Map

On this page

“Map” (Beta) shows the arguments people make in a project. It reads the project’s transcripts, finds the arguments, and places related arguments close together. Each point on the map is one argument, and each argument carries the quotes it came from.

“Map” is in the sidebar of every project. It needs no paid plan.

Generate a map

Press “Generate map”. Once a map exists the button reads “Regenerate”, and after a failure “Try again”.

Without transcripts the page says: “This project has no conversations with transcripts yet. Generate a map once conversations have been transcribed.”

What a point is

The header counts “{n} arguments from {m} conversations”. Each argument has:

A claim is a statement about the world that a web search could find evidence for. Anything else is a personal stance, a value judgement, a preference, an anecdote or reasoning from first principles. The map labels it “Opinion”.

When several arguments say the same thing, the analysis combines them itself and shows “combined from {n}”. You cannot merge or rename arguments by hand.

The two views

The same arguments have two views: “Cluster map” and “Argument tree”. The toolbar shows them “One at a time” or “Side by side”. The page starts on the cluster map.

Click an argument in either view and it opens in “Spotlight”, with “Quotes ({n})” and “Connections”. Each quote links to the page of its conversation.

Rest the cursor for 1.5 seconds on a group of three or more points and the group is saved as a cluster. A language model gives it a title (“Distilling core idea…”). Saved clusters also show on the room screen, the public page and the preview on “Present”.

The map draws 150 arguments by default. “Map budget” in the settings changes that. For a large map the page warns: “This map has {count} arguments. Rendering it may be demanding on your device.”

Colours and filters

“Color nodes by” in the settings colours the points by:

“Filter” narrows the map by tag (“Show only conversations with any of the ticked tags.”) or by conversation. The page then tells you what is left out: “Showing conversations tagged {tag}.” or “{n} conversations are hidden from the map.”

Fact-checks

Only claims can be checked. For an opinion the map says: “Arguments express stances or preferences and aren’t fact-checked.”

A fact-check is the work of a language model. One call searches the web with Google Search, and a second call classifies what it found. Read the justification and the sources before you rely on a result.

Tensions are not on this page

The Map page shows arguments. Tensions show as the “Tensions” outcome on the room screen, and in “Review results” and “Analysis”, each with its two sides and a line “To resolve:”.

Edits made elsewhere change the map

The Map page has no edit controls. You edit a finding under “Review results” on Present or under “Results” on Analysis, and the map follows.

Edit Effect on the map
Reword an argument The map shows the new wording. The earlier wording is kept in the finding’s history.
“Withdraw” an argument It leaves the analysis and the map. “Undo withdrawal” brings it back.
“Hide from this presentation” It stays on the Map page. It is left out of that presentation’s map for the audience.

The audience’s map never shows the history, and it has no fact-check buttons.

Who can do what

Action Admin Member External Observer
See the map Yes Yes Yes Yes
Generate a map, start or cancel a fact-check, save a cluster Yes Yes Yes No

Observers see the map without the “Generate map” button. See roles & permissions.

Related pages

Comments