Cognitive Policy-Driven LLM for Diagnosis and Intervention of Cognitive Distortions in Emotional Support Conversation

2026-04-18

AI Summary

This study addresses a critical limitation in existing emotional support dialogue systems, which often overlook cognitive distortions in help-seekers and provide only superficial emotional reassurance, thereby failing to alleviate deep-seated psychological distress. To bridge this gap, the authors integrate cognitive behavioral therapy principles into large language models, introducing CoPoLLM—a cognitive strategy–driven framework that enables precise diagnosis and targeted intervention. They also construct CogBiasESC, the first dataset annotated with cognitive distortion types, severity levels, and safety risk ratings. Experimental results demonstrate that the proposed approach significantly outperforms 15 state-of-the-art baselines in cognitive distortion recognition accuracy, intervention efficacy, and safety risk mitigation, establishing a novel paradigm that jointly optimizes diagnostic precision, therapeutic intervention, and safety assurance.

Technology Category

• Cognitive Modeling & Cognitive Systems: Social Cognition And Interaction
• Natural Language Processing: Safety and Robustness
• Intelligent Robots: Cognitive Robotics

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Emotional Support Conversation (ESC) plays a critical role in mental health assistance by providing accessible psychological support in real-world applications. Large Language Models (LLMs) have shown strong empathetic abilities in ESC tasks. Yet, existing methods overlook the issue of cognitive distortions in help-seekers' expressions. As a result, current models can only provide basic emotional comfort, rather than helping help-seekers address their psychological distress at a deeper cognitive level. To address this challenge, we construct the CogBiasESC dataset, the first dataset that expands existing ESC datasets by adding labels for cognitive distortions, includes their type, intensity, and safe risk level. Furthermore, we propose the Cognitive Policy-driven Large Language Model framework (CoPoLLM) to enhance LLMs' ability to diagnose and intervene cognitive distortions in help-seekers. We also analyze the safety advantages of CoPoLLM from a theoretical perspective. Experimental results show that CoPoLLM significantly outperforms 15 state-of-the-art baselines in terms of distortion diagnosis accuracy, intervention strategy effectiveness, and safety risk control.

Problem

• Cognitive Distortions
• Emotional Support Conversation
• Large Language Models
• Mental Health Assistance

Innovation

• Cognitive Distortions
• Emotional Support Conversation
• Large Language Models
• Policy-Driven Framework
• Mental Health Intervention