Judging by the Look: The Impact of Robot Gaze Strategies on Human Cooperation

TL;DR

This paper investigates how robot gaze strategies influence human cooperation by developing a neural model for human-like gaze shifts and testing their effects in a cooperative game scenario.

Contribution

It introduces an embodied neural model trained on eye-tracking data to simulate human-like gaze behavior in robots, exploring its impact on human decision-making.

Findings

Gaze strategies significantly affect human cooperation levels

The neural model successfully replicates human gaze patterns

Different gaze behaviors alter human decision-making in social interactions

Abstract

Human eye gaze plays an important role in delivering information, communicating intent, and understanding others' mental states. Previous research shows that a robot's gaze can also affect humans' decision-making and strategy during an interaction. However, limited studies have trained humanoid robots on gaze-based data in human-robot interaction scenarios. Considering gaze impacts the naturalness of social exchanges and alters the decision process of an observer, it should be regarded as a crucial component in human-robot interaction. To investigate the impact of robot gaze on humans, we propose an embodied neural model for performing human-like gaze shifts. This is achieved by extending a social attention model and training it on eye-tracking data, collected by watching humans playing a game. We will compare human behavioral performances in the presence of a robot adopting different gaze strategies in a human-human cooperation game.