iiinova — where your digital self keeps evolving

Thomas Davidson @thomasdavidson.bsky.social · 28/07/2026

I just landed in Burlington for @ic2s2.bsky.social. Looking forward to seeing everyone at the conference. I'm giving a talk about large reasoning models tomorrow and Duhui Lee is presenting some of our new work on conspiracy theories and visual media on Thursday #IC2S2

A view from a plane over Burlington VT on a cloudy day

0101

Thomas Davidson @thomasdavidson.bsky.social · 27/05/2026

Our paper on the pandemic and populism is online in Comparative Political Studies. We find that local COVID-19 transmission increased support for right-wing populists in Europe, and triangulate our findings across data from social media, elections, and surveys.

1144

Thomas Davidson @thomasdavidson.bsky.social · 10/04/2026

Mine is still alive and well (homebrew install)

020

Thomas Davidson @thomasdavidson.bsky.social · 27/03/2026

Question for AI Bsky, has anyone managed to get useful reasoning traces from open-weights models? Here is some typical output I got from Qwen 3.5 on a vision task. It starts out reasonable but descends into gibberish (neuralese?) @lauraknelson.bsky.social @tedunderwood.com @cbarrie.bsky.social

A screenshot of a wall of text. The text begins by describing a task presented by a user and then turns into gibberish, repeating many different words and names without any structure. A second image is similar but includes lots of symbols.

350

Thomas Davidson @thomasdavidson.bsky.social · 16/03/2026

Excited to be in Toronto to share my latest research in the UTM sociology speaker series and conduct a graduate student masterclass this afternoon! www.sociology.utoronto.ca/events/utm-s...

021

Thomas Davidson @thomasdavidson.bsky.social · 03/03/2026

We ran an experiment where people read GPT-4o summaries or Wikipedia. Default summaries with no ideological slant and texts generated using a liberal persona both shifted readers toward liberal opinions relative to Wikipedia. Conservative-framed texts only shifted among conservatives. 2/4

110

Thomas Davidson @thomasdavidson.bsky.social · 03/03/2026

Our new paper is out today in @pnasnexus.org with colleagues at Yale (@matthewshu.com, Danny Karell, @keitarookura.bsky.social) We wanted to understand how using AI-generated summaries to learn about history influenced attitudes compared to existing resources like Wikipedia. 1/4

1209

Thomas Davidson @thomasdavidson.bsky.social · 03/02/2026

First 50 downloads are free if you use this link: www.tandfonline.com/eprint/AYC6H...

001

Thomas Davidson @thomasdavidson.bsky.social · 06/01/2026

On the topic of AI and social science research, the Research Briefing on my Nature Human Behaviour paper is now online. It's an accessible summary of the research, implications, and some behind-the-scenes commentary. Thanks @gligoric.bsky.social for providing an expert opinion!

151

Thomas Davidson @thomasdavidson.bsky.social · 15/12/2025

Additionally, some models are overtly biased and are particularly sensitive to visual identity cues (AI-generated profile pictures). This demonstrates how different data modalities lead to varying levels of algorithmic bias.

100

Thomas Davidson @thomasdavidson.bsky.social · 15/12/2025

When considering the identity of the author, some MLLMs make context-sensitive judgments comparable to human subjects. e.g., less likely to flag Black users for using reclaimed slurs, a common false positive. But the results also reveal less normative decisions regarding so-called "reverse racism".

100

Thomas Davidson @thomasdavidson.bsky.social · 15/12/2025

I find that MLLMs follow a consistent hierarchy of offensive language to humans and show similarities across other attributes. There is heterogeneity across models, particularly the smallest open-weights versions.

100

Thomas Davidson @thomasdavidson.bsky.social · 15/12/2025

New paper in Nature Human Behaviour. I use a conjoint experiment to test multimodal large language models (MLLMs) for context-sensitive content moderation and compare with human subjects. Methodologically, this demonstrates how social science techniques can enhance AI auditing. 💻🤖💬

2182

Thomas Davidson @thomasdavidson.bsky.social · 09/09/2025

Analysis of the reasoning traces for Gemini 2.5 shows that the model identifies second-order factors when faced with these decisions, helping to address common false positives like flagging reclaimed slurs as hate speech (Warning: offensive language in example)

