SORS/WomenInBSC: Countering AI contamination and hallucination скачать в хорошем качестве

SORS/WomenInBSC: Countering AI contamination and hallucination

Speaker: Terhi Päivikki Esko, STS Helsinki Abstract: The presentation addresses prompting as situated human practice for mitigating AI hallucinations – the production of false or misleading outputs by generative models. It is connected to what is described as AI contamination: the result of a growing body of AI generated questionable, harmful, and misinformed content. This presentation launches the project Prompting as Practice for Countering AI Hallucinations (PATCH), which builds on prior work in the context of legal tech and the insights provided by BSC project AHEAD. By combining preliminary ethnographic inquiry and interviews with professionals experimenting with AI tools, the research investigates how users can develop local techniques to “ground” AI systems in various contexts. By studying prompting across diverse technical and institutional settings, I aim to examine how reliability and trustworthiness are co-produced through everyday human–machine interaction. Theoretically, I draw from science and technology studies (STS) focusing on scientific errors and uncertainty. The findings inform interdisciplinary discussions on responsible AI, emphasizing the creative, interpretive, and corrective work that sustains trustworthy and context-specific computation in the fields of health/life science and law. In doing so, this work bridges social studies of technology and science and technology studies, offering both conceptual insight and practical strategies for more accountable generative systems. Further information here: https://www.bsc.es/4Jq