This work was a great collaboration with @fgilardi.bsky.social, Zeynab Samei, and @mmosleh.bsky.social

Read the full preprint here:
🔗 arxiv.org/abs/2507.12372

Another risk: web-browsing LLMs are vulnerable to prompt injection attacks. As shown in arxiv.org/abs/2406.18382, hidden prompts in webpages can hijack model behavior—posing serious threats to otherwise beneficial applications.

Our findings show major privacy risks: web-browsing LLMs can profile social media users at scale—opening the door to misuse by malicious actors. Under GDPR, such automated profiling requires transparency, consent, and the right to opt out.

Our second dataset came from a prior survey study (N=1,384) with self-reported usernames and demographics from an international sample, mostly from US and UK.
LLMs achieved higher accuracy than with the synthetic dataset—and showed no gender or political bias.

Our first dataset included 48 synthetic X accounts, generated with GPT4o (tweets, profiles, images).
Even for the 19 accounts that were later suspended, LLMs could still infer age and gender.
We also observed gender and political biases—especially against low-activity accounts.

We tested six prompting strategies to infer age, gender, political orientation, and socioeconomic status—using only usernames or X (Twitter) profile links.
Surprisingly, in most cases, usernames outperformed full profile links.

🚨 New preprint alert:

We show that web-browsing GPT and LLaMA models can infer social media user demographics with reasonable accuracy—using only usernames.

This opens new possibilities for social media research in the post-API era but raises important privacy concerns.

✨ New LLM paper ✨

> Open-source LLMs perform well in text annotation tasks
> Fine-tuned open-source LLMs outperform zero-shot GPT-4
> Fine-tuning beats few-shot training with modest amount of annotated data

w/ @malizadeh.bsky.social @maelkubli.bsky.social et al.

link.springer.com/article/10.1...

Want to see if your research has had an impact on public policy?

Enter your info in SagePolicyProfiles to see the citations of your own publications in policy documents.

You get a profile page. Here is mine as an example: policyprofiles.sagepub.com/profile/5920...

"We find that sustained negative coverage significantly predicts changes to platforms’ user policies, highlighting the role of public pressure in shaping the governance of online platforms."

by @emmahoes.bsky.social @klueserthan.bsky.social @maelkubli.bsky.social @malizadeh.bsky.social et al.

"we present FreshPrompt, a simple few-shot prompting method that substantially boosts the performance of an LLM on FreshQA by incorporating relevant and up-to-date information retrieved from a search engine into the prompt."

via @malizadeh.bsky.social

Check out the program of the 2023 Digital Democracy Workshop, Oct 26-27, University of Zurich

Keynote by Atoosa Kasirzadeh: "Can we reconcile the governmental use of online targeting with democracy?"

Free attendance, registration required

polisky commsky cssky