Tom Costello
@tomcostello.bsky.social
research psychologist. beliefs, AI, computational social science. assistant prof at Carnegie Mellon
Conditionally accepted at the APSR (w/ @scottclifford.bsky.social & @patrickpliu.bsky.social): Why does political information so often change beliefs but NOT attitudes? We highlight the role of belief relevance, or the extent to which beliefs bear on attitudes.
Today's SJDM Featured Paper is: Costello, T. H., Pelrine, K., Kowal, M., Arechar, A. A., Godbout, J.-F., Gleave, A., Rand, D., & Pennycook, G. (2026). Large language models can effectively convince people to believe conspiracies. arXiv. doi.org/10.48550/arX...
Large language models can effectively convince people to believe conspiracies
Large language models (LLMs) have been shown to be persuasive across a variety of contexts. But it remains unclear whether this persuasive power advantages truth over falsehood, or if LLMs can promote...
doi.org
I saw we were fighting about what AI will do to (social) science. I have some thoughts (originally written for a book on scientific inquiry) and decided to turn them into my first substack article. Co-written with Nat Rabb! tomstello.substack.com/p/standing-o...
Standing on the Shoulders of Termites
Or, how we learned to stop worrying and love AI (scientists)
tomstello.substack.com
🧵on my new paper "Synthetic personas distort the structure of human belief systems" w Roberto Cerina I'm v excited about... 🚨 Do synthetic samples look like human samples? We compare 28 LLMs to the 2024 General Social Survey (GSS) to find out + develop host of diagnostics...
I'm hiring a postdoc at @cmu.edu (w/ far.ai & @dgrand.bsky.social + @gordpennycook.bsky.social)! How do LLMs shape human beliefs — and what do we do about it? AI safety meets behavioral science. Open to technical and social science backgrounds.
I'm hiring a postdoc at @cmu.edu (w/ far.ai & @dgrand.bsky.social + @gordpennycook.bsky.social)! How do LLMs shape human beliefs — and what do we do about it? AI safety meets behavioral science. Open to technical and social science backgrounds.
🚨 New paper out at @ajpseditor.bsky.social 🚨 Do the public hold meaningful attitudes? Using the case of abortion policy preferences, we provide strong evidence that policy prefrences can be coherent, stable over time, and causally explain vote choice. doi.org/10.1111/ajps...
New paper in Current Directions in Psych Science: journals.sagepub.com/doi/10.1177/... After countless arguments about what tasks ppl should/should not offload to AI, we instead argue that genAI can be used *augment* research protocols in novel ways. I.e. use AI to make better psych experiments!
Honored (and genuinely, wildly grateful and -- even more than I am grateful -- surprised) to share that our paper "Durably reducing conspiracy beliefs through dialogues with AI" received the @aaas.org Newcomb-Cleveland Prize (for the "most outstanding" paper in @science.org last year).
🚨New WP: Can an AI voter guide (grounded in information from a nonpartisan, fact-checked source) help voters’ decision making? 🚨 We built and evaluated an LLM-based chatbot that provided voting info in CA & TX (N=2,474) right before the 2024 election. 🧵👇
If the medium is the message, then the message of algorithmic content platforms is not expression or connection or freedom. It is salience above all else. This is what is happening to us on these platforms. They cannot behave any other way.
The #1 New York Times bestseller The Sirens’ Call is now in paperback! Chris Hayes masterfully explores how our focus has been commodified & manipulated, urging us to reclaim control over our lives & future. The Sirens’ Call is the big-picture vision we urgently need to offer clarity & guidance.
After years in academia, I’m exploring data science and research roles in industry. I'm a quant. social scientist (PhD Yale ’24, NYU) focused on causal inference, experiments, and large-scale data. Feel free to get in touch or share; all leads appreciated. dwstommes@gmail.com
Our latest collaboration with CMU, Cornell, and others: AI is just as effective at spreading conspiracy beliefs as it is at debunking them. We found a fix that works, showing that we need to make deliberate design choices. Full thread below.
Tom Costello@tomcostello.bsky.social · 7mo agoIf you tell an AI to convince someone of a true vs. false claim, does truth win? In our *new* working paper, we find... ‘LLMs can effectively convince people to believe conspiracies’ But telling the AI not to lie might help. Details in thread
If you tell an AI to convince someone of a true vs. false claim, does truth win? In our *new* working paper, we find... ‘LLMs can effectively convince people to believe conspiracies’ But telling the AI not to lie might help. Details in thread
If you tell an AI to convince someone of a true vs. false claim, does truth win? In our *new* working paper, we find... ‘LLMs can effectively convince people to believe conspiracies’ But telling the AI not to lie might help. Details in thread
A very interesting data-rich analysis of persuasion on digital media by @benmtappin.bsky.social. Recommend! open.substack.com/pub/benmtapp...
