closed round stays closed. Changing it afterwards would silently change a response rate that has already been shown and possibly analysed, and the comparison between rounds is the method itself.

Item 5 is not only guidance: the server refuses to change a closed round's dialogue, phase rules and schedule. That is why Collection shows only the open round, and a closed round's material is read in Analysis.

Everything else in this guide is reversible, and most of it is reversible by pressing the same button again.

13. What is folded, and when

Some sections are folded by default when they do not belong to the phase you are in. This is a promise, not a surprise: the table below is the whole rule, and it is generated from the same module the program reads (lib/sectionFolding.mjs). A guard test fails if the two disagree.

Three things are worth knowing before you read it.

1. Your own choice always wins. Open a section by hand and it stays open, whatever the phase says.
2. A folded section is not a hidden section. It leaves a chip and a line telling you how many sections the phase folded, so you can open them. In the CONTENTS row at the top of the view, a chip's name takes you to the section and opens it.
3. A section not listed here is never folded. The rule only removes; it never adds.

task is your work in the current phase (build · monitor · prepare · report), phase is the round's phase (response · dialogue · revision · closed), and mode is the Delphi variation. An empty column means that dimension does not restrict the section.

If a section you needed was folded, that is a fault in this table rather than in your work — and it is worth reporting, because the table is the rule.

14. The line that says where on the arc you are

Above the state card there is one short line. It is the only place that answers the question where in the whole process am I — not which screen you have open, but which stage of the study you are standing in.

It says one of these, and nothing else:

Real-Time Delphi and Barometer Delphi speak differently, because they do not walk the same arc (Classic Delphi and Argument Delphi walk it together):

Why it matters that this is a sentence and not a code. Internally each stage has a short code (A1, B1, A2 …), and it is in the page's markup where a test can read it. It is deliberately not on the screen: a code would be a third vocabulary beside the sentence and the guide, and the codes only exist in round-based Delphi — in the Barometer and in Real-Time Delphi the code is empty, which would make the screen look as though the code were missing rather than as though it does not exist.

So the line is written to be read, not decoded. If it disagrees with what you believe the study is doing, believe the line: it is computed from the study's own state, not from the round number.

15. Retirement is the Barometer's way, not a general one

A thesis whose time has passed can be retired — but only in the Barometer, and only at the turn of a wave. Everywhere else the program refuses, and the refusal says why: "Retirement is possible only in the Barometer, between waves. Let the thesis run the round to its end."

This is a methodological rule, not a limitation of the software.

• In a round-based Delphi a thesis runs the round to its end. The whole point of the method is the comparison between rounds: the same thesis, asked again after the panel has seen each other's reasoning. Removing a thesis mid-round would delete the very measurement the round exists to produce.
• In the Barometer the waves differ only by their date. The same theses are rated from one time to the next, so a thesis can genuinely reach the end of its usefulness — the future it asked about has arrived, or the question no longer divides anyone. The facilitator decides, and a new thesis can take its place.

What retirement does, exactly:

1. The thesis stays in the current wave to the end. Panelists can still rate it, and the wave's results are archived into the wave series as usual.
2. At the wave turn it moves to the archive and is no longer asked from the next wave onwards.
3. Its responses are kept for analysis. The export marks it lifecycle = retired, so the analysis can see that the thesis is no longer being asked — a retired thesis is not a deleted thesis, and the difference matters when someone later wonders why a series stops.
4. The decision can be cancelled before the wave turn.

The reason is asked for and stored ("time is up", "the future is visible"). It is worth writing properly: at the turn of the next wave it is the only thing that explains why the series ends there.

Written 26.8.2026, against the interfaces as rebuilt on 23.–25.8.2026. Button and tab names are taken from the application's own English strings, and a guard test (scripts/test_guides_coverage.py) fails if this guide names a control differently from the program. The specification behind this text is docs/UI_Vaihearkkitehtuuri_IA_v2.md; where it and this guide disagree, it is right and this is stale.

