True Causes in Numbers: Rooms, Claims and Causal Links

Numbers

Everything on this instance, counted from the record. Per-dialogue quality indicators follow the method’s own (Laouris & Metcalf 2025, Table 7); the usual ranges from their hundred and twenty dialogues are noted where they apply. No question from an unlisted room appears here, only counts.

Questions and rooms

dialogues (rooms)

49 (36 simulated, 12 AI)

distinct questions

federations

rooms inside federations

largest federation

reached the map

People

people (accounts, not stand-ins)

1141

seats across all rooms

1233

average per room

25.2

groups formed

Ideas

cards written

1698

average per room

34.7

carried between rooms

265

clusters formed

133

clusters per map, average (dimensions)

2.7

cards placed in a cluster, average

16.1%

cards structured

823

influence links certified

782

Findings

root causes per map, average

4.7

most in one map

pairs left contested

293

spreadthink, average

56.4%

demosensus, average

43.6%

Time, average per room of people

Generate open · 1 room

5.7 h

question to map

—

question to actions closed

—

Time, walkthroughs and rooms of models

Generate open · 46 rooms

17.3 h

Clarify open · 46 rooms

6 min

Cluster open · 46 rooms

23 min

Select open · 46 rooms

29 s

Structure open · 46 rooms

40 min

Map open · 46 rooms

1 s

Actions open · 45 rooms

8.9 days

question to map · 46 rooms

18.4 h

question to actions closed · 45 rooms

9.7 days

Stand-ins and models act at once, and the steward starts them when it suits, so these clocks describe the demonstrations, not people. Kept apart so they do not swamp the rooms above.

Effective time per member

members measured

median, rooms that closed Actions

—

median, every room so far

—

middle half of members

—

sittings per member, median

—

days with activity, median

—

Counted from each member’s own clicks, and a beat the page sends once a minute while it is open in front of them and they have scrolled, typed or clicked in the last two minutes. Time between two signals counts when it is under five minutes; longer is time away. So it is a floor: reading after the last click of a sitting is not in it. Walkthroughs and rooms of models are left out, and rooms that ran before it was recorded have no figures.

Where dialogues are

Generate

Clarify

Cluster

Select

Structure

Map

Actions

Judging

pairs judged

6786

votes cast

108603

votes per person, average

92.8

deadlocks sent to another group

711

of those, settled there

345

of those, published as contested

293

feedback loops found

Actions, in rooms that closed them

rooms that closed Actions

actions proposed

149

action × root squares settled

498 of 682 (73%)

root causes with an action that reduces them

root causes nothing acts on

105

actions per root cause, average

0.8

most root causes one action moves

actions that make some root cause worse

contradictions found (reduces one root, worsens another)

plans compared

rooms that chose a plan

root causes the chosen plan covers, average

—

Predictions

predictions made

with a measure and a date

guesses submitted

538

outcomes: as expected

partly

not at all

overdue

Evidence

sources nominated

backed by a second person

evidence seats taken at Select

Rooms

opened by taking up a question

opened to an email domain

with a call per group

sealed

public-record rooms

links found independently by several rooms

Minds moved

people who said where they stood, before and after

whose stance changed

average shift in certainty (−5 to +5)

3.0

Second pair of eyes

times a card or turn was handed back to its author

Usual ranges from practice: 50–100 cards per room of 15–25 people; spreadthink 35–55% at voting, 25–45% once meaning has been shared. Refreshed every minute.