Not a poll. Not a vote. A structured practice of collective
reasoning — with AI that reads the pattern of how a
group's thinking shifts across time.
Augur is the asynchronous equivalent of real-time swarm intelligence — but richer, because it captures not just where a group ends up, but how it got there.
"Real-time swarms produce
a snapshot. Augur produces
a trajectory."
— The Augur approach
A traditional consilium is humans deliberating. Augur is something different: humans deliberating alongside an AI that reads the structure of their reasoning — tracking not just positions, but confidence, influence, and the territory the group has not yet entered.
Platforms like Unanimous AI's Swarm require everyone present at the same moment, producing a single statistical snapshot. Augur works across days and weeks. The intelligence it surfaces is richer precisely because it has time to accumulate.
The name is deliberate. An augur read patterns and made meaning. Augur doesn't decide for the group — it tells the group what its own reasoning, read carefully, reveals.
How it works
When creating an Augur item, the organiser or the Knowledge Garden pipeline defines a convergence question — a single consequential question with three to five answer options. This becomes the structured pole around which discussion is anchored.
Each participant writes a response — their reasoning, in their own words — and optionally registers a position on the convergence question along with a confidence level from one to five. Position without reasoning is possible; reasoning without position is equally valid.
As members read each other's responses, they may update their position. A simple prompt — "did this change your view?" — captures the connection. The original position, the new position, and the response that provoked the shift are all preserved in the knowledge graph.
Influence trackingOver time, a map emerges: which responses influenced which participants, whose reasoning carried weight, where the strongest dissent clusters. This is not available in any synchronous swarm system — it requires time, and the graph to hold it.
Knowledge graphAt any point, an organiser can request an Augur synthesis. Claude reads the full thread alongside the concept graph from the originating source. The synthesis names where the group has converged, what the strongest dissenting reasoning is, and — crucially — which concepts the source introduced that the discussion never engaged with. These gaps become the seeds of the next question. The synthesis is not a conclusion; it is an invitation to continue.
Claude SonnetComparison
Four mechanisms
Members register a position on the convergence question alongside a confidence level from one to five. High confidence and low confidence are equally informative — a cluster of uncertain responses around a position tells a different story than a cluster of the certain.
When a member updates their position after reading a response, that relationship is recorded as an edge in the knowledge graph: Response A influenced Member B's shift from position X to position Y. Over time the argumentative structure beneath the thread becomes navigable.
Early responders who shift differ meaningfully from late responders who hold firm. The group's trajectory — who moved early, who held out, where consensus formed and where it fragmented — is the intelligence that synchronous swarms structurally cannot produce.
The AI synthesis reads the discussion against the concept graph from the originating source. Concepts the source introduced that the discussion never engaged with are surfaced explicitly. These gaps are not absences — they are the seeds of the next Augur question.
What gets built
Every response that carries a registered position stores that position and confidence level as properties on the Response node — alongside the reasoning text and the standard authorship and timestamp data.
When a member updates their position after reading another response, an INFLUENCED_BY edge is created between the two Response nodes. The prior position is stored as an edge property, making the full shift history queryable.
AI synthesis tags each response with the Concept nodes from the source graph that it engages with. Concepts with no ADDRESSES edges pointing at them are the gaps — surfaced in the synthesis as unexplored territory.
Example synthesis
After a ten-day discussion on ocean literacy and indigenous coastal knowledge, seeded by the Knowledge Garden. Convergence question: "Where should scientific and indigenous knowledge systems meet in ocean literacy education?"
Position distribution as of day 10. Five participants shifted from "scientific leads" to "co-equal start" during the discussion period.
This synthesis is generated on request, not continuously. It reads the full discussion against the concept graph from the originating source. It does not conclude — it illuminates. The conversation continues.
Augur and the Knowledge Garden are designed together. When a video is submitted to the Garden, it seeds both a discussion and an Augur convergence question — drawn automatically from the most consequential tension the source touches.
About the Knowledge GardenAugur is part of the Confabula platform, built by the College of Exploration. The platform is open to communities whose work requires genuine collaborative intelligence.