Designing a Socially Assistive Robot to Preserve the Dignity of Older Adults

Abstract

The population of older adults worldwide is expected to double between 2025 and 2050, whereas the total population is projected to grow only 34% over the same period. Many of these older adults prefer to age in place, to remain independent and living in their homes. However, many older adults need regular assistance to manage the activities necessary for independent living. With the likely decrease in availability of healthy adults to assist in the care of older adults, there will likely be a growing gap between the need for assistance and the availability of assistance.

These population trends are at least in part responsible for a growing interest in developing social robots that can assume some of the responsibilities and workload of human caregivers. However, in addition to benefiting humans, these robots also have the potential of impinging upon the dignity and autonomy of the person being assisted. To address this challenge, I have proposed the RSA Framework for designing and evaluating socially assistive robots for older adults. Each component of this framework -- Robot, Self, and Alliance -- takes a human-centric approach that prioritizes the perspective of the person being assisted. The rest of my dissertation then addresses each of these components. I present my work on a robot architecture, an approach to varying assistance to support feelings of the self, and relational behaviors that the robot may use to build rapport and facilitate the development of an alliance.

In the robot architecture, I introduce a multi-level verification of goals and actions that restricts the robot from pursuing forbidden goals and selecting actions that may cause harm (physical or otherwise). Also, the architecture has a mechanism to select an appropriate assistance for the robot to provide based on the need of the person that the robot is assisting. This mechanism is designed to support the autonomy of the individual while providing assistance on an important activity for independent living. I demonstrate this architecture with a robot that assists a person with sorting medications. The robot observes a person sorting medications, infers who much assistance the person needs to continue to make progress on the activity, and provides assistance that matches this need.

Lastly, I present my work on developing human-robot rapport using relational behaviors the robot performs to influence the human experience of rapport with the robot. In evaluating the relational behaviors in the context of a robot administering a personality survey, I found that a robot that is more engaging and provides verbal acknowledgment improves the experience of rapport.