Agent-based Models of Multi-lateral Social Interactions

In this Phase I SBIR proposal, Ejenta investigates the use of an agent-based modeling and simulation language to represent, analyze and predict participant activities and beliefs in social interactions. People are represented as individual behavioral actors (i.e. agents) that can reason based on beliefs about the world, act based on beliefs and communicate beliefs with others. We hypothesize that by exploiting models of human work practice we are able to narrow the gaps that arise from sensor data asymmetry. Our approach is to create a"template"practice model in the domain of the social interaction, based on ethnographic observations, video analysis and/or role-playing. We use the template model as input, together with the sensor dataset of a social interaction, to generate a situation-specific model of the interaction. This situation-specific model will include an encoding of the participants, belief changes, activities over time, communication, use of artifacts, movement and location, that together represent the social interaction.