Overview

Data sources do not always interpret a user’s information need neutrally. Product search, recommendation, and ranked retrieval systems can have financial or organizational incentives to surface biased results.

This work develops a formal framework for querying when the data source and user have conflicting interests. It studies when relevant information can still be extracted from biased sources, how biased information can be detected efficiently, and how input queries can be reformulated to improve the usefulness of returned results.

Research Themes

• Querying over biased or strategic data sources.
• Detecting conflict-driven bias in returned results.
• Reformulating queries to improve relevant information recovery.
• Building data systems that account for user and provider incentives.