Skip to content

Collective Intelligence

⚠ This page is a stub and needs development. It currently documents the concept primarily through DOD's own discussions, which are drawn from Western liberal-democratic reform contexts. It would benefit from: the wider academic literature on CI (Surowiecki, Lévy, Heylighen); examples of CI failure (how digital platforms can degrade collective reasoning through filter bubbles and tribal epistemology); and non-Western or indigenous collective knowledge traditions. Contributions welcome — see how to get involved.

Collective intelligence (CI) is the capacity of a group to know, decide, or coordinate more effectively than any individual within it. It is distinct from artificial intelligence — a distinction explicitly argued in DOD discussions (see below). CI is an emergent property: it arises from the interactions between participants, not from any single participant's capability.

Emergence and scale

Alexar Pendashteh, speaking at the December 2022 Basil's Table panel, offered an emergence-layer account: CI is one level in a stack that runs from quantum → chemistry → biology → psychology → societal phenomena. Each layer is governed by properties not visible from the layer below. The internet, on this view, is creating a single information layer that could enable a genuinely new form of collective intelligence in human societies — analogous to:

  • Bacterial quorum sensing — bacteria coordinate collective behaviour through chemical signalling once a density threshold is crossed
  • Wolf packs — distributed decision-making without a central commander
  • Shared human myths — the mechanism Yuval Noah Harari describes for large-scale human cooperation before modern institutions

The argument Pendashteh makes is that the hard problem in democratic design is building platforms that activate CI rather than mere information aggregation — and that this is where AI-as-tool becomes genuinely useful: not replacing human judgement but creating the conditions in which collective human judgement can emerge.

CI vs AI in democratic contexts

At the same panel, Pendashteh explicitly distinguished CI from AI:

"The goal is collective intelligence, not artificial intelligence — the question is whether these tools raise or lower the CI of the group."

This framing reframes the standard AI-in-democracy debate. The criterion is not whether AI is "used" but whether the system's design, including any AI components, increases or decreases the group's emergent capacity for considered judgement.

Prediction markets as CI platforms

Hubertus Hofkirchner, presenting Prediki at the 2017 Citizens' Democracy event, framed prediction markets explicitly as collective-intelligence infrastructure: markets that aggregate dispersed private knowledge by rewarding accuracy, moving the price toward what the crowd-as-whole knows better than any expert. The Prediki platform added argument capture and sentiment analysis on top of the market mechanism, attempting to surface not just the probability estimate but the reasoning behind it.

Nicholas Gruen's response at the same event drew a boundary: prediction markets surface CI on what will be (factual, resolvable questions); they cannot resolve what shall be (value questions) — which he argued requires a deliberating citizens' jury.

Trust, feedback loops, and CI preconditions

Alexar Pendashteh's analysis in the 2019 DOD Trust discussion approached CI from the preconditions side: trust is the substrate that enables collective intelligence to emerge. Without trust, groups cannot share information freely, cannot update beliefs cooperatively, and cannot coordinate at the scale CI requires. The "feedback loop" insight discussed at that event — the people who are influenced are not the people who are influencing — identifies one structural barrier to CI in large democratic systems.

See also

DOD discussions