Role play: a role in a box of the panel matrix

Facilitation (Pronoia)

What role play is for

In an ordinary Delphi the panelist answers as themselves: their expertise is what they bring. In role play the panelist answers from a role they write themselves. The role can be their actual position, a future position (Municipal finance director in 2035) or a perspective they deliberately choose to represent (Young rural entrepreneur).

Role play suits a study when

• you want to make assumptions visible: once the role is written out, the reader sees from which position an argument is made
• you are looking at future actors: the panelist reasons as a decision-maker of 2035 rather than as today's official
• the panel is missing perspectives you want included: a human panelist can take a role, and the facilitator can commission a role from an AI panelist
• the study is used for learning or workshops, and changing perspective is the aim itself.

Role play does not suit a study whose value rests on verified expertise. A self-selected role is a claimed perspective, not a qualification, and it is always marked as such.

Division of work: the facilitator gives the structure, the panelist chooses the box

Role play is built on the panel matrix. The facilitator defines the matrix axes and categories, for example the main axis Expertise with categories Technical, Economy, Administration, Experience and a second axis Stakeholder. The categories are boxes. The panelist writes their role freely and places it in the box it fits best.

The role is a name, the box is a group. Group comparison, pulse group scores and reports group panelists by box, not by role text. This is the key design choice of role play: if every freely written role were its own group, every group would have one member and k≥3 protection would hide them all. The boxes are what make role-play results open to structured analysis.

The facilitator's steps

1. Build the panel matrix (Panelists → Panel matrix). Choose categories with role play in mind: broad enough that each gets at least three panelists, and clear enough that panelists can choose.
2. Switch role play on in the settings and save. The setting explains what the boxes are. If the matrix has no categories, the setting warns you: panelists can then write a role but cannot choose a box.
3. Tell the panel what a role means in this study (template below).
4. Commission AI roles if you want to supplement the panel. On the AI panelists tab, in the add form under Role play: AI role, write the role in the AI role field and choose a box from the same matrix. The AI answers from its role's perspective.
5. Watch the boxes fill before opening the round: a box that stays empty or gets only one or two panelists is not shown as a group.

Template for the panel

In this study you answer from a role. On your own page, on the Role play card, write the role you answer from, for example your own position in 2035 or a perspective you want to represent. Then choose the box your role fits best. Other panelists see your role next to your reasoning, but not your name.

What the panelist sees

The panelist's own page has a Role play card. The panelist writes their role in Your role (at most 120 characters), picks a box on the main axis under Choose a box (and optionally on the second axis), and presses Save role. The card says: “Other panelists see your role next to your reasoning and comments.”

In the discussion the role appears next to the handle, marked:

• self-selected: the panelist wrote the role
• AI role: the facilitator commissioned the role from an AI panelist.

The markers never mix: a role a human chose is never shown as an AI role, or the other way round. When role play is switched off, roles are shown to no one, but the saved role and box are kept.

AI roles as panel supplements

An AI role is useful when the panel lacks a perspective you cannot get from human panelists: young people, future residents, actors from another field. The AI argues from the role's perspective, and its contributions are always marked.

Keep two things in mind:

• An AI role is a proposal, not a voice. It brings arguments into the discussion that people can respond to, but it does not tell you what people in that group actually think.
• The AI counts as a member of the box. The k≥3 threshold counts AI panelists in the group size (facilitator decision, 22 September 2026). A box with two humans and one AI is shown as a group. If you want the anonymity of the human group to be certain, keep at least three humans in each box.

Reading the results

• Compare boxes, not roles. Group comparison shows how the boxes differ. A single role can be read next to its reasoning, but it is not counted.
• A self-selected role is a frame, not a sample. A box's result tells you how the panelists who took that perspective argued, not what a whole profession thinks.
• Separate the AI share. In the distribution the AI share is hatched and the counts are split, so you can see how much of a box's result the AI produced.
• Use roles in interpretation. The combination of reasoning and role is often the richest material in the study: the same claim made from different roles shows where a view depends on position.

