Runway to build a METR-style research organization for AI's societal impacts, starting with a January event at Collider and a research pilot with SPAR
I'm building SIRA (Societal Impacts Research on AI). I want it to be a METR or Redwood for the societal side of AI, i.e., an independent research org measuring what happens to work, how people form beliefs and make decisions, to institutions, and to who holds power, as transformative AI is used everywhere.
Right now, frontier labs study their own products, but most of the research is about the systems themselves. Few orgs are tracking the effects on society, and I'm bullish that the next few years matter a lot here, because the choices that shape these effects may be hard to undo later.
This funding is a runway that lets me work on SIRA full-time through August 2027 while we set a research agenda, publish a first report, run a research pilot, launch a fellowship in June, and raise larger funding to hire SIRA's first full-time researchers.
Set a research agenda. Collider in New York is hosting our first event in January (more to follow on this). We're inviting ~10–15 researchers across economics, psychology, sociology, and AI safety to work out what SIRA should study first. Our preliminary ideas on focus areas can be found here.
Publish a first report. I've built a census of social science research on AI's societal effects, ~104,000 papers since 2015, up 17-fold from 2019 to 2025 (see it here). The report turns that into a short map of what we know, where the gaps are, and where we'll start.
Run a research pilot. We've applied to host 3–5 mentored projects in SPAR's spring round (February–May), across our focus areas.
Launch the inaugural fellowship in June with 8–12 researchers, and use everything above to raise funding for an in-house research team.
I'll know this worked if, by the end of May, SIRA has gone from an idea to a working research organization with things people can read and act on. Concretely, at least 10 researchers come to the January event; we publish a research agenda (that updates / refines our current focus areas) by the end of February; our evidence report is published by the end of March; at least 3 SPAR projects finish with a public-facing write-up by the middle of May. Further success could be if a larger grant for the research team is submitted or under review by June.
SIRA isn't incorporated yet. I'm currently looking for a fiscal sponsor to hold SIRA's grants and donations while we incorporate as a nonprofit and apply for our own 501(c)(3) status. Until then, funds raised here will be held and paid out by Manifund. The operations line below covers the sponsor's fees and our incorporation filings. If you've worked with a fiscal sponsor (or are from one), I'd love to hear from you.
Max goal ($72,000):
My time, full-time through August 2027: $5,000/month × 9 months = $45,000
January event at Collider: travel for ~6 invited researchers (~$400 each), meals, and my travel; Collider provides the venue = $4,500
SPAR fellow stipends: ~4 part-time fellows × $1,000/month × 3 months = $12,000
Help with the evidence report: ~200 hours × $30/hour = $6,000
Operations: incorporation, fiscal sponsor fees, tools = $4,500
At the minimum ($15,000): two months of my time ($10,000), the January event ($3,000), and basic operations ($2,000).
I'm Emilio Barkett! I've spent the past two and a half years in AI safety. I mentor research projects at SPAR, facilitate at BlueDot Impact, and was a research manager and fellowship facilitator at the Columbia AI Alignment Club. More about me --> emiliobarkett.github.io.
Since starting SIRA in July, I've written up the proposal on LessWrong (the post is more focused on the fellowship side), built the census and the website, and put out a fellowship interest form: 311 people signed up in the three weeks since launch, including at least 44 PhD students and 16 postdocs. Four researchers have committed to mentor, including people at Stanford GSB and Oxford, and Collider offered to host our first event.
Roshni Lulla (cognitive neuroscientist, PhD from USC, co-founder of the Institute for Humane Robotics)
Alexander Kimpton (geopolitical risk analyst at Dragoman and PIBBSS fellow)
Yusuf Kundgol (operations at ARENA).
If we can't raise enough for the research team or fellowship, SIRA will stay small. The agenda, report, and pilot projects still get published, and I'd look for a partner organization to host the fellowship. If SPAR doesn't take our projects, I'll run a smaller pilot directly with our mentors. And right now a lot runs through me, which is part of why the budget includes help with the report.
$0 so far.
There are no bids on this project.
Comments
No comments yet. Sign in to create one!