Emotional Supporters often Use Multiple Strategies in a Single Turn

TL;DR

This paper reveals that emotional supporters often use multiple strategies within a single turn in emotional support conversations, leading to a redefinition of the task and improved modeling approaches, including large language models that outperform humans.

Contribution

It introduces a new formulation of the ESC task accounting for multiple strategies per turn and demonstrates the effectiveness of advanced models, especially large language models, in this setting.

Findings

LLMs outperform supervised models and humans in the redefined task

Supporters frequently use multiple strategies, including questions and suggestions, within a single turn

The revised task better captures the complexity of emotional support conversations

Abstract

Emotional Support Conversations (ESC) are crucial for providing empathy, validation, and actionable guidance to individuals in distress. However, existing definitions of the ESC task oversimplify the structure of supportive responses, typically modelling them as single strategy-utterance pairs. Through a detailed corpus analysis of the ESConv dataset, we identify a common yet previously overlooked phenomenon: emotional supporters often employ multiple strategies consecutively within a single turn. We formally redefine the ESC task to account for this, proposing a revised formulation that requires generating the full sequence of strategy-utterance pairs given a dialogue history. To facilitate this refined task, we introduce several modelling approaches, including supervised deep learning models and large language models. Our experiments show that, under this redefined task, state-of-the-art LLMs outperform both supervised models and human supporters. Notably, contrary to some earlier findings, we observe that LLMs frequently ask questions and provide suggestions, demonstrating more holistic support capabilities.