The AI for Collective Sensemaking Unconference

on liberating social data, building AI tools to produce collective action, and human flourishing

I care a lot about how we as a society figure out the truth together, and how we make decisions together. I call it collective sensemaking.

Illustrating different layers where collective sensemaking happens. You can also factor it by the size of groups, it can happen at the scale of the individual, as dyads (2 people), at the level of squads (~5 people) , Dunbar scale (~150 people), City scale (7000+), etc.

To that end, I think there’s lots of juice in applying AI to social data.

Over the last year I’ve met founders and researchers tackling the same hyperobject from different sides and I wanted to get them in the same room so I did!

We all suspect that we’re working on complementary things but what is the underlying shape/

We hosted the AI for Collective Sensemaking Unconference on the 26th day of October, in Brookyn, New York. In unconference, participants offer their own talks and events by posting them on a schedule. Around 60 people came, and 15 brilliant talks and workshops were offered!

What we got up to

This field or movement is not quite legible yet, enjoying little hype, meaning our participants tended to be independent thinkers and passionate enough to find their own way to and invest their efforts in the space of collective sensemaking.

We we were treated to brilliant high-signal presentations with clear synergy between them! Below are some highlights with links to the slides of each speaker.

• Brandon Dunderstad from Calcifer Computing presented his work pioneering the field of Computational Semiotics, studying the spread and distortion of signs with computational methods.

• I talked about the Community Archive and Epistemic Garden, and what we can do applying AI to social data.

• Swapneel Mehta presented Arbiter, by SimPPL, an impressively featured set of tools to understand social media discourse.

• Ivan Vendrov teased his Applied Organic Alignment Lab and shared helpful frames for thinking about AI for coordination. I was struck by how he started by establishing some invariant constraints to human cooperation, like the Dunbar number, or the maximum human language bandwidth of 40 bits/sec.

• Seref presented Index Network, a protocol for matching people based on their intentions, unlocking deals, connections, and serendipity.

• Andrew Miller talked about liberating social data with Trusted Execution Environments (TEEs). It seems like TEEs will be critical for multiplayer coordination apps that work on sensitive data so I was very excited to connect the TEE community and the rest of the AI for collective sensemaking space.

• I was also excited about Raymond’s presentation of his fast twitter archive explorer that lets you take archives from community archive and send them to LLMs for analysis in bulk.

My interpretation of the conference

The energy was palpable. As far as I know, it was the first time many of these people were meeting their peers in person, with opportunities to discuss and coordinate!

Just for fun, this is roughly how I think the presented projects relate

I was especially pleased with connecting two clusters: people coming more from an AI and collective sensemaking background, and people coming from the side of protocols, cryptography, and data liberation. I think it’s a match made in heaven.

If I had to guess how the rough clusters of work we saw interrelate.

The conference was co-hosted by my lab Epistemic Garden and Calcifer Computing and graciously received by the Fractal Tech Hub in NYC.

To the participants, thank you so much for accepting my invitation and for your brilliant work.

group pic at 8pm, at about 50% of total attendants

Closing thoughts

As a final point, I believe there are abundant public goods in applying AI to social data, and that we must be sovereign over the data we generate.

If you resonate, I hope you will consider contributing your data to the community archive, or support it on opencollective.