MID seminar: guest Abdallah “Abdo” El Ali - Human-Centric AI Disclosures: Limits, Biases, and Design Beyond Labels

Abstract:
How should we disclose our use of AI? As AI-generated and AI-mediated contentbecomes increasingly indistinguishable from human-produced information, itbecomes critical to understand how such content is disclosed and interpreted,especially when it may be inaccurate or misleading. While regulatory effortssuch as the EU AI Act (Article 50) mandate disclosure, current approaches, suchas static “AI-generated” labels, remain limited. In this talk, I will highlightthe importance of considering ‘what’, ‘how’, and ‘when’ to disclose AIinvolvement. I will present a series of studies across domains including newsmedia and health, showing how AI disclosures can systematically bias people’sperceived trust, behavior, and physiological responses. I will further show howthese biases can extend to AI systems themselves when large language models areused as evaluators (LLM-as-a-judge). Based on these findings, I will highlightopportunities for designing human-centric AI disclosures that go beyond staticlabels by visualizing the nuances of human-AI collaboration. I will conclude bydiscussing how these approaches remain limited as AI becomes increasinglyassimilated into human expression, raising new questions for AI disclosures.

Bio:
Abdallah “Abdo” El Ali is a research scientist at Centrum Wiskunde &Informatica (CWI) in Amsterdam, and part-time Assistant Professor at UtrechtUniversity. He leads the research areas on Affective Interactive Systems andTrustworthy Human-AI Interaction, with a focus on human augmentation acrossphysical and digital environments and AI transparency. He is affiliated withthe AI, Media, and Democracy Lab, and serves as Chair of CHI Nederland (CHINL), the ACM SIGCHI Netherlands Chapter. Website: https://abdoelali.com

Photo: https://abdoelali.com/images/home_imgs/abdo_el_ali_cwi_grayscale.jpg