How Do We Decide What „Good” AI Is?

Evaluation plays a central role in the development of Large Language Models (LLMs) shaping nearly every stage of the lifecycle. Most evaluation practices rely on fixed metrics and scores as well as target examples of what is considered the desired and correct model behavior.

From a queer-feminist perspective, this raises critical questions: Who determines what is evaluated, on what basis, and by which methods? What norms underpin these evaluations, and what is deemed “problematic” or acceptable? And how (if at all possible) could a queer-feminist evaluation of AI look like? These questions extend beyond representation, touching on broader issues of power: Who holds the authority to influence how modern technology is designed and who is included in its development?

Together with guests, the Heart of Code and the Weizenbaum Institute will unpack these tensions, examining how current evaluation methods work, where they fall short, and what efforts are being made to challenge and counter their harmful tendencies.

Please register here for the free event

For this event, the Safer Space Policy of Heart of Code applies

Speaker

• Laura Pföhler, research associate, research group “Digitalization and Opening up Science” (WI)

• Ines Montani, developer specializing in tools for AI and NLP technology, and co-founder and CEO of Explosion and a core developer of spaCy

• Goda Klumbytė investigates possibilities and applications of community and participatory audits of AI systems, University of Kassel

Moderation

Tamara Atanasoska, research associate, research group “Digitalization and Opening up Science” (WI)