Exploring an Agent Interaction Modeling System (AIMS) for Human Autonomy Teams: Towards the Development of Intelligent Models of Interaction

摘要

This position paper describes an Agent Interaction Model and framework, i.e. the ability to define interaction roles (e.g. professor, student, parent, child, police, civilian) identified by varying amounts of standardized role features (e.g. Curiosity, Responsibility, Deference, Scientific Knowledge) mapped to physical interaction features (e.g. likelihood of interaction, likelihood of knowledge transfer). The framework can be used to enable exploration of amount, types and patterns of knowledge flow through agent interactions based on agent roles. The hypothesis is that in teams of humans and systems there is a distribution of roles that optimizes knowledge transfer. Empathy, i.e. the ability to change roles quickly based on the needs of others, is hypothesized to be an enabler of improved knowledge flow. The framework allows exploration of these hypotheses. In this paper, the authors report on initial development of the model and their proposed next steps to extend the model and the framework for use in human-agent experiments by mapping the features to user-interface components such as text, voice, gesture, appearance, signs and symbols..