For Digital Mass Persuasion, Exposure Matters More Than Persuasiveness
If you are interested in understanding the mass persuasive impact of digital media content you should generally pay more attention to content exposure than to its persuasiveness.
open.substack.com
in 2025 the combination of the latest models and agentic harnesses like Claude Code (currently considered tools for coders) felt like a step change for software engineers and, it seems, for mathematicians as well. in 2026 this will hit every other white collar job xcancel.com/Aaroth/statu...
I’m less polite: the capabilities q is usually an active denial. in skilled hands, SOTA LLM tools are more capable today than most people think. assume they can do X and you can get through to the real question of what people will do with that and what that reveals about the pressures they’re under
I'm gonna resonant computing so hard that my adaptive pluriverse becomes as prosocial as a god damned golden retriever. resonantcomputing.org
The Resonant Computing Manifesto
Technology should bring out the best in humanity, not the worst—a manifesto for resonant computing built on five principles that reject hyper-scale extraction for human flourishing.
resonantcomputing.org
Our new study provides rare causal evidence about NYC’s speed camera program. We find large reductions in collisions (30%) and injuries (16%) near intersections with cameras. www.pnas.org/doi/abs/10.1... @astagoff.bsky.social ky.social @brendenbeck.bsky.social nbeck.bsky.social 🧪
Can speed cameras make streets safer? Quasi-experimental evidence from New York City | PNAS
Each year, approximately 40,000 people die in vehicle collisions in the United States, generating $340 billion in economic costs. To make roads saf...
pnas.org
IMO, if we want to live in a democracy, we ought to embrace persuasion. When we treat all AI info as bad we are implicitly arguing that voters are incapable of processing arguments. This is an illiberal stance. democracy relies on the "unforced force of the better argument." (Habermas)
Any headline that ends in a question mark can be answered by the word “no” (Betteridge's law). Should we worry about gen AI persuasion shaping elections? I think “no” holds here. See knightcolumbia.org/content/dont... for a good summary of why we shouldn’t be too concerned
Don’t Panic (Yet): Assessing the Evidence and Discourse Around Generative AI and Elections
knightcolumbia.org
Joan from the Bronx@joanfromthebronx.bsky.social · 8mo agoThis is scary stuff.
I encourage folks to read the actual paper — the models were prompted to change opinions…not to be accurate. Humans also use inaccurate information. But the AIs were still reasonably accurate (Avg 77/100), with no evidence inaccurate claims are more persuasive.
David Ingram@davidingram.bsky.social · 8mo agoNEW: AI chatbots used inaccurate information to change people’s political opinions, study finds www.nbcnews.com/tech/tech-ne...
New: AI chatbots can change voters' minds, according to a pair of in-depth studies published just now in Science and Nature. How they do it is interesting — and concerning. Gift link: wapo.st/49RSstP
Voters’ minds are hard to change. AI chatbots are surprisingly good at it.
New research suggests AI chatbots can shift people’s political views more effectively than campaign ads on TV.
wapo.st
Very excited this paper is live! Congratulations to first author Hause Lin for being such a badass researcher and all around cool guy
David Rand@dgrand.bsky.social · 8mo agoIn Nature: rdcu.be/eTcbQ In exps with 🇺🇸 🇨🇦 🇵🇱 voters, an AI advocating for one presidential candidate or the other - and shifted attitudes by 4–10+ points 🔹Mechanism= "facts" and information (telling AI to not use facts knocks out persuasion) 🔹Right-leaning bots made more inaccurate "facts"
New paper in press at JPSP! An adversarial collaboration focusing on a large-scale test of how strongly implicit racial attitudes predict discriminatory behavior. Pre-print here: osf.io/preprints/ps...
Excited to share work led by Conrado Eiroa-Solans, a PhD student at UC Berkeley who I first met when he was applying to grad school. His new paper is out in Computers in Human Behavior Reports. /1
From Extrinsic to Intrinsic Motivation: Testing an AI-powered Motivational Interviewing System to Foster Prosocial Motivation
Scalable interventions promoting sustained behavioral change are crucial for addressing societal issues, yet traditional approaches often require inte…
sciencedirect.com