Technical description

• The panelist's role is stored in self_selected_role and the boxes in matrix_primary and matrix_secondary. Call: POST /api/respond/{invite}/role with the fields role, matrix_primary, matrix_secondary.
• An unknown box is rejected (role_play_box_invalid). When the matrix has categories, a role without a main-axis box is rejected (role_play_box_required).
• Adding an AI panelist (POST /api/studies/{study}/ai-panelists) takes the same fields; the role is stored as the profile's perspective.
• In the discussion response the role travels in panelist_role and commenter_role only while role play is on.

See also: Anonymity: identity modes and group reporting.

One-shot poll at a seminar: QR code and anonymous answers

Facilitation (Pronoia)

What the one-shot poll is for

The one-shot poll is Pronoia's lightest way to collect answers: no invitations, no sign-in, no handle. The audience scans a QR code with their phones, answers the theses and submits anonymously. Everyone answers once.

The one-shot poll suits

• taking the temperature of a seminar or workshop: where the room stands on the theses before and after a discussion
• the start or end of a Delphi: a wider audience's view alongside the panel's results
• teaching: presenting the method so that everyone gets to try it.

It does not suit a vote whose outcome is binding. Anonymity means no one can verify who answered, and the duplicate-vote guard only works per browser (see Limits).

Before the session

1. Write the theses. Scale theses are answered with a slider, other theses in text. Short theses about one thing work best in a hall; three to six are enough.
2. Choose Anonymous one-shot poll in the settings. Tick Poll open — accepting responses right now and save. An unsaved poll will not open.
3. Check the share card. A card appears under the tick with the QR code, the Public link and the poll's state. The state should say that the code works now.
4. Test with your own phone. Scan the code, answer and submit. A second attempt from the same browser says you have already answered; that is correct. Your test answer stays in the results, so allow for it when you read them, or test on a separate copy of the study.
5. Prepare the presentation view. Open presentation view opens a white full-screen page: the study's name, a large code, the address and the prompt “Scan the code with your phone and answer”. Open it in a browser tab you can project, or take a screenshot for a slide. Copy link puts the address on the clipboard if you want to send it to remote participants.

The address carries the study's language. Voters can switch language (Finnish, English, Swedish) at the top of the page.

During the session

What to tell the room:

Scan the code with your phone. Your answer is fully anonymous: no name, no sign-in, and no identifier is stored about you. Answer the theses and press Submit anonymously. Everyone answers once.

Keep the presentation view on screen throughout the poll: latecomers can still join, and the address is readable for anyone whose phone will not read the code.

Disruptions:

• The service is busy. The voter sees “The service is busy right now. Wait a moment and submit again – your answers are still in the form.” Ask people to wait a moment; nothing is lost. The server's limit is sized for a hall (see Technical description).
• The network drops. The message says the answer was not saved. The answers stay in the form, and submitting works once the connection is back.
• Someone presses submit without answering. The page asks them to answer at least one thesis.

Closing the poll

Remove the tick from Poll open — accepting responses right now and save. After that the code opens a page that reads “The poll is closed.” Anyone who submits at the very moment of closing sees a message that the poll was closed before their answer arrived.

You can reopen the poll the same way, for example for a second measurement after a discussion. Note that the same-browser guard applies to the whole poll: to measure the same audience twice, run the second measurement as a separate study.

Results

The answers go into the round's results and distribution like any other answers. Each submission is stored as an anonymous participant who does not appear on the panel roster or in group comparison. That is why one-shot poll results cannot be broken down by group: voters have no box in the panel matrix.

