Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

When and How to Express Empathy in Human-Robot Interaction Scenarios

Christian Arzate Cruz, Edwin C. Montiel-Vazquez, Chikara Maeda, Randy Gomez

TL;DR

This paper introduces whEE, a three-component framework that determines when and how a robot should express empathy in human-robot interaction, leveraging large language models to detect empathy cues and generate adaptive empathetic responses for the Haru robot. It combines behavioral empathy cues, EPITOME-based text generation, and a structured prompt design to classify empathy direction (seeking/providing/none) and to produce context-appropriate responses, evaluated across non-HRI and HRI datasets, including Empathetic Dialogues with a Robot and The Talking Room. The results show that LLMs are strong at identifying when empathy is sought, while providing empathy remains more challenging, with Haru able to modulate empathy in HRI scenarios and produce more empathetic utterances when appropriate. The work advances socially intelligent robot design by offering a practical framework for context-aware empathy that can mitigate empathy burnout and improve interaction quality in diverse scenarios.

Abstract

Incorporating empathetic behavior into robots can improve their social effectiveness and interaction quality. In this paper, we present whEE (when and how to express empathy), a framework that enables social robots to detect when empathy is needed and generate appropriate responses. Using large language models, whEE identifies key behavioral empathy cues in human interactions. We evaluate it in human-robot interaction scenarios with our social robot, Haru. Results show that whEE effectively identifies and responds to empathy cues, providing valuable insights for designing social robots capable of adaptively modulating their empathy levels across various interaction contexts.

When and How to Express Empathy in Human-Robot Interaction Scenarios

TL;DR

This paper introduces whEE, a three-component framework that determines when and how a robot should express empathy in human-robot interaction, leveraging large language models to detect empathy cues and generate adaptive empathetic responses for the Haru robot. It combines behavioral empathy cues, EPITOME-based text generation, and a structured prompt design to classify empathy direction (seeking/providing/none) and to produce context-appropriate responses, evaluated across non-HRI and HRI datasets, including Empathetic Dialogues with a Robot and The Talking Room. The results show that LLMs are strong at identifying when empathy is sought, while providing empathy remains more challenging, with Haru able to modulate empathy in HRI scenarios and produce more empathetic utterances when appropriate. The work advances socially intelligent robot design by offering a practical framework for context-aware empathy that can mitigate empathy burnout and improve interaction quality in diverse scenarios.

Abstract

Incorporating empathetic behavior into robots can improve their social effectiveness and interaction quality. In this paper, we present whEE (when and how to express empathy), a framework that enables social robots to detect when empathy is needed and generate appropriate responses. Using large language models, whEE identifies key behavioral empathy cues in human interactions. We evaluate it in human-robot interaction scenarios with our social robot, Haru. Results show that whEE effectively identifies and responds to empathy cues, providing valuable insights for designing social robots capable of adaptively modulating their empathy levels across various interaction contexts.

Paper Structure

This paper contains 31 sections, 5 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Empathy
  4. Empathy Cues
  5. Empathetic Robots
  6. Empathy Classifiers
  7. Empathetic Text Generation
  8. The whEE Framework
  9. When
  10. Assumptions
  11. How
  12. Assumptions
  13. Human-Robot Exchanges
  14. Assumptions
  15. Empathy Background
  16. ...and 16 more sections

Figures (5)

  • Figure 1: Overview of the design of our whEE framework.
  • Figure 2: The EDR dataset setting.
  • Figure 3: The Talking Room.
  • Figure 4: Empathy cues distribution comparison across HRI and non-HRI scenarios. Baseline methods (established affective computing methods) are shown in blue and LLaMA 3.3 in pink.
  • Figure 5: Empathy cues distribution comparing regular Haru (green) and empathetic Haru (orange).