Reprogramming Bias: Exploring Gender Representation and Inclusive Design Through AI-Generated Synthetic Personas

Speaker: Helena A. Haxvig, University of Trento

Abstract: The growing use of generative AI, particularly large language models (LLMs), has introduced new complexities around bias, representation, and accountability. In this talk I examine the implications of using LLM-generated synthetic personas in research and design. While these personas offer novel possibilities for participatory and co-creative research methods, they also risk perpetuating or amplifying existing societal biases embedded in training data and model architectures. Drawing on four empirical studies and interdisciplinary critiques, I highlight how gender and identity are often flattened or misrepresented by LLM outputs and propose frameworks for critically assessing bias through collaborative persona co-creation and exploration. Ultimately, I explore the ethical and methodological challenges of using AI-generated personas in research and design, arguing for more reflexive, inclusive, and transparent approaches.

Biography: Helena A. Haxvig (She/They) is currently a third-year PhD student at the University of Trento, Department of Information Engineering and Computer Science. With a background in Techno-Anthropology, they specialize in human-technology relations, focusing on Participatory Design and sustainable AI. Their research is grounded in feminist and queer Human-Computer Interaction (HCI), critically examining the ethical and societal implications of biased (AI) systems.

Host: Prof. Monica Landoni