Read the result like this:

• The distribution shows where the room stands, not the expert panel's position. If you set it beside Delphi results, say in the report that it comes from a different group.
• The reasons are anonymous and one-off. They cannot be clarified or answered, so they are material, not dialogue.
• The number of respondents is the number of submissions. Without an identifier, it does not tell you how many different people answered.

Limits

• The duplicate-vote guard is per browser. The same browser answers once. The guard stops accidental repeats, but not deliberate ones: another browser or a private window can answer again.
• No commenting. Voters can write a reason, but cannot comment on others' answers.
• No follow-up. A one-shot poll creates no own page, no revision and no next round. To engage participants further, invite them to the panel in the usual way.

Technical description

• The poll page is /poll/{study}?lang={language} and the presentation view /poll/{study}/qr. The QR code is drawn in the browser (error correction level M, four-module quiet zone); the address never passes through an outside service.
• GET /api/poll/{study} returns the theses of an open poll and POST /api/poll/{study} stores the answers. Each submission creates an anonymous participant (participant_type = anonymous).
• The server rejects a closed poll (403) and overload (429); the page tells the voter about both.
• The per-address limit is 1,500 submissions and 3,000 page loads in ten minutes. Phones in a hall often reach the server from the same address, so the limit is sized for a hall, not for a single user.
• The duplicate-vote guard lives in the browser's storage; the server stores no identifier for the voter.

See also: Anonymity: identity modes and group reporting.

Taking part in the open Delphi study

Participation (Pronoia)

Metodix always has one open Delphi study running, and anyone interested in the topic can ask for an invitation. This guide tells you in two minutes what you are joining. Answering itself takes about fifteen minutes.

No account needed

A personal link arrived in your email, and that is all you need. There is no password and nothing to register. The same link brings you back to the panel round after round, so keep the message.

Please don't pass the link on. It is your seat in the panel, and if someone else answers with it, their views are recorded as yours.

A Delphi study is not a survey

A survey adds opinions up. A Delphi study reads them. A round goes like this:

1. You answer first. You rate the theses before you see what anyone else said. The order is deliberate: the first number you saw would stick to your own.
2. You give your reasoning. The reasoning is the most important part of your answer. The rating says where you landed; the reasoning says why, and that is what the analysis reads.
3. You read the others' rationales. Under OTHERS' RATIONALES you see the rest of the panel anonymously. You can comment on a single rationale and ask its author for more.
4. You may revise. If some argument makes you think differently, change your answer and say what changed it. Changing your mind is not weakness but a result of the method, and it appears in the analysis as a finding of its own.

The aim is not unanimity. The aim is that those who disagree know why they disagree.

Anonymity is structural, not a promise

The other panelists see only a pseudonym. They see your rating and your reasoning; your name, your organisation and your email address stay out of sight. This is not politeness but a condition of the method: anonymous reasoning is judged by its content rather than by its author's standing.

In the distribution chart your own answer is marked you, and in the discussion your own turns appear to you as You. To everyone else the same row sits under a pseudonym.

Only the organisers at Metodix see your contact details, and they use them solely to send the invitation and the round messages.

What else is on the page

• DISCUSSION gathers the conversation around a thesis into a thread, and YOUR RATIONALE shows your own text, which you can refine.
• More services shows what has been stored about you, and it is also where you can withdraw from the study.
• Weak signals & wild cards is a place for observations that do not fit the theses but may still change the picture of what is coming.

If something doesn't work

Open Share an observation and write down what you tried and what happened. The message goes to the people who build Pronoia, not to the study's facilitator, and it has no effect on your answers. The most useful thing you can tell us is where you looked first — even if it was the wrong place.

Short answers

The facilitator trial in five stages

Facilitation (Pronoia)

You have a trial of Delphi Pronoia. In ten minutes this guide walks you through what running a Delphi study actually involves.

The stages are written as goals: each one says what you are trying to find out, not which button to press. Finding the route is part of the trial — and if you can't find it, that is the most valuable thing you can tell us.

