Position is Power - System Prompts as a Mechanism of Bias in Language Models

Explore how system prompts can introduce and amplify bias in language models through this 11-minute conference talk from the Association for Computing Machinery. Examine research findings from the University of Duisburg-Essen and Ruhr-University Bochum that demonstrate how the positioning and structure of system prompts serve as powerful mechanisms for bias propagation in AI systems. Learn about the evaluation practices used to assess bias in generative AI models and understand the implications of prompt positioning on model outputs. Discover methodologies for identifying bias patterns in language models and gain insights into the relationship between system prompt design and fairness in AI applications. Understand the critical role that prompt engineering plays in either mitigating or exacerbating bias issues in modern language models, with practical implications for developers and researchers working with generative AI systems.