This project lays the groundwork for understanding how existing tools for privacy-preserving data analysis interact with strategic and human aspects of practical privacy guarantees. When strategic individuals have privacy concerns about the use of their data, they may modify their behavior to ensure less, or perhaps more favorable, information is revealed. The project's novelties are an interdisciplinary approach, which combines tools from algorithm design, machine learning, and economics. The broader significance and importance of this work is to provide a critical step for society's ability to collect useful data and to interpret data via existing algorithms. As more personal data are collected, stored, and used in algorithmic decision making, these results are useful in the legal and policy landscape of personal data management. This work has two main technical thrusts. First, this project studies how privacy technologies can be designed and deployed to manage privacy concerns of strategic individuals. This yields insight into the design of optimal privacy technologies for strategic individuals in practical application areas. Second, this project develops data analysis techniques for settings where data are generated by privacy-aware individuals. This yields tools for the design and analysis of algorithms to efficiently learn and optimize from a strategic individual's data. This project also includes a significant educational and outreach component, including curriculum development, mentorship of students, and workshop organization.
CRII: SaTC: Data Privacy for Strategic Agents