1. Open your own example study

The invitation link lets you create an account and a password. After that your desk holds a study that already has a panel, theses and answers. It is yours: you can change anything in it without breaking anything.

Goal: find out what the study asks and how many people have answered.

Both answers are on the Theses and Panelists tabs. Note the basics of the method while you are there: a thesis is a claim to take a position on, not a question.

2. See where the panel landed

Goal: find one thesis the panel disagrees on, and find out on what grounds.

The distribution shows how the answers fell. Below it are the statistics: mean, median, standard deviation and the middle half. In a Delphi study the spread is often more interesting than the mean, because it tells you whether the panel is of one mind or split.

The reasoning, though, is why Delphi studies are run at all. Read it and see whether the disagreement lies in the ratings or only in the reasons behind them.

3. Read the dialogue

Goal: find a place where someone revised their position, and work out what changed it.

Panelists comment on each other's reasoning and may revise their own rating. The revision and its reason are stored separately. This is what separates a Delphi study from a survey: the material keeps a record of which argument moved whom.

4. Change something

Goal: add one thesis of your own and see how it looks to a panelist.

Use the panelist preview: you see the same view as your panel, and nobody is notified. It is the fastest way to check whether a thesis makes sense.

Keep one rule in mind: a thesis added while a round is open means that part of the panel answered a different set from the rest. Pronoia does not prevent it, but it does record it.

5. Build a study of your own

Goal: take one real question of yours as far as the theses.

Start from the dashboard button ➕ New study (wizard): choose a template, and the wizard takes you through seven steps to a study of your own. The illustrated guide is Your first Delphi with the wizard; if you already have a plan, see Which way to Pronoia? On the AI panelists tab you can add synthetic perspectives if you want to see the dialogue move without a real panel.

When you invite real panelists, the Invitations tab creates a personal link for each of them. You can send them from Pronoia or copy them into your own email.

The limits of the trial

Hitting a limit is not an error but the end of the trial. Tell us what you were doing, and we will agree on what comes next.

Tell us what didn't work

At the bottom of the screen there is Observation (the Share an observation window). Use it freely. The valuable part is not that something was broken, but where you looked first and what you expected to happen. Pronoia's most common defect is a feature that exists but cannot be reached from the screen where it is needed — and only a new user can see that.

DAE analysis: the pipeline, phases and quality gates

Analysis (DAE)

DAE (Delphi Analysis Ecosystem) is the ecosystem's interpretive layer. Where Pronoia describes — what the panel said and how views moved — DAE answers what it means. DAE is the methodological interpretive authority, and it works under quality gates (κ / QVG) and human review.

The boundary is kept strict on purpose. If description and interpretation blurred — or if assistants generated "ready" conclusions — a false sense of precision would arise. A Pronoia report is not an analysis but its source material; interpretation is done in DAE.

Handoff in — what DAE reads

DAE receives Pronoia's export contract: an eDelphi CSV, a Handoff Package (JSON, read by DAE CP-0/CP-2) and DAE_STATE.json. The package carries the orientation, the panel matrix type, the argument space (theses × members), the epistemic coverage, the perspective summary and the validation status. Downstream verifies this contract before starting — a malformed dataset cannot silently enter analysis.

The pipeline: 21 checkpoints, eight phases

The DAE orchestrator runs the material through 21 checkpoints (CP-0…CP-20), one phase at a time, with operational discipline and User Choice Points. The phases:

The phases group into trilogies: content (what the panel argued), system (depth · dynamics · purpose) and futures (signals · tensions). Synthesis integrates every lens.

The quality engine — when the human enters the loop

DAE's quality engine ensures the qualitative checkpoints are valid:

• Smoke and adversarial tests plus a gate validator gate every phase.
• κ (kappa) measures coding agreement; low κ routes to human coding (QVG — Quality Validation Gate).
• RED ZONE phases (synthesis, scenarios, reports) require special human judgement.

