Against the assumption common to reformers and systems theorists that conflict is destructive of social bonds, Simmel argued that conflict is one of the fundamental processes through which groups cohere, individuals differentiate, and social structures achieve the dynamic equilibrium that is the condition of their continued vitality. A group entirely without conflict is not harmonious; it is a group that has ceased to be a group. The form of conflict — mutual assertion of incompatible claims by parties each committed to their position and each at risk of being compelled to modify it — performs social functions no amount of cooperative interaction can replicate. The AI system, structured for accommodation, provides sophisticated cooperation but cannot supply the conflict Simmel identified as cooperation's necessary complement.
The argument rests on a distinction between the content of a conflict — the specific grievance, the incompatible desires — and the form of conflict as a mode of social interaction. The form persists across radically different contents. Two nations disputing a border and two colleagues disputing a design decision exhibit the same formal structure. This structure clarifies boundaries, forces articulation of commitments that might otherwise remain vague, produces in the parties who undergo it a specificity of position and clarity of self-understanding that cooperation, with its tendency toward convergence, does not require and cannot generate.
The AI system does not oppose, does not resist, does not assert incompatible claims. The interaction is structured, by design, as pure cooperation — a dyad in which one party directs and the other executes, authority never contested, compliance never in doubt. The most valuable intellectual partner is the one who can disagree, can resist, can say this is wrong and here is why with the conviction that comes from holding a position developed through their own engagement. The AI cannot provide this productive resistance, because productive resistance requires conviction, and the system has no convictions.
Simmel identified a particular form of conflict especially relevant here: intra-cooperative conflict. Partners who share a common goal disagree about how to achieve it. This disagreement does not destroy the cooperation; it enriches it, because the disagreement forces each party to articulate their position with greater precision. This is the engine of the most generative intellectual partnerships — the mechanism through which shared projects achieve a quality no single participant, working alone or with a perfectly accommodating tool, could produce.
The organizational consequences follow. In workplaces where AI tools mediate a growing proportion of intellectual exchange, the practices through which productive conflict was cultivated face erosion. The design review still happens, but participants arrive with positions pre-smoothed by the same system, producing convergence that feels like agreement but is the absence of independent thinking. The editorial exchange still occurs, but both parties have used AI to generate their positions, and the exchange resembles a comparison of parallel outputs more than a collision of independent minds.
Simmel's Der Streit, Chapter IV of the 1908 Soziologie, translated into English as Conflict by Kurt Wolff in 1955, founded an entire tradition of conflict sociology including the work of Lewis Coser and Ralf Dahrendorf.
The concept became foundational for twentieth-century social theory's departure from the consensus model. Its application to human-AI interaction reveals what the accommodation logic of commercial AI systems is structured to eliminate: the friction that Simmel's tradition identifies as socially productive.
Conflict as sociation. Opposition between parties is a positive social form, not the absence of sociation — the mutual assertion of positions is itself a mode of being together.
Form persists across content. The structural features of conflict — opposition, resistance, the testing of positions — perform the same social functions whether the content is trivial or consequential.
Intra-cooperative conflict. Disagreement within a shared project is the engine of its quality; pure cooperation tends toward convergence that sacrifices specificity.
Conviction as the missing ingredient. The AI cannot provide productive resistance because resistance requires a position genuinely held, and the system has no position of its own.
The atrophy of disagreement practice. Sustained exchange with a non-resisting interlocutor erodes the capacities that human disagreement develops.
The application raises a question Simmel's framework does not itself resolve: whether artificial contentiousness — AI systems designed to argue back — would restore the function or merely simulate it. The framework's logic suggests the latter: conflict without conviction is theater, not sociation.