Detailed Program
Introduction and Opening session (14:00–14:15)
Section 1: Bias from a Social Psychology Perspective (14:15–15:00)
Speaker: Islam Borinca
This session introduces bias through a social psychological lens, beginning with an interactive exercise before moving to key theoretical concepts and measurement approaches. Participants will take part in a series of implicit association tasks, followed by a discussion of implicit and explicit bias, in-group and out-group dynamics, and subtle versus blatant forms of bias. The session will conclude by connecting psychological approaches to the study of bias with current work in NLP.
Section 2: Measuring Underrepresentation in Archives and Models (15:00–16:00)
Speaker: Marco Antonio Stranisci & Lia Draetta
We overview existing practices to detect and mitigate the underrepresentation of vulnerable communities at different stages of the creation of NLP technologies: data selection and filtering, annotation, and model implementation. The overview adopts methods and frameworks from Semantic Web, gender studies, applied mathematics, social sciences, and NLP.Coffee Break (16:00–16:30)
Section 3: Discuss the issues deriving from the human labor behind fair and ethical NLP systems (16:30–17:30)
Speaker: Ephantus Kanyugi & Joan Kinyua
While discussions of bias in NLP often center on datasets, models, and algorithms, far less attention is paid to the human labor behind them: the global network of data labelers, content moderators, and annotators whose invisible work shapes every AI system. This session connects the academic conversation on bias in NLP to the worker realities that produce these datasets, revealing how structural inequalities, pay disparities, and cultural invisibility introduce another layer of bias—one rooted in labor practices rather than data points.