AAAI Publications, 2017 AAAI Fall Symposium Series

Font Size: 
Towards Socially Intelligent HRI Systems: Quantifying Emotional, Social, and Relational Context in Real-World Human Interactions
Jesse Parent, Yelin Kim

Last modified: 2017-10-09

Abstract


In this paper, we present current issues and findings for developing HRI systems with capabilities to measure and interpret emotional, social, and relational context in real-world human interactions. We discuss our ongoing work to create a dataset that will contribute to achieve this overarching goal. Our source data are documentary films of face-to-face interactions of two-person pairings of real relationships. We select an example video from the source data to explore labeling methods to describe its emotional, social, and relational phenomena. Extending from previous studies, we propose a social label, synthesizing emotion and relationship labels to provide interpretable descriptions of social-relational context in human interaction. We demonstrate how our proposed social label, engagement label, is associated with salient emotional and social dynamics during interactions. We further discuss open questions and provide insight into future research directions where a robust and multifaceted approach is necessary. Building from past efforts in classifying and interpreting affective and relational context, this paper opens a new gateway for developing HRI systems that can understand and adapt to real-world interactions.

Keywords


emotion recognition; engagement; social context;

Full Text: PDF