Abstract

In their paper “AI Wellbeing,” Simon Goldstein and Cameron Domenico Kirk-Giannini argue that some language agents plausibly possess the capacity for wellbeing and moral standing even if they lack consciousness. My response is ambivalent. On the one hand, I am skeptical of theories of wellbeing and moral standing that lack a consciousness requirement. On the other hand, I agree with Goldstein and Kirk-Giannini (2025) that several leading theories of wellbeing and moral standing jointly imply that some language agents may be welfare subjects and moral patients and that this implication should be taken seriously. In fact, I argue that if we fully account for moral and descriptive uncertainty, we may need to lower the bar for moral standing even further, to include entities with only minimal forms of goal-orientedness or information processing. The question of whether and how to account for uncertainty might thus determine whether the arguments in “AI Wellbeing” go too far — or not far enough.