Computer Science, Rutgers
Realtime co-adaptation of external media and sensory motor control in closed loop as a gateway into the hidden potentials of the non-verbal autistic child
One of the open questions in dealing with the low-functioning part of the population with Autism Spectrum Disorders (ASD) is how we can facilitate learning under the condition that no verbal instructions can be used and that no verbal feedback can be given by the low-functioning individual. In this work we explore the possibility of eliciting exploratory and eventually intentional behaviors with minimal or no explicit instructions. We designed a system that responds to unconstrained human motion and has a hidden uninstructed task that the subject has to figure out in order to get a sensory reward, such as a video or audio of their preference. We found that non-verbal low-functioning ASD population is capable of figuring out the hidden task. Moreover, we found that the media of greater preference for the subject elicited more statistically predictive motor signatures. This difference in the motor signatures can be treated as the feedback from the subject, which would allow us to pick better reward media, thus increasing motivation of the subject.
In our previous work we have shown that in general the motor signatures of the individuals with ASD are less statistically predictive than the motor signatures of the typically developing individuals. If the change in the motor signatures towards a more predictive range can be made persistent (i.e. present in absence of the reward media) that would make the movements of the ASD individuals closer to those of the typically developing population.