🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

When Community Notes inform users about falsehoods on X posts, the replies to the post have more negativity, anger, distrust, and moral outrage, finds Yuwei Chuai Anastasia'Asya'Sergeeva Lenzini Nicolas Pröllochs doi.org/10.1145/370659…

GPT-4's responses align more with left-wing political values compared to the average American, with degree and direction varying by topic, finds Fabio Motoki et al. doi.org/10.1016/j.jebo…

People negatively evaluate AI moderators that use emotional arguments rather than rational arguments, finds Silver et al. doi.org/10.1145/370659…

Fascinating look at differential exposure to election fraud claims, both questioning and supporting the false claim, by Mathieu Lavigne Fogarty John Carey Brendan Nyhan (@BrendanNyhan on 🟦☁️) Jason Reifler business.columbia.edu/sites/default/…

In co-developing conspiracy theories, Reddit sustains narratives while Twitter amplifies emotional content and broadens the audience, finds Dror Walter et al., in the case study of intermedia agenda setting of the Wayfair conspiracy doi.org/10.1177/146144…

Misinformation debunking messages from a public health institution were more effective than AI recommendations; In the US, peer messages enhance source and message evaluations among conspiracy believers, finds Dr. Xinzhi Zhang Winson Peng Dr Qinfeng Zhu doi.org/10.1177/107769…

Fascinating look at computational analysis of visual framing by Arnav Arora Srishti@CVPR’25 Maria Antoniak Serge Belongie Isabelle Augenstein doi.org/10.48550/arXiv…

In-person grassroots communications with politicians are 47x more impactful than social media communications, finds Lee Kaiping Chen (@kaipingchen.bsky.social) Marble & Curtis Bram, highlighting how social media can increase engagement but may be discouted by policymakers doi.org/10.1093/poq/nf…

Multiple exposure to political candidates' messages are more powerful than a single exposure across three different types of online platforms, finds Dr. Robert Epstein et al. doi.org/10.1371/journa…

Joe Rogan's audience "transcends" politics, instead being built on shared cultural orientations to dimensions such as institutional distrust, conspiracism, and attitudes toward gender and race, finds Dominik Stecuła Dr. Danna Young🇺🇸✌🏻 Mike Wagner Dhavan Shah preprints.apsanet.org/engage/apsa/ar…

Removing exposure to political ads on Facebook and Instagram in the 2024 US Election had no detectable effects on political knowledge, polarization, participantion, candidate favorability ,or turnout, finds Hunt Allcott Talia Stroud Joshua Tucker et al. nber.org/system/files/w…

Over the past decade, cable news stations are increasingly diverging in their topical coverage and linguistic patterns, unlike broadcast news, demonstrating the polarization in the supply of media content, finds Hosseinmardi Wolken David Rothschild 🌻 Duncan Watts nature.com/articles/s4159…

"Deep canvassing" conversations with AI can durably reduce prejudice and change policy opinions, finds Tom Costello Gordon Pennycook Robb Willer David G. Rand @dgrand.bsky.social osf.io/preprints/osf/…

Republicans penalize gay candidates while Democrats prefer them as long as the candidate looks and sounds "straight," showing the differential political biases between homosexuality and gender nonconformity, finds Martin Naunov in Journal of Politics @[email protected] doi.org/10.1086/736697

Fascinating look at internet censorship from a network connection by Saha, Afsah Anwar & Mueen doi.org/10.1145/371786…

During the 2024 Election, Democratic-aligned Discord servers focused on equity issues while Republican Discord servers centered on economic policies, with toxic language (e.g., sexism) increasing after Harris's nomination, finds Buzelin et al. doi.org/10.1145/371786…

LLMs highly agree on assessing accuracy and reliability of information sources, but their ratings only moderately correlate with human expert evaluations, with credibility ratings having a liberal bias under default configurations, finds Kevin Yang & Menczer doi.org/10.1145/371786…

Excited to be on this panel discussing surveillance capitalism today!

LLMs can effectively depolarize social media content while maintaining textual coherence, finds Santos et al., using a between-subjects experiment doi.org/10.1145/371786…

Personal narratives are particularly used by "politically disinclined individuals" on Reddit, with personal narratives from these individuals being better received in political communities than their other posts, finds Tejasvi Chebrolu Ponnurangam Kumaraguru “PK” Ashwin Rajadesingan doi.org/10.1145/371786…