Using users for online evaluation and learning

Summary

This presentation explores the application of online controlled experiments for both information retrieval and recommender systems, addressing how implicit user feedback can bridge the gap between user information needs and system queries. The talk demonstrates how real user interactions can inform system learning and evaluation, moving beyond traditional offline metrics to capture authentic user satisfaction and engagement. This approach is particularly valuable for understanding the translation of implicit information needs into explicit queries and optimizing both search and recommendation algorithms based on genuine user behavior.