When automated confidence is not enough, the analyst is brought into the loop — and the facilitator decides how far interpretation should reach. This is the arc's methodological guardrail: delegation is not a loss of control.

Outputs

The pipeline ultimately produces scenarios (several paths: causal layers, multi-level perspective, tensions, soft systems) and audience-specific reports with data-provenance markers. The synthesis is falsification-tested and keystone-validated — not an impression but a traceable, gated interpretation.

What if DAE is not in use?

Not everyone has the DAE ecosystem. For them, Pronoia offers descriptive report formats (summary, per-thesis result cards with arguments, round comparison) in HTML/PDF/Word. They give a proper summary but stay deliberately descriptive — they do not replace DAE's interpretation.

Sources: help/06-dae; the DAE core (dae-core) and phase modules.

Glossary

Planning (DPE) Facilitation (Pronoia) Analysis (DAE)

Abductive thesis — an open, exploratory thesis (E orientation): what is emerging.

Argument Delphi — a variation: dialectical theses and dialogue over rounds; aims at dissensus — maps tensions and reasoning (lead indicator CDI). The same process and phases as Classic Delphi.

Argument Simulator (Route B) — generator of synthetic, orientation-aware arguments when no real panel exists.

Barometer Delphi — a variation: the same theses rated from one time to the next, wave by wave; change over time is the finding (BCI). The only variation in which a thesis can be retired.

Basket (thesis material for the next round) — the facilitator's picks between rounds; does not change the comment and is not visible to the panel. Opens beside the thesis editor as source material.

Blind-facilitator mode — a per-panel setting that hides panelists' names and email addresses from the whole study team, the owner included. It locks once the first human panelist has been invited or has answered. Not the same as the blind visibility phase.

Classic Delphi — a variation: the same process and phases as Argument Delphi; aims at reasoned consensus and measures it with consensus indicators (CCI). The value stored in the contract is Classical.

Commensurability — enough shared ground for genuine dialogue; the goal of panel design (not full consensus).

Convergence — the narrowing of positions across rounds.

CP-0…CP-20 — the DAE pipeline's 21 checkpoints, grouped into eight phases.

Dialectical thesis — a claim taken a position on by probability and desirability (N/S/R/D).

Dissensus — structured, well-understood disagreement; in Delphi as valuable a result as consensus.

ECI (Epistemic Coverage Index) — how many of the 21 argument cells (7 knowledge bases × 3 functions) are filled.

Export contract — Pronoia's validated payload to DAE (eDelphi CSV, Handoff Package JSON, DAE_STATE).

Home cell and additional membership — a panelist's primary place in the panel matrix and any additional groups on the same axis; group size is counted in people, not memberships.

Handoff — a structured transfer between stages: Planning Handoff Package (in), export contract (out).

k-anonymity (k = 3) — no sub-group of fewer than three people is reported at group level.

κ (kappa) — a coding-agreement measure in DAE; low κ routes to human coding.

Lock-in — a single perspective claiming universality and crowding out the rest.

Logging policy — a public page (pronoia.fi/lokikaytanto): what is recorded in the event log and what is not, who sees the log and what the service administrator has committed to.

Closing — the study's final pause: the facilitator writes a letter about the results and next steps, and it appears on the panelist's home page. Not a state that is set but a task that is done.

Orientation (N/S/R/E/D/D+) — the study's locked standpoint; sets theses, panel type and between-round stance.

Panel matrix — a two-axis designed epistemic space (e.g. expertise × stakeholder).

Pause — the state between rounds (round_closed), in which the facilitator reads the discussion, picks and builds the next round; for the panelist an interim result, not a waiting room.

Perspective = position + view — the same phenomenon looks different from different positions; the aim is to identify what each can/cannot see.

PHS (Perspective Health Score) — panel diversity on 0–12; guides supplementation.

Pick — starring (☆) a reasoning or comment as thesis material for the next round, with a reason; a methodological move from dialogue to a new instrument, not a bookmark.

