Natasha Goel

Can ChatGPT change your mind? I wrote for @techpolicypress.bsky.social about what we find + where AI persuasion research should go from here:

How will trust in AI evolve? What can it really persuade us on? Will it reach resistant audiences?

Can ChatGPT change your mind? I wrote for @techpolicypress.bsky.social about what we find + where AI persuasion research should go from here:

How will trust in AI evolve? What can it really persuade us on? Will it reach resistant audiences?

Why Don’t Russian Soldiers Revolt? Astonishing death rates and brutal abuse have not kept troops from following orders.

My first Foreign Policy essay is out - it examines the role of violence in Russia's military, which both creates docility in the face of meatgrinder tactics, and encourages increased violence towards fellow soldiers and Ukrainian prisoners of war and civilians. foreignpolicy.com/2025/04/09/r...

Partisans often seem unwavering in their support for a politician/policy, even when faced with opposing evidence. But recent studies show that partisans can be persuaded. So how can both be true? My new @bjpols.bsky.social ky.social paper explores this Q: doi.org/10.1017/S000...

If you're at #MPSA2025 and interested in AI as a tool for persuasion, I will be presenting our paper comparing effects of AI and human source cues in reducing certainty in false beliefs on Saturday at 5:10pm (Political Knowledge, Palmer House 7th floor)

Pre-print: osf.io/preprints/os...

🚨 New paper out w/ @ericmerkley.bsky.social in Political Behavior, "The Nature of Online Talk: Incivility of Opposing Views and Affective Polarization"

Link to paper: link.springer.com/article/10.1...

Short thread below:

PDF can be found here: rdcu.be/d4U8M

In terms of co-partisans, although it was cognitively satisfying to see belief reinforcing-information, there was a limit: individuals recognized norm-defying incivility and punished it accordingly.

We see that incivility is consequential, but not always in the ways we expected. In terms of the out-partisans online, ppl likely have some expectation of civility, absent other information, and their expectations can be violated with minimal exposure.

Interestingly, we find that people do punish the incivility of co-partisans. The presence of the control here suggests that respondents rewarded co-partisans for being critical of the out-party, but this benefit is extinguished when they are expressed in an uncivil manner.

We do, however, see some effect of the uncivil treatment on broader out-party evaluations (feeling thermometer).

Point allocation difference between uncivil out-party profiles and civil ones were not significant. Comparison with the control (no tweets) suggests that ppl may punish out-partisans for any opposing views, regardless of how they are stated.

All participants were Player 1 in our trust game. Participants were assigned to either control (demographics + no Tweets), civil condition (demographics + civil Tweets), or uncivil condition (demographics + uncivil Tweets). They played 2 rounds of the game (out-party & in-party).

🚨 New paper out w/ @ericmerkley.bsky.social in Political Behavior, "The Nature of Online Talk: Incivility of Opposing Views and Affective Polarization"

Link to paper: link.springer.com/article/10.1...

Short thread below: