🚨 PhD Position at the University of Amsterdam 🚨 Join my team as a computer scientist / computational social scientist working on LLMs, social media, and politics. We offer freedom, impact, and an inspiring environment at one of Europe's leading universities. 🔗 werkenbij.uva.nl/en/vacancies...

In the literature, there are two competing explanations for "echo chambers": 1️⃣ Algorithms curate what we see (“filter bubbles”) 2️⃣ People choose like-minded peers (“selective exposure”) Our new study suggests something surprising: both explanations might be wrong. 🧵 arxiv.org/abs/2508.10466

Ever since I started thinking seriously about AI value alignment in 2016-7, I've been frustrated by the inadequacy of utility+RL theory to account for the richness of human values. Glad to be part of a larger team now moving beyond those thin theories towards thicker ones.

🔔 Often our tech policy interventions operate from linear, top-down assumptions that don't account for the complexity they seek to govern. To dig into this I'm co-organizing a workshop at ACM CI 2025 with Jason Burton, Joe Bak-Coleman, + Naomi Shiffman. You should come! ci-x-tp.github.io

Proposals for Build Peace (arguably the main conference on digital peacebuilding) are now open. This year it's near Barcelona in November. Consider applying! https://howtobuildpeace.org/attend-the-conference/register/

🚨 WEBINAR ALERT 🚨 Join KGI on March 25th for a live discussion on designing algorithmic feeds that put people first. As legislation and litigation around algorithms heats up, it’s never been more important to learn how they can be improved.

EVENT: Join us for Artificial Intelligence and Democratic Freedoms on April 10-11 at @columbiauniversity.bsky.social & online. Hosted with Senior AI Advisor @sethlazar.org. Co-sponsored by the Knight Institute & @columbiaseas.bsky.social. Panel info in 🧵. RSVP: knightcolumbia.org/events/artif...

Connected by Data are seeking examples of participatory digital governance to map out such projects around the world. https://connectedbydata.org/blog/2025/03/05/participatory-digital-governance

🔔 (new paper!) You might have heard of the "Habermas Machine", an AI-human pipeline that is really good at finding common ground between ideologically diverse groups, at least in lab settings. But can this kind of approach help in real world conflicts?

Thanks to TIME and Tharin Pillay for this great coverage of our work with @audreyt.org, @mchangama.bsky.social, @lukethorburn.com, Divya Siddarth and Emilie de Keulenaar: time.com/7258238/soci... Read the full paper here: www.arxiv.org/abs/2502.10834

Live near Chicago? Or going to ISA 2025? Want to geek out on conflict theory + coauthor a landmark paper? We're finding all the ways people distinguish between "good conflict" and "bad conflict" — but we need help, so we're hosting a workshop! You should come! https://forms.gle/BKxziu8zbZ1oZZqU8

Re-aligning SOCIAL MEDIA & AI: "Whom does it serve? That is the overarching issue as we confront "enshittification" & threats to democracy. Subject of 4 publications covered in this thread. 🧵 Most recent applies to PERSONAL AI AGENTS - "Have your AI call my AI" www.techpolicy.press/new-perspect...

Register your interest in our LLMs & Public Discourse: Convening & Workshop by mid-December. This exclusive, invite-only event will occur in Berkeley, California, February 27-28, 2025. Complete the interest form here: docs.google.com/forms/d/e/1F...

Build Up are hiring a Digital Peacebuilding Associate! You should apply. They're the best. https://docs.google.com/document/d/1WH9-zE7tt7EsFe-W_DFAhX3MGuhloiSSAhf0c_y4G_w/edit?usp=sharing

Today the European Commission proposed how Art.40 of the Digital Services Act (#DSA) could work in practice.In a worldwide first, this article in the DSA mandated very large platforms to grant researchers access to a wide range of previously undisclosed data.Some key points - now also on Bluesky :)👇

Reframing the issues of governing social media. Human discourse is a social process that depends on three pillars: agency, mediation, & reputation. Without these three strong pillars, + middleware enabling open innovation and interoperation, social media will struggle to balance chaos and control.

Human discourse is a social process, writes @rreisman.bsky.social , that depends on three pillars: agency, mediation, and reputation. But without strong pillars and middleware enabling open innovation and interoperation, social media will likely struggle to balance chaos and control.

Applications are open for fellowships at the Applied Social Media Lab at BKC, Harvard: https://asml.cyber.harvard.edu/asml-fellowship-program/

There are opportunities for regulators to dive deeper into less-emphasized areas for intervention around social media, write Nathaniel Lubin, Kalie Mayberry, Manon Revel, and @lukethorburn.com:

"we need people in our lives, not the simulation of people" https://theconvivialsociety.substack.com/p/embracing-sub-optimal-relationships

The social media platform Sparkable is implementing bridging-based ranking! https://blog.sparkable.cc/building-the-first-bridging-based-social-media-platform/

New Facebook deactivation study (in France) just dropped. https://www.jonathanmladd.com/uploads/5/3/6/6/5366295/effects_of_facebook_access_french_presidential_election-_mpsa_2023.pdf

YouTube is experimenting with a Community Notes-style feature using a bridging algorithm! https://blog.youtube/news-and-events/new-ways-to-offer-viewers-more-context/

"Experiments are the Best Kind of Transparency" New piece in @techpolicypress.bsky.social with Jonathan Stray and Pri Bengani (@acookiecrumbles.bsky.social)! www.techpolicy.press/experiments-...

