A Cognition-based Model for More Forgiving Human-Machine Interactions through Embodied Cooperation

University Of California-Davis, Davis CA

Investigators

Abstract

This award will provide a new framework to promote more forgiving relationships between humans and autonomous machines by blurring the lines between the machine, the operator, and their cooperative actions. As humans we are wired to effectively use autonomous systems as the body and brain employ complex autonomous networks to accomplish every action we perform. These networks are managed by mechanisms that link our intentions, actions, and their sensory outcomes. Together, this create a sense of embodiment; the perception that our body and actions are indeed our own. Most autonomous machines do not communicate or behave in ways that establish these same links. Operators are acutely aware that a machine’s cooperative actions are not their own, and our natural human tendencies to become frustrated with and blame machines when errors occur can promote the abandonment of these same technologies that promise to extend our capabilities. This award supports fundamental research to characterize the relationships between frustration, blame and the embodiment of autonomous machines. It has implications in driving our willingness to cooperate with error-prone autonomous systems across a growing number of applications including prostheses, powered exoskeletons, and other robotic technologies designed to augment human capabilities. This work will provide interdisciplinary research opportunities for underrepresented groups as well as positively impact engineering and neuroscience education. A unique human neuro-robotic model in which participants with upper limb amputation and targeted reinnervation surgery (a neural-machine interface) pilot cooperative robotic limbs will be used. Here, participants can operate artificial limbs by thinking about moving their missing limbs while also feeling movement and touch. It has been previously demonstrated that these sensorimotor channels can be manipulated to promote the embodiment of prostheses. Using this model, it will be investigated how embodiment can be promoted using touch and movement sensory feedback when operating robotic limbs with varying degrees of autonomy and in the face of deteriorating control. Additionally, using cohorts of able-bodied participants and amputees with targeted reinnervation surgery, a robust data set will be built that links measures of embodiment to user frustration and blame while performing cooperative tasks with virtual human-like and non-anthropomorphic robotic limbs. From this data regression modelling techniques will be applied to develop a quantitative index that links the severity of blame and user frustration to the degree of embodiment of cooperative machines. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.