A Cognitive Model that Derives Diverse Interpretations from a Single Fact for Narrative Approach

Abstract

In this paper, we present a novel computational cognitive model that enables robots to generate multiple interpretations of single events―a key capability for narrative therapy applications. Despite narrative therapy’s success across clinical, educational, and other domains, existing robotic counselors lack the perspective-taking abilities essential for this approach. Our model addresses this gap by incorporating cognitive bias mechanisms that simulate how different individuals interpret the same facts based on their experiential frameworks. Through simulation experiments with six distinct cognitive bias profiles interpreting a three-character scenario, we demonstrate that our model successfully generates diverse, bias-consistent narratives. Moreover, observers with similar cognitive biases produce similar interpretations, validating our approach. This work advances human-robot interaction by providing a computational foundation for therapeutic robots that can understand and respond to the multiplicity of human perspectives.