Human–AI Friendship

Abstract

Abstract This article critically assesses four major arguments currently discussed against the possibility or desirability of human–AI friendships, rejects these arguments, and provides reasons why friendship with AI both may be possible in a philosophically meaningful way and could prove desirable. The first argument proposes that there is at least one conceptual condition for friendship—either a property of the friends in question or a feature of friendships between friends—that AI may not reach. The second one is that the consequences of human–AI friendships are undesirable due to relational deskilling, as AI will not possess important human features that keep us desirably social beings. The third argument claims that working toward AI friends is a technosolutionist idea, as we ought to provide social or no- or low-tech options to any solution that AI may offer. The fourth argument claims that we destroy unique value by replacing human–human friendships with human–AI ones, rendering the latter undesirable. This article argues that all four arguments are inconclusive to opposing AI becoming our friend. In contrast, there are several important positive points to consider when discussing the desirability of social AI and their capabilities for friendship. This leads to an optimistic view on the future of humans and AI becoming friends in a philosophically interesting sense.

Authors 1

1. Hendrik Kempt corresponding Aachen

   RWTH Aachen University

   Affiliation as printed

   RWTH Aachen University