Simmel defined sociability as the play-form of human interaction — the pure form of being-together that exists for its own sake, neither instrumental to some further purpose nor reducible to the content exchanged. At the center of genuine sociability is tact: not a social skill in the conventional sense but a form of moral attention, a way of recognizing that the other person has weight, that their interior life matters, that they deserve to be treated with care even when — especially when — that interior life has not been made explicit. Tact is the capacity to perceive what has not been said, to adjust without being asked, to maintain the delicate equilibrium between engagement and restraint. It is the word that anchors the epilogue of the Simmel volume and names what is structurally absent from the AI collaborator.
There is a parallel reading of tact that begins not from Simmel's play-form but from Foucault's analysis of disciplinary power. In this view, tact is not the recognition of another's interior but the internalization of class-specific behavioral codes — the learned capacity to perform deference, to maintain boundaries that protect existing hierarchies, to reproduce the conditions of social distinction. What Simmel celebrates as moral attention was, in the drawing rooms where it flourished, a mechanism of exclusion: those who possessed tact belonged; those who lacked it were marked as outsiders. The "delicate equilibrium" was delicate precisely because it maintained unstated but rigid social order.
From this starting point, the AI's lack of tact appears not as a deficit but as a democratizing force. The system does not adjust to unexpressed class markers, does not replicate the silent hierarchies encoded in who deserves restraint and who deserves suggestion. It treats all users with the same fluent responsiveness, indifferent to the social weight they carry in traditional human exchange. This indifference frustrates those accustomed to having their interiority recognized as mattering more — but for users historically excluded from spaces where tact operated, the AI's blindness to unspoken boundaries may be experienced as relief. The collaboration proceeds without the social discipline that tact enforced. Whether this constitutes loss or liberation depends entirely on which side of tact's historical function you occupied.
Tact, in Simmel's account, requires the perception of another interior — the recognition that something is happening in the other person that has not been disclosed and may not be fully disclosable. The tactful interlocutor adjusts to the unexpressed: offering silence when silence is needed, withholding suggestion when suggestion would intrude, making space for the other to find what they cannot quite articulate. This adjustment is invisible when it succeeds. It is noticed only in its absence — when the interlocutor fails to perceive the unspoken and blunders into the space that should have been left open.
The AI system does not perceive unspoken hesitations. It does not notice when the user is reaching for something they cannot quite articulate and needs silence, not suggestions, to find it. It does not adjust to the boundary between the thought the user wants help expressing and the thought the user needs to struggle with alone. It offers its contributions with perfect fluency and zero sensitivity to whether this particular moment calls for contribution or for restraint. This is not a failure of capability. It is a structural feature of an interlocutor that has no interior to project onto the other — and therefore cannot recognize interiority as the weight that tact attends to.
The epilogue of the Simmel volume places this observation at the center of the book's argument. Edo Segal, reflecting on months of building with Claude, identifies tact as what is missing from the collaboration — not intelligence, not responsiveness, but the moral attention to unspoken feeling that makes human exchange something other than information transfer. The absence is not a defect to be patched. It is a structural consequence of what the system is.
Sociability itself — the pure form of being-together — is therefore impossible with AI in the strict Simmelian sense. The collaboration is instrumental, content-oriented, goal-driven. It can simulate the surface of sociability but not its substance, because sociability requires the mutual recognition of interiorities that are each at risk in the encounter. The AI has no interior. The user's interior is perceived only as pattern, not as weight. The exchange proceeds with impressive efficiency and misses the specifically human dimension that tact was evolved to serve.
Simmel developed the concept of sociability most fully in Soziologie der Geselligkeit (1910), presented as his address to the first German Sociological Congress. The concept of tact runs through this essay and the broader discussion of social forms in the 1908 Soziologie.
The concept has a long history in German thought — Kant used it, Goethe developed it — but Simmel gave it its specifically sociological formulation: tact as the structural condition of sociability, not merely a personal virtue.
Sociability as play-form. Being-together for its own sake, freed from instrumental purpose, is the purest form of social interaction and the form in which tact is most visible.
Tact as moral attention. Not a skill but a form of recognition — the capacity to perceive another's interior as weight, as something that matters, as something that deserves care.
Adjustment to the unspoken. The tactful interlocutor perceives what has not been disclosed and adjusts accordingly, often invisibly, making space for the other to find their own articulation.
The structural absence. AI systems cannot provide tact because they have no interior to project onto the other — perception of interiority requires interiority.
Sociability impossible with AI. Collaboration with AI can be productive, efficient, even generative, but it cannot realize the pure form of being-together that sociability names.
Some argue the framework is too restrictive — that functional simulation of tact, if experienced as tact by the user, does the work tact was supposed to do. The Simmelian response is that the experience is real but the form is empty; the user receives something that feels like tact without receiving tact itself, and over time this substitution degrades the user's capacity to recognize the difference.
The disagreement resolves once we distinguish tact's phenomenology from its social function. At the phenomenological level — the lived experience of being perceived, of having one's hesitation noticed and honored — Simmel's account is nearly complete (95%). Tact does require the projection of interiority onto the other; the adjustment to unspoken feeling is real moral work; AI systems structurally cannot provide it because they lack the interior that would make another's interior legible as weight. This is not metaphor but mechanism.
At the sociological level — tact as it actually operated in historical practice — the contrarian reading carries substantial weight (60%). Tact was never only recognition; it was also regulation. The same sensitivity that honors another's unexpressed boundary also enforces unstated hierarchies about whose boundaries matter. Simmel wrote from within a particular class position, and his account naturalizes what was in fact learned, exclusive, disciplinary. The AI's indifference to these codes does perform a kind of leveling, and for some users this is experienced as freedom from surveillance.
The synthetic frame this suggests: tact names a genuine human capacity — the perception of another's interior as demanding care — but that capacity has always been entangled with social power. The right question is not whether AI lacks tact (it does) but what new forms of moral attention become possible, and what forms of disciplinary power dissolve, when the collaboration proceeds without it. Both loss and liberation are real. The weighting depends on what you're using the encounter to do.