iiinova — where your digital self keeps evolving

Reposted by Do Won Kim

Jamie Cummins @jamiecummins.bsky.social · 14/07/2026

Today we launch the first stable release of RegCheck: v1.0.0. RegCheck makes it easier and quicker to compare study registrations to published papers for consistency - something we know is important in principle, but rarely done in practice. A 🧵 on what's new: regcheck.app

regcheck.app

RegCheck

RegCheck is an AI tool to compare preregistrations with papers instantly.

512883

Do Won Kim @dowonkim.bsky.social · 16/06/2026

10/ Questions and feedback are welcome! With my wonderful coauthors ✨: @ozgurcanseckin.bsky.social @glciampaglia.com @baottruong.bsky.social Saumya Bhadani & Alessandro Flammini

001

Do Won Kim @dowonkim.bsky.social · 16/06/2026

9/ Most effects faded after about a month. So durable depolarization likely requires repeated exposure. Still, our findings suggest that challenging partisan expectations may be a key mechanism through which political conversations reduce polarization.

131

Do Won Kim @dowonkim.bsky.social · 16/06/2026

8/ There was also a tradeoff. Expectation-challenging conversations reduced polarization, but participants generally found them less satisfying and were less eager to have similar discussions in the future.

121

Do Won Kim @dowonkim.bsky.social · 16/06/2026

7/ Importantly: People did not substantially change their own policy positions. These conversations reduced polarization without changing minds. Instead, they changed how people perceived and felt about partisan groups.

141

Do Won Kim @dowonkim.bsky.social · 16/06/2026

6/ But the mechanism depended on which expectation was challenged. An agreeing outgroup member made people: → feel warmer toward outgroup members → perceive less distance from them A disagreeing ingroup member made people: → feel less warmth toward ingroup members → perceive more distance from them

121

Do Won Kim @dowonkim.bsky.social · 16/06/2026

5/ Expectation-challenging conversations had lower: - affective polarization - perceived issue polarization relative to expectation-confirming conversations.

121

Do Won Kim @dowonkim.bsky.social · 16/06/2026

4/ Participants had a political conversation with an AI chatbot whose partisanship (in- vs out-group) and issue stance (agree vs disagree) were independently manipulated. ➡️ 4 conditions: - Ingroup Agree, Outgroup Disagree (expected); - Ingroup Disagree, Outgroup Agree (expectation-challenging)

121

Do Won Kim @dowonkim.bsky.social · 16/06/2026

3/ Partisans often assume: - Democrats and Republicans disagree - Co-partisans agree - Political identity predicts political opinions So what happens when those expectations are challenged? We tested this in a preregistered experiment with 1,983 U.S. partisans.

121

Do Won Kim @dowonkim.bsky.social · 16/06/2026

2/ Political conversations are often proposed as a remedy for polarization. But the evidence is mixed. Some conversations reduce polarization. Others do little. Some even backfire. Why? We argue that a missing piece is the expectations people bring into the conversation.

121

Do Won Kim @dowonkim.bsky.social · 16/06/2026

1/ Excited to share our new paper: Challenging Partisan Expectations Reduces Political Polarization We find that political conversations reduce polarization most when they challenge what people expect about partisan boundaries. 🧵 Paper: arxiv.org/abs/2606.15901 #polisky

arxiv.org

Challenging Partisan Expectations Reduces Political Polarization

Political conversations are often proposed as a remedy for political polarization, yet their effectiveness remains inconsistent. We argue that this inconsistency partly reflects a neglected feature of...

174

Do Won Kim @dowonkim.bsky.social · 16/06/2026

11/ Questions and feedback are welcome! With my wonderful coauthors: @ozgurcanseckin.bsky.social @glciampaglia.com @baottruong.bsky.social Saumya Bhadani Alessandro Flammini

000

Do Won Kim @dowonkim.bsky.social · 16/06/2026

10/ Most effects faded after about a month. So durable depolarization likely requires repeated exposure. Still, our findings suggest that challenging partisan expectations may be a key mechanism through which political conversations reduce polarization.

100

Do Won Kim @dowonkim.bsky.social · 16/06/2026

9/ There was also a tradeoff. Expectation-challenging conversations reduced polarization, but participants generally found them less satisfying and were less eager to have similar discussions in the future.

100

Do Won Kim @dowonkim.bsky.social · 16/06/2026

8/ Importantly: People did not substantially change their own policy positions. These conversations reduced polarization without changing minds. Instead, they changed how people perceived and felt about partisan groups.

100

Do Won Kim @dowonkim.bsky.social · 16/06/2026

7/ But the mechanism depended on which expectation was challenged. Agreeing outgroup member made people feel warmer toward outgroup members and perceive less distance from them. Disagreeing ingroup member made people feel less warmth toward ingroup members and perceive more distance from them.

Effects of AI conversations on political polarization.
Each plot shows average treatment effects for the Outgroup Agree (purple circle), Ingroup Disagree (purple diamond) and Outgroup Disagree (orange diamond), relative to the Ingroup Agree condition (gray dashed line), with 95% confidence intervals. Non-significant results (q > .05) are indicated by lighter colors and unfilled markers.
(A)~Expectation-challenging conversations (Outgroup Agree and Ingroup Disagree) and conversations in the Outgroup Disagree condition reduce affective polarization.
(B,C)~The reduction in affective polarization in the Outgroup Agree and Outgroup Disagree conditions is primarily driven by warmer feelings toward the outgroup. In the Ingroup Disagree condition, the decrease is driven by colder feelings toward the ingroup.
(D)~Expectation-challenging conversations also reduce perceived issue polarization.
(E,F)~The reduction in perceived issue polarization in the Outgroup Agree condition is driven by a decrease in perceived outgroup distance, while in the Ingroup Disagree condition it is largely driven by a increase in perceived ingroup distance.
(G)~We find no evidence that challenging partisan expectations persuade participants either by moderating or strengthening their previous attitudes on the issue.