PIRE — Pronoia's orientation-aware round-gap assistant.

Planning Handoff Package — the planning payload into Pronoia: orientation, matrix, theses, perspective context.

PQS (Panel Quality Score) — DAE's panel-validation quality score (CP-1).

Preview (Try as panelist) — the facilitator sees the survey through a panelist's eyes under a preview identity of its own; nothing is saved, and the previewer does not appear in the panel's figures.

Pseudonym (handle) — a stable word handle (e.g. Aurora); anonymous to others, recognisable in dialogue.

QVG (Quality Validation Gate) — DAE's gate that routes to human review when AI validity is not enough.

Real-Time Delphi — a variation: one round in which answering and dialogue are open at the same time from the start; no comparison between rounds (RCI). The aim is set thesis by thesis.

Retirement (retiring → retired) — a Barometer Delphi concept: a thesis that has lost its future relevance is archived between waves. The decision and its taking effect are two moments; the replacing thesis names the one it replaced.

Revision rate — the share of people who changed their position after the dialogue among everyone who responded, humans and AI panelists together; computed in one place. Not the same as the number of revisions.

Round 0 — the orientation: an AI-assisted intake that motivates participants and maps the phenomenon before the instrument is locked. Belongs to every variation; the only way to skip it is the same in all.

Source material (provenance) — the contributions recorded on a thesis card as the ones the thesis came from; a deleted contribution keeps its record, its content is emptied.

Trilogy — a DAE phase group: content, system (depth/dynamics/purpose), futures.

Variation — the cadence of rounds: Argument Delphi, Classic Delphi, Real-Time Delphi or Barometer Delphi. The name is a proper name and is not translated. studies.variation is the truth from which delphi_mode is derived.

Wave — one measurement in a Barometer Delphi: the same theses rated again, the waves differing only in their date. A wave is not called a round, so that the words do not blur.

Phases (1a/1b) — a round's response (1a) and dialogue (1b) phases. A phase adds a capability, it never removes one: answering stays open in the dialogue.

Visibility (blind / after-submit / open) — the per-phase control of when a panelist sees others.

Frequently asked questions

Can I use just one part of the ecosystem?

Yes. The three products work together as a pipeline but also apart: you can produce only a design (DPE), run Pronoia without DAE (descriptive reports), or analyse an existing dataset with DAE. The handoffs join the parts when you use them in sequence.

What is the difference between Pronoia and DAE?

Pronoia is descriptive (what the panel said and how views moved). DAE is interpretive (what it means), under quality gates and human review. A Pronoia report is DAE's source material, not an analysis.

Do I need DAE to get results?

No. Pronoia produces descriptive reports (summary, per-thesis result cards, round comparison) in HTML/PDF/Word. They are a proper summary but stay descriptive.

How do orientation and variation differ?

Orientation (N/S/R/E/D/D+) gives the standpoint — from where the future is studied. Variation (Argument Delphi, Classic Delphi, Real-Time Delphi, Barometer Delphi) gives the cadence — how rounds are paced. They are different axes. Classic and Argument Delphi run the same process and differ only in the aim (consensus / dissensus); Real-Time Delphi runs one round; Barometer Delphi rates the same theses again from one wave to the next.

Why is the response phase protected from others' answers?

To protect the core sequence independent judgement → feedback → reconsideration from social pressure (anchoring, bandwagon, dominance). The default preset (Classic, used by both Classic Delphi and Argument Delphi) uses after-submit: no anchoring before you commit, feedback right after. The protection concerns seeing, not answering: in the dialogue phase an unanswered thesis can still be answered.

Can the facilitator see panelists' real identities?

By default yes (for reminders and quality control). A per-panel blind-facilitator mode hides them, so the facilitator also sees only pseudonyms. Blind mode locks once the first human panelist has been invited or has answered, so it cannot be lifted during the study. In every mode, only the panelist can change their own response; see also the Logging policy.

