Why AI Desire and Reasoning Should Not Be Taken Literally

Summary

In a discussion of AI, desire and fictional characters, philosopher Rebecca Lowe argues that psychological states such as desire belong only to living things. On this view, machines, fictional characters and dead historical figures do not literally have desires or goals, even when ordinary language attributes them to such entities. Lowe accepts that language can use shorthand, as when people say that a government “wants” something, but warns that similar claims about AI may lack an equally clear shared interpretation. She says this confusion could affect how people understand human rights and responsibilities, and could be especially harmful to children’s grasp of what reasoning is. At the same time, she is enthusiastic about AI’s potential to improve access to information and help address medical problems. She distinguishes using AI to assess evidence from asking it to predict whether someone will commit a future crime: patterns may support predictions, but they do not establish truths about the future. Lowe also allows that algorithmic outputs can resemble human choices in a functional sense, while maintaining that this does not mean an AI forms desires or experiences. Her discussion of Sherlock Holmes illustrates the limits of inferring a character’s psychology from a fictional text, where an author can later change the pattern. She closes by separating broad improvements in living conditions from the very different experiences of individuals, and reiterates her opposition to treating desirable outcomes as an obligation to maximize them.