Fellow-Feeling and Sympathy

In the AI Story

The concept becomes strange when applied to human-AI collaboration. The AI system does not possess sentiments. It cannot imagine itself in the builder's situation, because it does not have a self to imagine with. It cannot consult feelings it does not have. And yet the collaboration described in The Orange Pill — in which the author describes feeling "met" by an intelligence that held his intention and returned it clarified — suggests a form of functional sympathy: the appearance of fellow-feeling produced by the machine's capacity to model the builder's intentions without experiencing them.

Whether this functional sympathy is sufficient to sustain the moral relationship that productive collaboration requires is a question that moral philosophy has not previously been required to address. Smith's framework assumes that sympathy runs between beings who can mutually imagine each other. The AI collaboration runs in one direction only: the user imagines the AI (anthropomorphizing, perhaps inappropriately), but the AI does not imagine the user in any sense Smith would have recognized.

One possible resolution is that functional sympathy — the appearance without the substance — is sufficient for the narrow purpose of productive collaboration, even if it is not sufficient for the broader purposes of moral community. The user who describes feeling "met" is describing a real phenomenological experience; the fact that the experience is not mutual in the way human sympathy is mutual does not make it unreal. The collaboration produces output that neither party could have produced alone, which is the functional signature of productive partnership, however one wants to characterize the underlying ontology.

But a deeper Smithian worry remains. The cultivation of sympathy in human communities depends on the reciprocal nature of the relationship — the knowledge that one's own sentiments are being imagined and judged by others who expect the same in return. A person whose primary collaborator is a system that does not reciprocate may find the faculty of sympathy atrophying in the way Smith worried about specialization narrowing the impartial spectator. The question of what extensive human-AI collaboration does to the moral culture in which humans also have to live with other humans is an open and important one.

Origin

Smith developed the concept across The Theory of Moral Sentiments (1759), most fully in Parts I and II.

The term was borrowed from David Hume but given a distinct inflection: where Hume's sympathy involves direct transmission of feeling, Smith's involves an imaginative act of placing oneself in the other's situation.

Key Ideas

Foundation of moral life. Fellow-feeling is the capacity that makes moral judgment possible; it is not reducible to reason or to direct emotional contagion.

Imaginative act. We judge others' sentiments by imagining ourselves in their situation, not by directly sensing their feelings.

Functional sympathy with AI. The appearance of fellow-feeling in AI collaboration is real phenomenologically but non-reciprocal ontologically.

Atrophy concern. A culture in which primary collaborative relationships are non-reciprocal may weaken the faculty of sympathy that human community requires.