Exploring the Relation Between Stiffness Perception and Action Using Models and Artificial Neural Networks

Abstract

We use sensory feedback to form our perception, and control our movements and forces (actions). There is an ongoing debate about the relation between perception and action, with evidence in both directions. For example, there are cases in which perceptual illusions affect action signals and cases where they do not. However, even when they do, it is unknown if perceptual information can be inferred from action signals alone. To answer this question, we utilized a perceptual illusion created by artificial tactile skin stretch, which increases stiffness perception, and affects grip force. We used data recorded in a stiffness discrimination task in which participants compared pairs of virtual objects, comprised of force and artificial skin stretch and indicated which they perceived as stiffer. We explored if models could predict participants' perceptual responses, and the increase in stiffness perception caused by the skin stretch, solely from their recorded action signals. That is, with no information provided about the stimuli. We found that participants' perceptual augmentation could be predicted to an extent from their action signals alone. We predicted the average augmentation effect across participants, and a general trend of increased predicted perceptual augmentation for increased real perceptual augmentation. These results indicate that at least some perceptual information is present in action signals. Furthermore, of the action signals examined, grip force was necessary for predicting the augmentation effect, and a motion signal (e.g., position) was needed for predicting human-like perception, shedding light on what information may be present in the different signals.