PSY-STEP: Structuring Therapeutic Targets and Action Sequences for Proactive Counseling Dialogue Systems
Summary: arXiv:2604.04448v1 Announce Type: new
Abstract: Cognitive Behavioral Therapy (CBT) aims to identify and restructure automatic negative thoughts pertaining to involuntary interpretations of events, yet existing counseling agents struggle to identify and address them in dialogue settings. To bridge this gap, we introduce STEP, a dataset that models CBT counseling by explicitly reflecting automatic thoughts alongside dynamic, action-level counseling sequences. Using this dataset, we train STEPPER, a counseling agent that proactively elicits automatic thoughts and executes cognitively grounded interventions. To further enhance both decision accuracy and empathic responsiveness, we refine STEPPER through preference learning based on simulated, synthesized counseling sessions. Extensive CBT-aligned evaluations show that STEPPER delivers more clinically grounded, coherent, and personalized counseling compared to other strong baseline models, and achieves higher counselor competence without inducing emotional disruption.
Introduction
The development of artificial intelligence (AI) in mental health care has gained significant traction in recent years, particularly in the implementation of dialogue systems aimed at providing therapeutic support. Among these, Cognitive Behavioral Therapy (CBT) has emerged as a leading framework, focusing on the identification and restructuring of automatic negative thoughts. However, traditional counseling agents often struggle to effectively engage with these thoughts in real-time dialogue.
Introducing STEP
To address the limitations faced by existing counseling technology, the authors introduce the STEP dataset. This innovative dataset is designed to model CBT counseling by:
- Explicitly reflecting automatic thoughts.
- Incorporating dynamic, action-level counseling sequences.
By leveraging these features, researchers aim to create a more effective dialogue system that can engage users in a meaningful therapeutic process.
Development of the STEPPER Agent
Building upon the STEP dataset, the team has developed STEPPER, a counseling agent that proactively elicits automatic thoughts from users. The agent is designed to execute interventions grounded in cognitive therapy principles, ensuring that interactions are both relevant and impactful.
To enhance the performance of STEPPER, the researchers employed preference learning techniques. This method involved using simulated, synthesized counseling sessions to train the agent, allowing it to refine its decision-making capabilities and empathic responses.
Evaluation and Results
Extensive evaluations aligned with CBT principles have been conducted to assess STEPPER’s effectiveness. The findings reveal that STEPPER significantly outperforms other leading baseline models in several key areas, including:
- Clinical grounding of responses.
- Coherence in dialogue.
- Personalization of therapeutic interactions.
- Overall counselor competence without causing emotional disruption.
These results highlight the potential for STEPPER to serve as a vital tool in the future of mental health counseling, offering a more responsive and tailored approach to user needs.
Conclusion
The introduction of PSY-STEP and the STEPPER agent marks a significant advancement in the field of AI-driven mental health support. By focusing on the intricacies of CBT and automating the elicitation of automatic thoughts, this research paves the way for more effective therapeutic dialogue systems. As technology continues to evolve, the integration of such proactive counseling agents could transform the landscape of mental health care, providing individuals with accessible and personalized support.