The role of appearance and motion in action prediction

Abstract

We used a novel stimulus set of human and robot actions to explore the role of humanlike appearance and motion in action prediction. Participants viewed videos of familiar actions performed by three agents: human, android and robot, the former two sharing human appearance, the latter two nonhuman motion. In each trial, the video was occluded for 400 ms. Participants were asked to determine whether the action continued coherently (in-time) after occlusion. The timing at which the action continued was early, late, or in-time (100, 700 or 400 ms after the start of occlusion). Task performance interacted with the observed agent. For early continuations, accuracy was highest for human, lowest for robot actions. For late continuations, the pattern was reversed. Both android and human conditions differed significantly from the robot condition. Given the robot and android conditions had the same kinematics, the visual form of the actor appears to affect action prediction. We suggest that the selection of the internal sensorimotor model used for action prediction is influenced by the observed agent's appearance.

Visual inference of arm movement is constrained by motor representations

“…Previous studies of our group [1][2][3] evidenced that ending movement estimation relies on endogenous input throughout the recall of internal models of action rather than the quantity of visual input [2]. These results support the more general idea that perception is action-dependent and that motor competencies are involved in such predictive tasks [4,5], through the recruitment of a frontoparietal action perception network [6].…”

Movement kinematics affect action prediction: comparing human to non-human point-light actions

“…The human shape of the point-light agent in the present study might have led observers to expect human movement dynamics (see Saygin and Stadler 2012). Thus, the artificial condition employed a highly familiar agent with an unfamiliar movement.…”

“…For the current experiment, this means that the perception of a human body may have activated associated motor programs. Saygin and colleagues (2011) and Saygin and Stadler (2012) discuss in more detail how top-down expectancies induced by the agent's appearance could influence action perception and prediction. They investigated differences in action prediction accuracy between a human agent, an android with a realistic human appearance (i.e., skin, hair, clothes etc.)…”

Simulating and predicting others’ actions

“…Good performance on this task requires detailed simulation of the action continuation during occlusion, and factors that change performance can then be measured. For example, Saygin and Stadler (2012) examine the influence of human, robot or humanoid form and motion on action prediction, and suggest that slower prediction is used for robot actions compared to human actions.…”