How are AI panelists marked?

Always and everywhere: "AI" plus a 2–3 word role description (e.g. AI · Economist). Never hidden, never presented as human.

How does group reporting protect anonymity?

With k-anonymity (k = 3): sub-groups of fewer than three are not reported, and the rule applies independently on each matrix axis. On small panels the pulse is shown at whole-panel level only.

When do I close the study?

When the panel has converged, or when dissensus is stable and well-understood — further rounds would add cost without insight. Read the pulse (convergence/dissensus per thesis) before deciding.

Are AI assistants' suggestions finished conclusions?

No. Generative assistants (e.g. PIRE's deepening suggestions) are always drafts that require an audit and that the facilitator owns. Interpretation is done in DAE under quality gates.

Role quick-guides

Short entry points: where to start, the most important guides, and the one next line whose crossing raises your level.

Researcher / designer

Planning (DPE)

Your job: the research design — orientation, phenomenon, panel, questions.

Start: Orientation, variation and question architecture → Building the panel.

Next line: from assembling a single study → the Planning Handoff pipeline that seeds Pronoia without rebuilding.

Facilitator

Facilitation (Pronoia)

Your job: running Pronoia — Round 0, rounds, dialogue, anonymity, reports.

Start: The iterative process → Anonymity. Whole picture: Integration guide.

Next line: from building each round by hand → let PIRE suggest the between-round operations (you confirm); from reading descriptive reports → hand a round to DAE for interpretation.

Developer-partner

Analysis (DAE)

Your job: integration, handoffs, MCP/plugins, installation.

Start: Integration guide (handoff schemas) → DAE pipeline → Anonymity: data model.

Next line: from handling individual transfers → the handoff contracts (Planning Handoff Package, export contract) that downstream verifies automatically — a malformed dataset cannot slip through.

Panelist

Facilitation (Pronoia)

Your role: you take part in the panel — answer the theses, read others' reasoning and revise your position.

What you need to know: you answer independently first (others' answers are hidden until you commit), then you see an anonymized picture of the panel and can comment and update. You are anonymous to others under a pseudonym (e.g. Aurora); the facilitator may see your identity unless blind mode is on. AI panelists are always labelled "AI". The Home view shows a "since your last visit" digest — come back to see how the panel moved.

Next line: from a one-shot answer → coming back to revise: Delphi's value comes from reconsidering in light of others' reasoning.

→ A light participation guide is also in the app's in-app help.

Decision-maker / stakeholder

Analysis (DAE)

Your role: you read the results and reports to support decisions — you don't run the study yourself.

What you need to know: distinguish descriptive from interpretive. A Pronoia report tells you what the panel said and how views moved (distributions, consensus/dissensus, arguments) — it is source material. A DAE report interprets: themes, scenarios, synthesis under quality gates. Dissensus is a result, not a failure: stable disagreement signals a real tension. Scenarios are alternative paths, not forecasts.

Next line: from reading a single figure → ask what is description and what is interpretation, and what a conclusion rests on (data provenance).

Start: Overview (descriptive vs. interpretive) → DAE guide (outputs).

Slides and deep dive

Planning (DPE) Facilitation (Pronoia) Analysis (DAE)

Presentation slides

The slides are generated from the same markdown source as these guides — in one build, as reveal.js decks. They are always in sync with the guides, not a separate slide library to maintain. Navigate with the arrow keys or space; the down arrow opens a section's inner slides.

Worked example

One deeper learning artifact is Food Paradigm 2050 — the whole planning pipeline run end to end (orientation D, Route B, 19 roles, PHS 12/12, 27/27 quality gates). It shows the method in practice; see also the Simulation block.

Note — the worked example is from 2026-05. Its method (orientation, panel, theses, perspective layer, handoff) still holds, but product specifics have moved on since. The most up-to-date source is always this guide (DPE/Pronoia/DAE).