Abstract Summary
In conversational search, user simulation is an important method for evaluation and it offers three key benefits: (i) considering conversational interactions, (ii) avoiding user privacy issues, and (iii) reducing evaluation costs compared to online evaluation. However, most prior work has focused on simulating generic users, while few studies have explored to simulate specific users with simple user traits, which is an essential step toward achieving more personalized evaluations. Moreover, simulating specific users is beneficial for developing personalized conversational search systems by enabling training on simulated user data. With the advancements of large language models (LLMs), there is a trend to leverage the abilities of LLMs for user simulation. This work aims to explore how to simulate specific users with complex traits using LLMs, going beyond the simple user traits considered in previous studies.
