Exploring Data-Driven Components of Socially Intelligent AI through Cooperative Game Paradigms

Casey Bennett,Eunseo Yoon,Jaeyoung Suh,Yejin Chae,Jihong Jeong,Benjamin Weiss

doi:10.3390/mti6020016

Abstract

The development of new approaches for creating more “life-like” artificial intelligence (AI) capable of natural social interaction is of interest to a number of scientific fields, from virtual reality to human–robot interaction to natural language speech systems. Yet how such “Social AI” agents might be manifested remains an open question. Previous research has shown that both behavioral factors related to the artificial agent itself as well as contextual factors beyond the agent (i.e., interaction context) play a critical role in how people perceive interactions with interactive technology. As such, there is a need for customizable agents and customizable environments that allow us to explore both sides in a simultaneous manner. To that end, we describe here the development of a cooperative game environment and Social AI using a data-driven approach, which allows us to simultaneously manipulate different components of the social interaction (both behavioral and contextual). We conducted multiple human–human and human–AI interaction experiments to better understand the components necessary for creation of a Social AI virtual avatar capable of autonomously speaking and interacting with humans in multiple languages during cooperative gameplay (in this case, a social survival video game) in context-relevant ways.

Highlights

There are many ways in which two agents can interact, from cooperatively to competitively and many variations in between
We focused on identification of missing interaction components beyond the speech system content itself
Each category and/or sub-category in the hierarchy became the basis of various “use cases” that we can use to define capabilities and goals for the artificial intelligence (AI) agent, so that the autonomous social interaction of the virtual avatar and in-game actions are aligned, without the need for a

Summary

Introduction

There are many ways in which two agents (whether artificial or human) can interact, from cooperatively to competitively and many variations in between. There is a growing interest in understanding how we can create artificial intelligence (AI) that emulates natural human social behavior in diverse settings to produce better interactive technology [1] Those various types of interactions provide different arenas to explore that question, such as through competitive or cooperative game play. One advantage of cooperative game paradigms is that they expose the constraints between the designer’s conception of sociality in AI and the user’s embedded expectations, since in a cooperative game environment, the AI agent and human user must work together to achieve some goal in a “coordinated” fashion [2] Those constraints include indirect non-verbal aspects of the otherwise direct verbal interactions that form a core part of the interaction context within which the communication content must be understood.

Methods

Results

Conclusion