Synthetic Users, Real Differences: an Evaluation Framework for User Simulation in Multi-Turn Conversations

TL;DR

This paper introduces realsim, a comprehensive evaluation framework for assessing the realism of user simulations in multi-turn chatbot dialogues, highlighting the differences from real user interactions.

Contribution

It presents a new distributional evaluation framework and a curated dataset to rigorously compare real and simulated dialogues across multiple dimensions.

Findings

Simulated users struggle to replicate communication frictions present in real dialogues.

Evaluation reveals variability in simulation quality across different chatbot domains.

Simulations may lead to overly optimistic assessments of chatbot performance.

Abstract

There is growing interest in exploring user simulation as an alternative to gathering and scoring real user-chatbot interactions for AI chatbot evaluation. For this purpose, it is important to ensure the realism of the simulation, i.e., the extent to which simulated dialogues reflect real dialogues users have with chatbots. Most existing methods evaluating simulation realism produce coarse quality signal and remain solely at the level of individual dialogues. To support more rigorous evaluation in this area, we propose realsim, an evaluation framework that enables practitioners to take a distributional view of real vs. simulated dialogues along 8 dimensions, covering attributes related to the communicative functions of the interaction, user states, and the surface form of user messages. We then instantiate the framework with a curated dataset of 1K multi-turn task-focused real user-chatbot dialogues that cover 16 domains of chatbot applications. Overall, we find that simulated users tend to struggle at capturing communication frictions that real users introduce to interactions, which could make evaluations based on such simulations overly optimistic. We also observe variability in performance across different domains, which may indicate a need for domain-specific user simulators.