Murder on the Dancefloor? AI, Embodiment, and Remorse

When we see someone on the dance-floor we learn something about them: their personality—their mind—is revealed in how they move. We are fundamentally embodied creatures and become aware of each other’s—and even our own—thoughts and feelings by means of the way they are expressed on our faces and by our bodies. In this paper, I will explore the significance of this mind-body connection for the debate about AI welfare. Drawing on my previous discussions of the Turing Triage test, I will argue that a necessary condition of attributing moral standing to machines is that we could conceive of someone being wracked by remorse after realising they had “killed” an AI. Imagine that, while standing on the edge of the dance floor, I summon a chatbot to ask it for some feedback on my attempts to do the Funky Chicken, and then, convinced of my new found prowess, end the chat session. An emerging school of thought in AI ethics implies that when I close the chat session, I murder a sentient being. Yet if, upon realising this, I fall to the (dance) floor weeping, yelling, and tearing my hair out, most people, would, I suspect, think that I was deranged. If I spend a year wracked by remorse, people would struggle to understand why. I argue that this is because AI is not embodied in the right way to allow us to use the language of remorse, or other related, moral emotions reliably. I will distinguish this argument from the “relational turn” advocated by Gunkel and Coeckelbergh and draw out some of its counter-intuitive consequences, as well as its virtues.