Lectures and seminars SFOepi Seminar Series in Biostatistics: Tibor Varga, University of Copenhagen

Abstract

Algorithmic fairness is usually discussed as a technical property, e.g., a metric to optimize or a bias to correct. In this talk, I will argue that algorithmic fairness is better understood as a systems problem. In healthcare algorithms, for example, clinical risk scores or AI models, bias can emerge through the interaction of at least four levels: the social structures that generate inequities, the data that captures these inequities, the algorithms trained on that data, and the healthcare institutions that deploy and act on algorithmic outputs. A complexity science lens, attentive to feedback loops, emergence, and dependency across these levels, offers epidemiologists a more accurate vocabulary for work than fairness metrics alone. This talk introduces what fairness and bias mean across the social, data, algorithmic, and institutional levels, and helps the audience begin engaging with these challenges as systemic ones.

60 mins including questions

Welcome also to a workshop this afternoon with Tibor Varga. For details and registration, see Workshop in the SFOepi Seminar Series in Biostatistics: Tibor Varga, University of Copenhagen. 

Speaker

Tibor Varga from Department of Public Health, Copenhagen Health Complexity Center, University of Copenhagen

Registration: No registration is needed for the seminar

If you have any questions, please contact Marie Jansson at marie.jansson@ki.se