Critical AI Lab

I am thrilled to be recognized by @TIME as one of the 100 most influential people in the field of artificial intelligence for my work with Data Workers' Inquiry. #TIME100AI time.com/time100ai I want to take this opportunity to share a few reflections on my work👇🧵

Dicen que “la IA no destruye empleos; sólo los transforma”. Lo que no dicen es que esa “mágica transformación” de la que hablan es lisa y llanamente más precarización. Detrás de los algoritmos hay legiones de trabajadores precarizados e invisibilizados: los trabajadores de datos.

Next Monday, we are thrilled to welcome @alexhanna to the @JWI_Berlin for a conversation about her new book, THE AI CON: How to Fight Big Tech's Hype and Create the Future We Want (thecon.ai) 🔗Register here: tickets.weizenbaum-institut.de/sgvbf/

One year ago, on #LaborDay, we launched Data Workers' Inquiry. The process was messy and uncertain, but driven by the need to do research differently. I wrote a reflection on what this first year taught us data-workers.org/about/how-to-d… Thanks for reading and for supporting this work.

1/4 Exciting News! David Hartmann et al.'s paper, "Lost in Moderation: How Commercial Content Moderation APIs Over- and Under-Moderate Group-Targeted Hate Speech and Linguistic Variations," has been accepted to CHI Conference 2025.

4/4 This work has been a joined Project between the @JWI_Berlin , CAIS and the @thehertieschool with the co-authors Amin Oueslati, Dimitri Staufer, Lena Pohlmann, @simonsaysnothin , Hendrik Heuer.

3/4 They conclude that using commercial APIs for content moderation risks silencing legitimate speech and failing to protect online platforms from harmful speech! Read the preprint here: pdf.arxiv.org/pdf/2503.01623

2/4 This Paper evaluates five content moderation APIs. They find that all APIs have problems correctly identifying implicit hate speech - like sarcasm. At the same time, the APIs tend to misclassify content related to Black, LGBTQIA+, Jewish, and Muslim people as hate speech.

Amid the hullabaloo caused by DeepSeek, its esential that I remind you that none of these systems are truly "artificial" or "intelligent" Without human intervention, they cant distinguish a cup from a puppy. Set your calendars and hear more at the DLA launch!

Let's take care of the Internet! David Hartmann was recently featured in the "Intenet Governance Explained"-Podcast. Dive into the episode to learn more about the risks of using AI in content moderation and the working conditions of content moderators. open.spotify.com/episode/3u2iq9…

Data Labellers and annotators are literally the lifeblood of any AI systems, Machines are stupid, they don't know a tree from a cup, yet the people who enable this recognition to happen are often neglected and invisibilised. We are elated to support this initiative

The Data Labelers Association advocates for the rights, fair pay, job security, and mental and occupational health of data labelers and annotators! Join the Launch on February 13, 18h CET / 12PM ET twitch.tv/dair_institute

📢We're proud to support the upcoming launch of the Data Labelers Association of Kenya with an online event! Join us to meet the workers and support their struggle. 📅Feb 13 ⏰8pm Kenya/ 6pm EU/ 12pm ET ✅Watch and submit your questions to the speakers: twitch.tv/dair_institute

Just published with a Best Paper award! @ella_es_adriana, @TianlingY & @MilagrosMiceli examine the discursive strategies in the papers that combine social media data and computational methods. doi.org/10.1145/3701216

I am thrilled to announce our (w/@TianlingY @MilagrosMiceli) research article "What Knowledge Do We Produce from Social Media Data and How?" won Best Paper Award at @ACM_GroupWork (dl.acm.org/doi/10.1145/37…)

New article from @fatmaElsafoury: Dive into the different facets of discrimination and bias in AI algorithms, focusing on the growing deep-fake industry. Stay informed—read and share widely! bpb.de/lernen/bewegtb… (english: efatmae.github.io/posts/2024/11/…)

New publication by @dawietman et al.! Their research exposes a critical regulatory gap in the AI Act & DSA, calling for a diverse AI audit ecosystem that empowers civil society and affected groups to ensure oversight and accountability doi.org/10.1007/s43681…

Paper alert! @TianlingY & @MilagrosMiceli explore guilds on Chinese data work platform as a form for both worker empowerment and control. Dive into their ethnographic work about a crowdsourcing platform in China doi.org/10.1145/3686904