Opinion for collective intelligence

Abstract

This short piece shares thoughts on some recent research and books related to collective intelligence - on topics ranging from democracy and institutions to LLMs and animals.

Collective intelligence stimuli – part of a series of comments by editors on what they’ve found interesting recently

The collective intelligence field is simultaneously widening, deepening, and running into some interesting dilemmas, which is good as all fields thrive as much through questions as through answers. I recently wrote a review of some interesting recent books on collective intelligence which shows the range of current CI work: one on democracy (the big collection on CI and democracy edited by Stephen Boucher and others), one on animals (Martin Wikelski’s ‘The Internet of Animals’ on the ICARUS project), and one a popular introduction (which I was a bit critical of).

Experiment around democracy and collective intelligence remains very lively (despite the countertrends!), with some good recent books (e.g. Plurality by Audrey Tang and Glen Weyl) and a flood of new material on how best to combine AI and CI (e.g. this on how AI can help with consensus and this broader review of AI/CI). New tools like SocraSynth and DeepMind’s Habermas Machine are pointers to a future where CI and AI complement each other to dramatically speed up intelligent reasoning.

On another front, John Kay’s ‘The Corporation in the 21st Century’, one of the best recent books on business, has a lot in it about collective intelligence – and how essential CI is for understanding both firms themselves and the ecosystems they thrive in. Many economists still struggle to absorb the insights of CI – so it’s refreshing to see this presented as common sense.

I loved Robert Sapolsky’s recent book ‘Determined', which isn’t about CI as such, but is a reminder of the non-centred nature of thought within the individual brain, which functions without a single locus of free will. As that becomes a mainstream view, it will be easier for people to realise that societies and organisations are similar: assemblies of sometimes competing and sometimes cooperating modules rather than being guided by a central control room, and increasingly supported by LLMs that can enhance questioning, collaboration, and decisions.

This is also relevant to one of my current areas of work: the theory and practice for designing next generation public institutions in cities, nations, and globally (which is being done with the UNDP, various governments and collaborators). One premise is that the organisation of intelligence needs to become much more central to institutional design. We’ve done many case studies, from Careblocks in Colombia to AESIA, the new AI regulator in Spain, from China’s tech investment Government Guidance Funds to Digital Public Infrastructures in India. One of my favourites is the International Solar Alliance, a prime example of an ‘intelligence assembly’ that orchestrates data, evidence, tacit knowledge and more, to accelerate solar implementation in the global south. Another example is a report I wrote with Ales Cap recommending how to use a combination of collective and artificial intelligence to combat the risk of elections being disrupted by deepfakes and misinformation: part of a much broader programme to map out a path between the chainsaws of Elon Musk on the one hand and defence of the status quo on the other.

But CI is also throwing up important dilemmas. In democracy, how to balance expertise and citizen input (where the answers are bound to vary depending on the topic)? How to design against collective stupidity and the folly of crowds? How to understand the impact of scale on the forms and functions of CI? And how to reconcile the limited bandwidth of human intelligence with the fire hoses of information we’re now subjected to and the limitless capabilities of AI? These and many other questions should keep us all busy.