Workshop: The Politics of Models – Bias, Representation & Transparency in Generative Models

Generative models can capture an abstract representation of various modalities such as sound, images, video, text, and data. These models are now globally utilized for knowledge and content organization and synthesis. The generative AI models are integrated rapidly into research, industry, culture, and the public sectors; where the accelerated growth hinders the considerations into societal implications of these rapid advancements. These models are new forms of knowledge representation, in which the technical choices such as applied ontology or categorizations as well as the data sources and gathering processes, influence, imply, and afford new forms of power and control on representation, history, culture, and knowledge associations. Addressing these issues requires both technical advancement and research as well as societal grounding, awareness in positionality, and critical engagements. This workshop creates a shared space for WASP and WASP-HS researchers with the goal to build interdisciplinary research directions to influence a better and healthier societal integration of generative AI.

Topics

Representation & Bias
Analysis of representation and bias in generative models across modalities – sound, images, video, text, and data.

Power & Control
Understanding the new forms of power and control enabled through generative AI over representation, history, and culture.

Transparency & Accountability
Building frameworks for transparent and accountable deployment of generative models in research and society.

Societal Alignment
Aligning technical solutions with societal expectations, needs, and values for a healthier integration of AI.

Schedule

Kivanç Tatar, Chalmers University of Technology
Kathlén Kohn, KTH Royal Institute of Technology
Md Fahim Sikder, Linköping University
Francis Lee, Södertörn University
Bodil Formark, Södertörn University