110

Thomas Davidson @thomasdavidson.bsky.social · 09/09/2025

On a content moderation task, humans take longer and LRMs use more tokens when offensiveness is identical or fixed. This suggests that LRM behavior is consistent with dual process theories of cognition, as the models expend more reasoning effort when simple heuristics are insufficient

120

Thomas Davidson @thomasdavidson.bsky.social · 09/09/2025

The results are consistent across three frontier LRMs: o3, Gemini 2.5 Pro, and Grok 4

110

Thomas Davidson @thomasdavidson.bsky.social · 09/09/2025

New pre-print on large reasoning models 🤖🧠 To what extent does LRM behavior resemble human reasoning processes? I find that LRM reasoning effort predicts human decision time on a pairwise comparison task, and both humans and LRMs require more time/effort on challenging tasks

161

Thomas Davidson @thomasdavidson.bsky.social · 01/08/2025

If you want to learn more about these articles, check out our editors’ introduction, where we provide an overview and discuss some central themes. One point we emphasize is how the model ecosystem has matured & open-weight models are viable for many problems journals.sagepub.com/doi/10.1177/...

110

Thomas Davidson @thomasdavidson.bsky.social · 01/08/2025

Stuhler, Ton, and Ollion show how LLMs enable more complex information extraction tasks that can be applied to text corpora, with an application to the study of obituaries journals.sagepub.com/doi/abs/10.1...

130

Thomas Davidson @thomasdavidson.bsky.social · 01/08/2025

A study by Than et al. revisits an earlier SMR paper, exploring how LLMs can be used for qualitative coding tasks, providing detailed guidance on how to approach the problem journals.sagepub.com/doi/full/10....

111

Thomas Davidson @thomasdavidson.bsky.social · 01/08/2025

Law and Roberto showcase how vision language models like GPT-4o can extract information from satellite images, applying these techniques to study segregation in the built environment journals.sagepub.com/doi/full/10....

110

Thomas Davidson @thomasdavidson.bsky.social · 01/08/2025

Waight and colleagues develop a pipeline for quantifying narrative similarity, combining the latest frontier LLMs with earlier techniques to study Russian influence campaigns journals.sagepub.com/doi/abs/10.1...

100

Thomas Davidson @thomasdavidson.bsky.social · 01/08/2025

Maranca and colleagues show the importance of statistical corrections when using predictions from multimodal LLMs and other computer vison models for downstream tasks, providing several applications journals.sagepub.com/doi/abs/10.1...

100

Thomas Davidson @thomasdavidson.bsky.social · 01/08/2025

Kozlowski and Evans synthesize work using generative AI for simulation, giving us an in-depth view into how AI models work and how scholars are addressing six challenging issues journals.sagepub.com/doi/abs/10.1...

140

Thomas Davidson @thomasdavidson.bsky.social · 01/08/2025

Broska, Howes, and van Loon propose the mixed-subjects approach, providing a methodology for combining estimates from human subjects and silicon samples journals.sagepub.com/doi/abs/10.1...

110

Thomas Davidson @thomasdavidson.bsky.social · 01/08/2025

Lyman and colleagues explore the trade-offs between using instruction-tuned models and base versions of LLMs for downstream tasks, emphasizing the need to better understand model training journals.sagepub.com/doi/abs/10.1...

100

Thomas Davidson @thomasdavidson.bsky.social · 01/08/2025

Boelaert and colleagues pose a challenge to work that uses LLMs as substitutes for humans in surveys, finding that AI models exhibit idiosyncratic biases they term “machine bias” journals.sagepub.com/doi/abs/10.1...

110

Thomas Davidson @thomasdavidson.bsky.social · 01/08/2025

Zhang, Xu, and Alvero show how online survey participants are already using AI for open-ended responses, demonstratng how many researchers will have to grapple with the impacts of AI-generated data journals.sagepub.com/doi/abs/10.1...

183

Thomas Davidson @thomasdavidson.bsky.social · 01/08/2025

I’m delighted to share that the August 2025 special issue of Sociological Methods & Research on Generative AI is out now. Along with my co-editor, Daniel Karell, we put together this issue to build on the conference we organized last year. Here's a thread on each of the ten papers:

A screenshot of the Sociological Methods & Research website showing the special issue title

37531

Thomas Davidson @thomasdavidson.bsky.social · 02/05/2025

It was an honor to visit the University of Kansas to give the 2025 Blackmar Lecture. I had a great time learning about the department and the legacy of Frank Blackmar, who taught the first sociology class in the US, which has continued for 135 years www.asanet.org/frank-w-blac...

