What Cares?

Abstract

Abstract Chapter 8 imagines future challenges of designing and deploying carebots for eldercare and related artificial intelligence (AI) applications. We examine the objection that sociable robots cannot function adequately as caregivers because they do not care about care recipients. We show that what matters most is that care gets expressed, and we defend a version of care ethics which holds that the form care takes varies, depending on the capacities, contexts, and needs of those who give and receive care. While robotics literature stresses using machines to extend autonomy, we stress extending care and supporting not only autonomy but a range of human capabilities. We evaluate the goal of aligning machine behavior with human values (values alignment) and argue that this leads to algorithmic biases. We consider whether sophisticated AI should be trusted, since it operates as a black box, and address concerns about the ethics of robot deception.