Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Cactus: Towards Psychological Counseling Conversations using Cognitive Behavioral Theory

Suyeon Lee, Sunghwan Kim, Minju Kim, Dongjin Kang, Dongil Yang, Harim Kim, Minseok Kang, Dayi Jung, Min Hee Kim, Seungbeen Lee, Kyoung-Mee Chung, Youngjae Yu, Dongha Lee, Jinyoung Yeo

TL;DR

This work introduces Cactus, a multi-turn dialogue dataset that emulates real-life interactions using the goal-oriented and structured approach of Cognitive Behavioral Therapy (CBT), and creates a diverse and realistic dataset.

Abstract

Recently, the demand for psychological counseling has significantly increased as more individuals express concerns about their mental health. This surge has accelerated efforts to improve the accessibility of counseling by using large language models (LLMs) as counselors. To ensure client privacy, training open-source LLMs faces a key challenge: the absence of realistic counseling datasets. To address this, we introduce Cactus, a multi-turn dialogue dataset that emulates real-life interactions using the goal-oriented and structured approach of Cognitive Behavioral Therapy (CBT). We create a diverse and realistic dataset by designing clients with varied, specific personas, and having counselors systematically apply CBT techniques in their interactions. To assess the quality of our data, we benchmark against established psychological criteria used to evaluate real counseling sessions, ensuring alignment with expert evaluations. Experimental results demonstrate that Camel, a model trained with Cactus, outperforms other models in counseling skills, highlighting its effectiveness and potential as a counseling agent. We make our data, model, and code publicly available.

Cactus: Towards Psychological Counseling Conversations using Cognitive Behavioral Theory

TL;DR

This work introduces Cactus, a multi-turn dialogue dataset that emulates real-life interactions using the goal-oriented and structured approach of Cognitive Behavioral Therapy (CBT), and creates a diverse and realistic dataset.

Abstract

Recently, the demand for psychological counseling has significantly increased as more individuals express concerns about their mental health. This surge has accelerated efforts to improve the accessibility of counseling by using large language models (LLMs) as counselors. To ensure client privacy, training open-source LLMs faces a key challenge: the absence of realistic counseling datasets. To address this, we introduce Cactus, a multi-turn dialogue dataset that emulates real-life interactions using the goal-oriented and structured approach of Cognitive Behavioral Therapy (CBT). We create a diverse and realistic dataset by designing clients with varied, specific personas, and having counselors systematically apply CBT techniques in their interactions. To assess the quality of our data, we benchmark against established psychological criteria used to evaluate real counseling sessions, ensuring alignment with expert evaluations. Experimental results demonstrate that Camel, a model trained with Cactus, outperforms other models in counseling skills, highlighting its effectiveness and potential as a counseling agent. We make our data, model, and code publicly available.
Paper Structure (86 sections, 24 figures, 13 tables)

This paper contains 86 sections, 24 figures, 13 tables.

Table of Contents

  1. Introduction
  2. Design Considerations for Cactus
  3. Can LLMs be Competent Counselors?
  4. AI counselors have limitations in selecting CBT techniques.
  5. AI counselors tend to suggest direct reframing of the clients' thoughts.
  6. Can LLMs act like Real Clients?
  7. AI clients tend to express the provided information explicitly.
  8. AI clients tend to be overly positive.
  9. Cactus: A Psychological Counseling Dataset using CBT
  10. Dataset Construction
  11. Source dataset.
  12. Counselor simulation.
  13. Client simulation.
  14. Dialogue generation process.
  15. Filtering.
  16. ...and 71 more sections

Figures (24)

  • Figure 1: Comparing a previous counseling dataset with real-world scenarios: The dataset shows counselors conveying large amounts of information in a single turn, whereas real-world counseling involves active, collaborative communication between a counselor and client.
  • Figure 2: Comparison of the distribution of CBT techniques selected by GPT-4o and psychological experts. The results of GPT-3.5-Turbo are shown in Figure \ref{['fig:cbt_tech_distribution_chatgpt']}.
  • Figure 3: Empirical investigations into the problems of using ChatGPT as an AI counselor and AI client. Details of experiments are in Appendix \ref{['app:desing_consider2']}.
  • Figure 4: The overview of the data collection process of Cactus.
  • Figure 5: Comparision of Two-Agent mode and Script mode.
  • ...and 19 more figures