Evaluating Role-Consistency in LLMs for Counselor Training
This work addresses the need for effective training methods for future counselors in online services, but it is incremental as it builds on existing research on virtual clients.
The paper tackles the problem of training counselors by testing large language models' ability to maintain consistent roles during simulated client interactions, introducing an adversarial dataset and finding that Vicuna and other open-source models show varying performance in role-consistency.
The rise of online counseling services has highlighted the need for effective training methods for future counselors. This paper extends research on VirCo, a Virtual Client for Online Counseling, designed to complement traditional role-playing methods in academic training by simulating realistic client interactions. Building on previous work, we introduce a new dataset incorporating adversarial attacks to test the ability of large language models (LLMs) to maintain their assigned roles (role-consistency). The study focuses on evaluating the role consistency and coherence of the Vicuna model's responses, comparing these findings with earlier research. Additionally, we assess and compare various open-source LLMs for their performance in sustaining role consistency during virtual client interactions. Our contributions include creating an adversarial dataset, evaluating conversation coherence and persona consistency, and providing a comparative analysis of different LLMs.