Location: Zoom (email presenter for link)
Add to Calendar 2021-08-02T13:30:00 2021-08-02T13:30:00 America/New_York Ph.D. Proposal Presentation by Yiling Jia Toward Reliable Interactive Learning for Search and Recommendation Systems   Abstract: Zoom (email presenter for link)

Toward Reliable Interactive Learning for Search and Recommendation Systems



Nowadays information service systems are increasingly being used in everyday decision-makings, from recommending the movies to watch, products to purchase, to even providing the treatment to receive in healthcare. Such systems enable us to have every kind of data we could possibly want at our fingertips by highlighting a small number of particularly relevant or valuable instances from a vast catalog. However, as such systems become more involved in our lives in almost all the aspects, it brings up a big factor on whether the systems could provide reliable service and gain users' trust. Arguably, users' confidence and trust in the system will rapidly deteriorate if the system cannot provide expected service, such as providing inferior recommendation and ranking results, making unfair decisions, or lack of explanation for the results, which makes a reliable system urgent and important.


In this proposal, I focus on improving the reliability of interactive learning in information service systems, focusing on three aspects: stability, fairness and explainability. Firstly, for explainable recommendation systems, I propose to perform multi-task learning for both user preference modeling for personalization, and content modeling for explanation. Secondly, to provide stable online search service, I propose to estimate a pairwise ranker on the fly and explore the unknowns based on the model uncertainty, which enables efficiently exploration and thus expedites the learning process. Finally, I propose to incorporate the fairness metric into existing utility-oriented online ranking policies to provide fair ranking list to both users and the content providers online. Overall, we aim to improve the explainability, stability and fairness of the interactive learning to provide reliable information service and thus gain user trust.



  • David Evans, Committee Chair (CS/SEAS/UVA)
  • Hongning Wang, Advisor (CS/SEAS/UVA)
  • Haifeng Xu (CS/SEAS/UVA)
  • Cong Shen (ECS/SEAS/UVA)
  • Liangjie Hong (LinkedIn)