Abstract
Humans naturally pay attention to other animate agents in their environment, a prosocial behavior that has been documented as early as a few weeks. What internal mechanisms give rise to this behavior? A standard hypothesis is that the human brain has a built-in module that specifically detects animacy from visual input. We hypothesize that animate attention naturally arises from a more general process of curiosity driven learning. We ran experiments to measure important features of animate attention. Observers (N = 12, A gerange = 43 years) wore a mobile eye tracker while watching a display consisting of four self-propelled, spherical robots travelling along a mat. Using kinematics alone, the robots were made to appear animate, random, periodic, or static. Average fixations proportions were animate 55%, random 24%, periodic 14%, and static 7%, with 10 of 12 observers fixating most on animate. We also administered an autism assessment scale and found that observers scoring high on this scale attended to animate robots significantly less (r = −0.65, p = 0.02), suggesting that our attentional-fraction metric is an indicator of variance in social behavior. We then built a neural network agent embodying concepts of intrinsic curiosity, and embedded it within a virtual environment emulating the real world robot display. The agent was tasked to predict the virtual robot trajectories and was constrained to focus on one robot at a time. Results show that the neural network agent produces an aggregate attentional fixation pattern identical to that of human adults. Crucially, the network achieves this without the requirement for any specific built-in modules favoring animacy. Instead, the pattern arises from the agent’s discovery of the inherent relative “interestingness” of animacy as compared to random, periodic, and static motion. More broadly, our results suggest that key characteristics of social behavior may emerge naturally from intrinsically-motivated learning agents.