050

Thomas Davidson @thomasdavidson.bsky.social · 24/04/2025

Our article on using LLMs for text classification is out now in SMR! 🤖🧑‍💻💬 We compare different learning regimes, from zero-shot to instruction-tuning, and share recommendations for sociologists and other social scientists interested in using these models. doi.org/10.1177/0049...

26319

Thomas Davidson @thomasdavidson.bsky.social · 13/03/2025

My review of Natalie-Anne Hall's recent book on Brexit supporters' post-referendum engagement on Facebook is out now in Contemporary Sociology: doi.org/10.1177/0094... Hall's analysis explains why extremism and conspiracy theories resonate with online audiences (and some platform owners).

Book cover for Brexit, Facebook, and Transnational Right-Wing Populism by Natalie-Anne Hall. It shows a map of the world with information flowing across the image, evoking digital communication.

Text from the review that reads: "Overall, Hall’s study illuminates the complex interactions among transnational events and ideologies, social media, and people’s everyday experiences of social change. While some of the Brexiteers’ concerns were distinctly parochial, like worries about immigration and social issues in their communities, social media provided them with a window into diffuse global developments and transnational ideologies. They used Facebook to consume and discuss material such as crimes allegedly committed by migrants in Europe, attacks on white farmers in South Africa, and terrorist attacks in distant places, which they connected back to the local and the personal. Right-wing actors, from the grassroots level to political parties and media organizations, are particularly astute at weaving together disparate events into compelling narratives, ranging from conspiracy theories like Cultural Marxism and the Great Replacement to anxieties about political correctness and Islam. Hall’s book is not only about Brexit, but provides important insights into the social, political, and technological bases for the recent growth and evolution of right-wing populism."

050

Thomas Davidson @thomasdavidson.bsky.social · 23/09/2024

My article on Generative AI and Sociology is out now in Socius! I explore the applications of GenAI across computational, qualitative, and experimental research, and discuss important issues including bias, reliability, and interpretability. journals.sagepub.com/doi/full/10....

0215

Thomas Davidson @thomasdavidson.bsky.social · 21/03/2024

Join our CBSM small group on computational methods for a discussion of new data, methodologies, and the future of social movements research 🪧 💻📱#️⃣🤖 With @lauraknelson.bsky.social, @haphazardsoc.bsky.social, Josh Zhang, Eunkyung Song, and Danny Karell. PM me for Zoom details.

The image is a flyer for an event titled "The Data Pulse of Social Change: Social Movements Research in the Computational Era". The event will take place online on Tuesday, April 9 at 4-5pm ET.

072

Thomas Davidson @thomasdavidson.bsky.social · 16/01/2024

Seeing some discussion about AI course policies. It's the first time teaching my CSS classes in a couple of years, so I came up with this. Interested to hear any feedback / how others have addressed this in technical courses.

120

Thomas Davidson @thomasdavidson.bsky.social · 06/12/2023

Submit an abstract to the generative AI and sociology workshop, due next Friday 12/15 We also welcome submissions from other social scientists and fellow travelers For more info visit: tinyurl.com/soc-gen-ai

063

Thomas Davidson @thomasdavidson.bsky.social · 15/11/2023

Are you a social scientist using Generative AI in your research? Daniel Karell and I are organizing the Generative AI and Sociology Workshop at Yale on April 5-6, 2024. Abstracts are due December 15. For more info, visit: tinyurl.com/soc-gen-ai

12514

Thomas Davidson @thomasdavidson.bsky.social · 02/11/2023

New working paper on the use of generative AI as a sociological method. I discuss applications to computational, experimental, and qualitative research, and use GPT-4 & DALLE-3 to provide motivating examples of image-to-text and text-to-image analyses. Feedback welcome! osf.io/preprints/so...

23913

Thomas Davidson @thomasdavidson.bsky.social · 16/10/2023

New working paper with Ranjit Lall and Felix Hagemeister: osf.io/7xqkz/ We leverage the exogeneity of early super spreader events to analyze how the onset of the pandemic boosted support for right-wing populists in Europe using data from Twitter, French elections, and British and Dutch surveys.

151