In this talk, Yan will discuss using the theory of Contextual Integrity

(CI) to develop and evaluate socio-technical systems that incorporate a socially meaningful conception of privacy. I will discuss CI-based methods for identifying contextual privacy norms and examining how these norms may differ from policies and regulations. I will then present our recent work on using CI to assess the encoded norms in LLMs and the impact of model properties (e.g., hyperparameters, capacity) and optimization strategies.

If you are interested in learning more about contextual integrity, please consider attending the CI symposium, a week earlier, on May 19-20 in Brussels: https://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fprivaci.info%2Fsymposium%2F2025%2Fprogram.html&data=05%7C02%7CLaura.hoekjan%40ru.nl%7Ca64afb8d70d041f0cc2808dd8dfed1e5%7C084578d9400d4a5aa7c7e76ca47af400%7C1%7C0%7C638822847768064660%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7C%7C&sdata=SXWxqz8o6OHxr1Y9m9OQcgVpbCcFdqslzRRNH6T1CB4%3D&reserved=0

About the speaker:
Yan Shvartzshnaider is an Assistant Professor in the Department of Electrical Engineering and Computer Science, Lassonde School of Engineering at York University, where he leads the Privacy Rhythm research lab. His research interests include sociotechnical systems that incorporate a socially meaningful conception of privacy that meets peoples’ expectations and is ethically defensible.