Web agents ask smart questions to learn user preferences for future tasks
Learning What to Ask for Tomorrow: Active Preference Acquisition for Long-Term Personalized Web Agents
Computation and Language
Summary
Web agents help people with repeated online tasks like booking travel or shopping, but they only know what users tell them. The authors study a way for these agents to actively ask questions that might be worth it because the answers help not just now but also in future tasks. They introduce a method called Active Preference Acquisition, which decides which questions to ask by estimating the long-term benefit minus the cost of asking. Their experiments show this method asks fewer questions but learns more useful preferences over many tasks than methods focusing only on the current task.
What this means in practice
- •For web service developers: Design web assistants that ask questions strategically to improve user experience across repeated tasks over time while minimizing user effort.
- •For e-commerce platform teams: Enhance shopping agents by acquiring key buyer preferences that deliver better recommendations not only immediately but also in future interactions.
Authors
Yiqi Wang, Zizhi Zeng, Jiaqi Zhang, Yiqun Duan, Mingkai Zheng, Taotao Cai
Abstract
Web agents increasingly serve the same user across repeated online tasks such as booking travel, shopping, and scheduling, so their long-term usefulness depends on how well they know the user's preferences. Passive memory and personalization methods cannot obtain preferences the user never reveals, and clarification methods that value a question by its benefit to the current task overlook the benefit of reusing the answer in later tasks. We study Active Preference Acquisition (APA): which missing preferences are worth the cost of asking? We define the Prospective Personalization Value (PPV) of a missing preference as its expected cumulative utility gain over future tasks minus the effort and disclosure-risk cost of asking, and select questions under a per-interaction budget. We characterize when cumulative reuse justifies acquisition despite insufficient single-task benefit, and evaluate this regime with an initial-profile diagnostic. APA computes the counterfactual gain of each candidate preference on public task prototypes with the agent's own task solver and aggregates it by the expected reuse of each task family. We evaluate it on ActiveWebPref, a controlled longitudinal benchmark with hidden but queryable user profiles and exact task utilities. At a horizon of 20 tasks, 34.4% of budget-feasible valuable missing preferences are amortization-dependent; APA acquires 80.3% of them, whereas current-task value of information acquires 5.1%. Under per-user asking-cost caps matched to APA's realized expenditure, APA achieves higher net utility than four heuristic selection baselines. In our cost setting, asking about every unknown preference yields lower net utility than not asking, while APA recovers 84.2% of its utility gain with 27.6% of its questions. The value of acquiring a user preference should account for its reuse across future tasks, not only its benefit to the task at hand.