Using Role-Playing Scenarios to Identify Bias in LLMs
• Podcast
Publisher
Software Engineering Institute
DOI (Digital Object Identifier)
10.58012/y68p-cr82Topic or Tag
Listen
Watch
Abstract
Harmful biases in large language models (LLMs) make AI less trustworthy and secure. Auditing for biases can help identify potential solutions and develop better guardrails to make AI safer. In this podcast, Katie Robinson and Violet Turri, researchers in the SEI’s AI Division, discuss their recent work using role-playing game scenarios to identify biases in LLMs.
About the Speaker
Katherine-Marie Robinson
Katherine-Marie Robinson is an assistant design researcher in the SEI’s AI Division. Since joining the SEI in September 2022, Robinson has worked on a wide variety of projects where she aims to bring a responsible AI (RAI) lens to the work at hand including researching and developing tools, curriculums, and …
Read more
Violet Turri
Violet Turri is an assistant software developer in the SEI AI Division where she works on multiple machine-learning engineering projects with an emphasis on explainability, test and evaluation strategies, and computer vision. Turri holds a bachelor’s degree in computer science from Cornell University and has a research background in human-computer …
Read more