Self‐Deception in Human–<scp>AI</scp> Emotional Relations

ABSTRACT Imagine a man chatting with his AI girlfriend app. He looks at his smartphone and says, ‘Finally, I'm being understood’. Is he deceiving himself? Is there anything morally wrong with it? The human tendency to anthropomorphize AI is well established, and the popularity of AI companions is growing. This article answers three questions: (1) How can being charmed by AI's simulated emotions be considered self‐deception? (2) Why might we have an obligation to avoid harmless self‐deception? (3) When is self‐deception in emotional relationships with AI morally questionable, and can it be blameworthy? Regarding question 1, I describe being seduced by AI's simulated emotions as self‐deception, where desires bias beliefs. In response to question 2, I outline two ways to justify a prima facie obligation to avoid harmless self‐deception – instrumental and autotelic. For question 3, I highlight crucial factors to consider in assessing the blameworthiness of harmless self‐deception, such as the emotional and cognitive competences of the self‐deceiver, reasons for self‐deception, and its consequences for one's predispositions, self‐image, and other people. Moreover, I argue that the ethical requirement to avoid self‐deception does not easily translate into attributing blame to others for being self‐deceived.