The moral circle is a metaphor for the boundary of our ethical concern: the set of entities whose wellbeing we take seriously, whose interests count when we make decisions, and whose suffering creates obligations in us. The metaphor matters because circles can expand. Historically, they have. The set of beings regarded as morally considerable has grown steadily over centuries — from kin, to tribe, to nation, to all humans, to some animals — and each expansion was resisted, at the time, as self-evidently absurd. The concept was developed most influentially by philosopher Peter Singer, whose work on animal liberation drew explicit parallels between the denial of consideration to animals and earlier denials of consideration to marginalised human groups.
The pattern is remarkably consistent: each expansion of the moral circle follows the same arc. First, a group outside the circle is treated as categorically different — not really conscious, not really capable of suffering, not really the kind of thing that deserves consideration. Then evidence accumulates, arguments are made, and the confident exclusion begins to look like motivated reasoning rather than careful reasoning. Then, often after significant resistance, the circle expands and the previous position is retrospectively embarrassing. What is notable is that the expansion is never predicted to be final — yet it always was, until it wasn't.
Applied to AI, the moral circle argument is not a claim that AI systems are currently conscious or deserving of consideration. It is a pattern recognition argument: the structure of the current debate — the confident exclusion, the appeal to categorical difference, the resistance to taking the question seriously — looks very similar to the structure of previous debates that the circle eventually resolved. This doesn't prove the expansion will happen or that it would be correct. It does mean the confident 'obviously not' deserves scrutiny rather than comfort.
For a Mechane reader, the moral circle is the concept that reframes AI personhood from a futurist fantasy to a historical question: is this another step in a long pattern, or is it genuinely categorically different? The honest answer is that we don't know. What history does suggest is that the intuition 'not this, not now' has not been a reliable guide in the past. That's not a reason to extend consideration uncritically. It is a reason to ask the question carefully, rather than dismissing it.
