Human-AI friendship: Rejecting the appropriate sentimentality criterion

Abstract

Most traditional philosophical views of friendship deny the possibility of human-AI friendship because they consider feeling love or something similar towards the other to be a requirement for genuine friendship. We call this the appropriate sentimentality criterion for friendship. Coupled with the claim that AI cannot and will not ever be able to feel what a friend should feel, the appropriate sentimentality criterion is the key to an argument that genuine human-AI friendship will never be possible. We argue against the requirement of appropriate sentimentality for friendship, suggesting that the feelings commonly associated with friendship are only a proxy for what really matters. We go on to present an inclusive account of friendship that requires only mutual positive intentions and a preponderance of rewarding interactions. We conclude that the appropriate sentimentality criterion for friendship should be rejected and that real human-AI friendships are currently possible.