A new law, the "Prohibiting Social Media Manipulation Act" in Minnesota, requires social media platforms to release the results of product experiments they run. https://www.revisor.mn.gov/bills/text.php?number=SF4097&version=latest&session=ls93&session_year=2024&session_number=0

Are you a funder looking to support basic research into the societal impacts of algorithms? This project (for testing alternative recommenders during the US election) by Jonathan Stray et al. is a great bet. https://humancompatible.ai/news/2024/05/23/sponsor-a-better-social-media-algorithm/

One hypothesis why many attempts to measure the impacts of social media find small or null results is that the mean effects are small, but that effects on some individuals are much larger. Enter this paper: https://www.nature.com/articles/s41598-024-55064-y

There's a workshop on social media ethics with cool people happening at Oxford next week! Sorry I'll miss it. https://www.oxford-aiethics.ox.ac.uk/event/workshop-social-media-corporations-risks-rights-and-responsibilities-magdalen-college-oxford

This is such a great case—really excited that @knightcolumbia.org gets to litigate this on behalf of @ethanz.bsky.social. www.nytimes.com/2024/05/05/o...

The Perspective API from Jigsaw now includes classifiers for bridging! Affinity, compassion, curiosity, nuance, personal stories, reasoning, and respect. https://medium.com/jigsaw/announcing-experimental-bridging-attributes-in-perspective-api-578a9d59ac37

New evidence that optimizing for engagement (specifically clicks) creates an incentive for negativity in online news. https://www.nature.com/articles/s41562-023-01538-4

The new Knight–Georgetown Institute are hiring an Associate Director to work on tech policy. Would be a cool job. https://kgi.georgetown.edu/work-with-us/318-2/

Survey from the Polarization Research Lab suggests that 2% of USA-ians support murder of individuals from opposing political parties. This seems too high. Hopefully just Lizardman's constant (insincere responses). https://prlpublic.s3.amazonaws.com/reports/political_violence_comparative_public.html

I like this genre of journalism that models how divisive topics can be navigated with humanity. What are other examples? https://www.theguardian.com/lifeandstyle/series/dining-across-the-divide

Cool new paper uses LLMs to correct misinformation by Tim Althoff, Amy Zhang, and others: arxiv.org/pdf/2403.111...

Sunday is your last day to apply to join our editorial team. Find the details of the vital role an editor plays in our organisation here: www.bellingcat.com/editor-posit...

There's a research scientist role focused on societal impacts open at Anthropic. https://jobs.lever.co/Anthropic/16ec5111-778c-4a89-9c74-6d7c29091a74

Excited to dive deeper into the AT Protocol and share what I’ve learned — then hack away with others. I’ve got some ideas brewing. Top contender is a bridging-based ranking feed, inspired by: bridging.systems #BridgingSystems

Bluesky sees this first 🫣 my Mozilla Creative Media Award project with @lujain.bsky.social is live; check out the announcement! foundation.mozilla.org/en/blog/the-...

A new interactive game, "The Algorithm", by Alia ElKattan & Lujain Ibrahim helps build intuitions about how recommenders operate. https://whatsthealgorithm.com/ This kind of work to make technical concepts accessible and engaging is super valuable! Looking forward to what they build next.

There's an opening for a post doc at the Polarization Lab at Duke University (with Chris Bail et al.). https://academicjobsonline.org/ajo/jobs/27242

Last year a day-long "NonEngagementCon" was held to pool insights from social media platforms on using non-engagement signals in recommenders. The findings have just been published — it's rare to get this much industry insight shared publicly. https://arxiv.org/pdf/2402.06831.pdf

New study comparing the effects of bridging-based ranking with engagement-based ranking and other algorithms. From Jason Burton, Stefan Herzog, and Philipp Lorenz-Spreen. https://osf.io/preprints/psyarxiv/5yfbt

Amy Zhang, Michael Bernstein and others have cool new framework — the 'Form-From' framework — for mapping the design space of social media platforms. https://arxiv.org/pdf/2402.05388.pdf

X is hiring for two positions to work on Community Notes. https://twitter.com/CommunityNotes/status/1752056816014643321

A study of what happens when people feel outnumbered by people who disagree with them online — the opposite of an echo chamber. https://osf.io/preprints/osf/tygec

Even when the *objective function* of a recommender system is perfectly aligned with user welfare, the *learning algorithm* can still cause negative impacts. https://arxiv.org/abs/2401.05304

Study on the effect of showing 'misleading counts' as well as 'like counts' on social media posts. https://royalsocietypublishing.org/doi/10.1098/rstb.2023.0040

Bertelsmann Stiftung have published an excellent visual explainer on bridging algorithms! https://bridging.reframetech.de/en Includes a side-by-side comparison of bridging and engagement-based approaches that does a great job of conveying how much less divisive bridging-based ranking could be.

Is anyone keeping a database of political deepfake incidents?