Multi-View Attention Multiple-Instance Learning Enhanced by LLM Reasoning for Cognitive Distortion Detection

2025-09-22 · Source: cs.AI updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, AI for Mental Health Applications · Depth: Expert, quick

Summary

A new framework has been developed for the automatic detection of cognitive distortions, which are often linked to mental health disorders. This framework integrates Large Language Models (LLMs) with a Multiple-Instance Learning (MIL) architecture to improve interpretability and expression-level reasoning. It operates by decomposing each utterance into Emotion, Logic, and Behavior (ELB) components. LLMs then process these components to infer multiple distortion instances, each assigned a predicted type, expression, and a model-assigned salience score. These instances are subsequently integrated using a Multi-View Gated Attention mechanism for final classification. Experimental evaluations on both Korean (KoACD) and English (Therapist QA) datasets indicate that incorporating ELB components and LLM-inferred salience scores significantly enhances classification performance, particularly for distortions with high interpretive ambiguity. The dataset and implementation details are publicly available.

Key takeaway

For research scientists developing NLP solutions for mental health, this framework offers a robust approach to tackle the inherent ambiguity in cognitive distortion detection. By integrating LLMs with a MIL architecture and ELB components, you can achieve more accurate and interpretable results, especially for nuanced distortions. Consider adopting this psychologically grounded method to enhance fine-grained reasoning in your mental health NLP applications, leveraging the publicly available dataset and implementation for rapid prototyping.

Key insights

Combining LLMs with MIL and ELB components improves cognitive distortion detection by enhancing interpretability and reasoning.

Principles

• Decompose utterances into ELB components.
• LLM-inferred salience scores improve classification.
• Multi-View Gated Attention integrates instances.

Method

The method involves decomposing utterances into Emotion, Logic, and Behavior (ELB) components, using LLMs to infer distortion instances with type, expression, and salience, and then integrating these instances via Multi-View Gated Attention for classification.

In practice

• Apply ELB decomposition to text analysis.
• Utilize LLMs for fine-grained instance inference.
• Employ Multi-View Attention for complex data integration.

Topics

• Cognitive Distortion Detection
• Large Language Models
• Multiple-Instance Learning
• Multi-View Gated Attention
• Emotion, Logic, Behavior

Best for: Research Scientist, AI Scientist, Machine Learning Engineer, NLP Engineer

Related on AIssential

• Listening Between the Lines: Joint Learning of ASR Embeddings and LLM-Augmented Linguistics for Dementia Detection — Jointly learning ASR embeddings and LLM-augmented linguistics significantly improves dementia detection from speech.
• Formalizing and Mitigating Structural Distortion in LLM Attention for Zero-Shot Graph Reasoning — LLM performance on graphs is bottlenecked by structural distortion from sequence linearization, correctable by attention alignment.
• Mechanistic Decoding of Cognitive Constructs in LLMs — LLMs encode complex emotions like jealousy as structured linear combinations of psychological antecedents.
• QG-MIL: A Gated Transformer Aggregator for Domain-Agnostic Multiple Instance Learning in Medical Imaging — QG-MIL stabilizes attention in MIL aggregators for medical imaging, improving prediction stability and cross-domain performance.
• Lazy or Efficient? Towards Accessible Eye-Tracking Event Detection Using LLMs — An LLM-driven pipeline automates eye-tracking event detection, making complex analysis accessible without specialized programming.
• Mental-R1: Aligning LLM Reasoning for Mental Health Assessment — Aligning LLM reasoning with human cognitive processes significantly improves mental health assessment accuracy.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.AI updates on arXiv.org.