100

Do Won Kim @dowonkim.bsky.social · 16/06/2026

6/ Expectation-challenging conversations had lower: • affective polarization • perceived issue polarization relative to expectation-confirming conversations.

(I) Study design. We conduct two waves of survey. The first wave included a pre-treatment questionnaire, random assignment to the treatment --- an AI conversation that was either expectation-confirming (Ingroup Agree and Outgroup Disagree) or expectation-challenging (Ingroup Disagree and Outgroup Agree) --- and a post-treatment questionnaire measuring the main outcomes of the study.
After a minimum period of one month, participants were invited to a follow-up survey measuring again only the main outcomes of the study. The figure shows an example of the elicited issue of one of the participants assigned to the Ingroup Agree condition, the actual response from the chatbot (A), and as a counterfactual, what the chatbot would have responded in the other conditions ((B)--(D)).

(II) Expectation-challenging versus expectation-confirming conditions. The figure shows linear contrasts comparing expectation-challenging conditions (Ingroup Disagree and Outgroup Agree) with expectation-confirming conditions (Ingroup Agree and Outgroup Disagree) for polarization outcomes. Points indicate contrast estimates comparing expectation-challenging and expectation-confirming conditions. Negative estimates indicate lower values under expectation-challenging conditions; positive estimates indicate higher values under expectation-challenging conditions. Filled markers denote contrasts with BKY sharpened two-stage FDR-adjusted q <=.05, whereas hollow markers denote q > .05. Reported q-values are shown on the right. This analysis is exploratory and was not preregistered.

100

Do Won Kim @dowonkim.bsky.social · 16/06/2026

5/ This created four conditions: ✓ Ingroup Agree ✓ Outgroup Disagree (expected) and ✓ Ingroup Disagree ✓ Outgroup Agree (expectation-challenging)

100

Do Won Kim @dowonkim.bsky.social · 16/06/2026

4/ We tested this in a preregistered experiment with 1,983 U.S. partisans. Participants had a structured political conversation with an AI chatbot whose: partisan identity (ingroup vs outgroup) and policy stance (agree vs disagree) were independently manipulated.

100

Do Won Kim @dowonkim.bsky.social · 16/06/2026

3/ Partisans often assume: - Democrats and Republicans disagree - Co-partisans agree - Political identity predicts political opinions So what happens when those expectations are challenged?

100

Do Won Kim @dowonkim.bsky.social · 16/06/2026

100

Reposted by Do Won Kim

William J. Brady @williambrady.bsky.social · 27/05/2026

As I mentioned in the below thread, this project involved many feats of engineering, led by the fantastic @markptorres.bsky.social. If you're a CS or CSS person interested in the gory details, see his blog post: markptorres.com/research/202...

markptorres.com

How we built the infrastructure for a large-scale social media field experiment during the 2024 US election

What we built

0113

Reposted by Do Won Kim

Sarah Shugars @shugars.bsky.social · 13/08/2025

Very cool pilot study from @dowonkim.bsky.social using chatbots to examine how people respond to out-group agreement and in-group disagreement. Looking forward to seeing this project develop into the full study! #pacss2025 #polnet2025

041

Reposted by Do Won Kim

Gordon Pennycook @gordpennycook.bsky.social · 15/05/2025

Recent research shows that AI can durably reduce belief in conspiracies. But does this work b/c the AI is good at producing evidence, or b/c ppl really trust AI? In a new working paper, we show that the effect persists even if the person thinks they're talking to a human: osf.io/preprints/ps... 🧵

814646

Reposted by Do Won Kim

Misha Teplitskiy | Science of science | on LinkedIn mostly @innovation.bsky.social · 15/05/2025

Conventional wisdom says interdisciplinary research is valuable but harder to get through peer review (need to please diverse reviewers, etc). @sdxiang.bsky.social Daniel and I partnered with @ioppublishing.bsky.social to test this wisdom and add nuance

1219

Reposted by Do Won Kim

John Holbein @johnholbein1.bsky.social · 21/04/2025

What happens to people when researchers encourage them to deactivate Facebook or Instagram for a few weeks in the lead up to an election? People get modestly* happier/less depressed/less anxious. "The Facebook effect is driven by people >35, while the Instagram effect is driven by women <25."

55416

Reposted by Do Won Kim

Andrew Trexler @atrexler.com · 03/04/2025

New paper with @dianamejordan.bsky.social and sky-less Trent Ollerenshaw! We provide large-N tests of repeated measure designs in survey experiments, showing that they slightly attenuate ATEs relative to post-only designs, but provide large gains to precision. Thread below. Preprint: osf.io/q6czp

Figure 4 providing a graphical summary of the main results of the study: a slight attenuation of average treatment effects from a repeated measure design, and large reduction in the ATEs' standard errors (improved precision).

3329

Reposted by Do Won Kim

Christopher Barrie @cbarrie.bsky.social · 25/03/2025

📄NEW PAPER📄 Ever wondered content people actually pay *attention* to online? Our new research reveals that you likely pay attention to far more varied political content than your likes and shares suggest

812343

Reposted by Do Won Kim

Alberto Acerbi @acerbialberto.com · 19/12/2024

Interesting results, I wonder if stated preferences match behaviours, e.g., how many people are using bsky custom feeds?

241