Hesitator: A User Simulator for Conversational Recommender Systems
Researchers have introduced Hesitator, a user simulation framework for evaluating conversational recommender systems. Unlike existing simulators that unrealistically assume perfect information processing and high acceptance rates, Hesitator explicitly models human decision-making under choice overload. Its modular architecture separates utility-based item selection from overload-aware commitment decisions, enabling more realistic simulations of hesitation and deferral. Experiments across multiple frameworks, domains, sales modes, and LLM backbones demonstrate its effectiveness.
Key facts
- Hesitator is a theory-grounded user simulation framework.
- It addresses the limitation of LLM-based simulators exhibiting unrealistically strong information-processing capabilities.
- The framework introduces a modular Decision Module that separates utility-based item selection from overload-aware commitment decisions.
- Experiments were conducted across multiple user simulation frameworks, domains, sales modes, and LLM backbones.
- The work is published on arXiv with ID 2605.05250.
Entities
Institutions
- arXiv