Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

CauESC: A Causal Aware Model for Emotional Support Conversation

Wei Chen, Hengxu Lin, Qun Zhang, Xiaojin Zhang, Xiang Bai, Xuanjing Huang, Zhongyu Wei

TL;DR

CauESC introduces a causal-aware framework for Emotional Support Conversation that explicitly identifies emotion causes and reasons about emotion effects within a dialogue. The model combines a cause-aware encoder, a causal interaction module leveraging COMET-based effects, and independent-integrated strategy executors to generate strategically guided, empathetic responses. Empirical results on the ESConv dataset show state-of-the-art performance in both strategy selection and response quality, supported by comprehensive ablations and human evaluations. The approach enables fine-grained emotion understanding from cause to effect and demonstrates practical impact for more effective emotional support in conversational AI.

Abstract

Emotional Support Conversation aims at reducing the seeker's emotional distress through supportive response. Existing approaches have two limitations: (1) They ignore the emotion causes of the distress, which is important for fine-grained emotion understanding; (2) They focus on the seeker's own mental state rather than the emotional dynamics during interaction between speakers. To address these issues, we propose a novel framework CauESC, which firstly recognizes the emotion causes of the distress, as well as the emotion effects triggered by the causes, and then understands each strategy of verbal grooming independently and integrates them skillfully. Experimental results on the benchmark dataset demonstrate the effectiveness of our approach and show the benefits of emotion understanding from cause to effect and independent-integrated strategy modeling.

CauESC: A Causal Aware Model for Emotional Support Conversation

TL;DR

CauESC introduces a causal-aware framework for Emotional Support Conversation that explicitly identifies emotion causes and reasons about emotion effects within a dialogue. The model combines a cause-aware encoder, a causal interaction module leveraging COMET-based effects, and independent-integrated strategy executors to generate strategically guided, empathetic responses. Empirical results on the ESConv dataset show state-of-the-art performance in both strategy selection and response quality, supported by comprehensive ablations and human evaluations. The approach enables fine-grained emotion understanding from cause to effect and demonstrates practical impact for more effective emotional support in conversational AI.

Abstract

Emotional Support Conversation aims at reducing the seeker's emotional distress through supportive response. Existing approaches have two limitations: (1) They ignore the emotion causes of the distress, which is important for fine-grained emotion understanding; (2) They focus on the seeker's own mental state rather than the emotional dynamics during interaction between speakers. To address these issues, we propose a novel framework CauESC, which firstly recognizes the emotion causes of the distress, as well as the emotion effects triggered by the causes, and then understands each strategy of verbal grooming independently and integrates them skillfully. Experimental results on the benchmark dataset demonstrate the effectiveness of our approach and show the benefits of emotion understanding from cause to effect and independent-integrated strategy modeling.
Paper Structure (32 sections, 13 equations, 5 figures, 6 tables)

This paper contains 32 sections, 13 equations, 5 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Emotional Support Conversation Model
  4. Commonsense Reasoning
  5. Emotion Cause Extraction
  6. Problem Formulation
  7. Approach
  8. Cause Aware Encoder
  9. Emotion Cause Encoding
  10. Emotion Effect Acquisition
  11. Causal Interaction Module
  12. Independent-Integrated Strategy Executors
  13. Experiment
  14. Dataset
  15. Evaluation Metrics
  16. ...and 17 more sections

Figures (5)

  • Figure 1: An Emotional Support Conversation Example, which intra and inter effects gradually change the emotional state of the seeker.
  • Figure 2: The overview of the proposed CauESC which consists of a cause aware encoder, a causal interaction module, and independent-integrated strategy executors. Gray dotted curves indicate the detail of the corresponding module.
  • Figure 3: The strategy distribution on ESConv.
  • Figure 4: The strategy distribution at different conversation progress on ESConv.
  • Figure 5: The few-shot prompt of ChatGPT.