Abstract:
In an age of autonomous driving and robotics, we are increasingly engaging with robots that deploy autonomous assistance. Cognitive science and human-computer interaction literature tells us that, when we apply autonomy in assistive settings, we are often augmenting the user’s sense of agency over the system. Sense of agency is a phenomenon from cognitive science that represents the experience of being in control of one’s environment. If our assistive systems had some notion of the user’s sense of agency, we could deploy robotic assistance that maintains the user’s feeling of control over their assistive device, which is an important factor in technology adoption and use. To build systems that deploy assistance while maintaining a strong user sense of agency during the interaction, we must understand assistive robot users’ experiences of control in the context of the assistance the robot provides.

Although we know that sense of agency experiences are prevalent in human-machine interactions, assistive robotics literature rarely addresses this concept explicitly. Gap 1: The assistive robotics literature lacks an understanding of the subjective sense of agency experience among users with disabilities during assistive interactions. Gap 2: Additionally, we lack measures of sense of agency that can be computed online, which might enable us to monitor and maintain sense of agency throughout assistive interactions. Gap 3: Finally, assuming we had a real-time measure for sense of agency, it is unclear how we could use the measures to dynamically adjust the level of assistance throughout assistive interactions. This thesis addresses the above three research gaps in two assistance domains: physical and cognitive assistance.

To quantitatively address Gap 1, we run experiments in both assistance domains that investigate the effects of level of assistance on the user’s sense of agency. To qualitatively address Gap 1, we will employ semi-structured interviews and scenario-based design techniques to query participants with disabilities about common scenarios in which assistive technology violates their sense of agency. For Gap 2, we propose real-time proxy metrics for sense of agency based on cognitive science theories in both the physically and cognitively assistive domains. Finally, for Gap 3, we implement novel assistive paradigms that dynamically adjust the level of assistance based on an estimation of the user’s sense of agency, which is computed from our proxy metrics. These novel paradigms are then compared to conventional assistive systems which do not consider sense of agency.

In this thesis, we contribute a quantitative and qualitative understanding of the subjective experience of sense of agency among assistive robot users with disabilities in the cases of physical and cognitive assistance. Additionally, we contribute some of the first real-time proxy metrics for sense of agency in assistive robotics. Finally, we contribute novel assistance paradigms that consider the user’s sense of agency when providing assistance to the user, and we compare these paradigms to conventional assistance approaches. We hope to build assistive systems that respect the user’s sense of agency, potentially leading to improved assistive interactions that leave users feeling in control of their assistive devices.

Thesis Committee Members:
Henny Admoni, Chair
Aaron Steinfeld
Patrick Carrington
Elaine Short, Tufts University

More Information