I study how platforms, creators and AI systems shape what reaches people online, and how to measure that from the outside. Computer scientist by training, postdoc at Utrecht University (ERC project HUMANads), somewhere between computational social science, platform governance and law.
Most of what lives here is research software: pipelines, datasets, benchmarks and a few public tools, so you can poke at the work rather than just read about it.
- dfa-dashboard: LLM pipeline and dashboard for the EU Digital Fairness Act consultation. 4,322 submissions, every claim linked to its quote. Try it · paper
- enelvo: normaliser for messy Portuguese text. On PyPI since 2016, still maintained.
- Human or LLM?: can you tell a human post from an AI one? A game built on instasynth (ICWSM 2024).
- The Sponsored Scroll: what should be visible in a social feed? Interactive project with Utrecht colleagues.
- ChildSafeAds: shared task at NLLP @ EMNLP 2026 on ads in child-facing YouTube videos. Lead organiser.
- Datasets: ALYT (abusive language on YouTube), toxmon (toxicity and monetisation), instagram-disclosure-trends (influencer disclosure, four countries).
thalesbertaglia.com · Scholar · LinkedIn · Bluesky · ORCID · contact@thalesbertaglia.com