Project Details: Representing Social Interaction

Researcher: Alan Wagner

This project explores the challenge of computational representating a robot or agent's interactions. A representation of interaction suitable for implementation of a robot must be computational in nature, and yet, if the robot's interaction will also involve humans, then this representation must have meaningful connections to social psychology. For many reasons, we settled on the outcome matrix (aka the normal-form game) as a suitable representation of interaction which is implementable on a robot. The outcome matix depicted to the right explicitly represents important information about an interaction, such as whom is interacting and how the selction of specific actions will impact the other individual. Moreover, this representation has an extensive history in the fields of game-theory, economics, operations research, neuroscience, and psychology. It is therefore a strong candidate for use on a robot.

The goals of this project are to develop algorithms and software that allow human-robot interaction researchers to 1) create outcome matrices of any and every situation the robot encounters and 2) use outcome matrix notation to describe their experiments formally in manner useful and repeatable by other scientists.




Software for this project is going to be released as part of an interdependence API coming in the next couple of months.