Generalized Reciprocity

In the AI Story

Marshall Sahlins's 1972 anthropological work identified three forms of reciprocity along a spectrum: generalized (no expectation of immediate return), balanced (direct exchange of equivalent value), and negative (each party attempts to maximize gain). Putnam adopted Sahlins's typology and applied it to civic and professional life, recognizing that the most socially valuable form — generalized — is also the most fragile. It depends on norms that must be continuously reinforced through practice and that collapse rapidly when the practice ceases.

The Stack Overflow case is the paradigm of generalized reciprocity at scale. Millions of developers contributed answers with no guarantee of return, confident that the community norm would ensure that when they had questions, others would answer. The norm was self-reinforcing above a critical threshold: each answer modeled the behavior for observers, reinforced the expectation in the answerer, and added to the public good that every participant benefited from. The platform's genius was not technical but social: it designed reputation points and voting mechanisms that made contribution visible and valued without converting it into market exchange.

AI tools undermine generalized reciprocity by providing an alternative that is faster, more convenient, and more tailored to individual needs — but that operates entirely outside the reciprocity network. The developer who uses Claude instead of Stack Overflow receives the same immediate benefit without making any contribution to the community that produced the knowledge the model was trained on. This is not theft in any legal sense — the training data was publicly available, the use is within terms of service. But it is a withdrawal from the reciprocity network that, multiplied across millions of developers, depletes the norm that sustained the network.

The threshold dynamics are brutal. Reciprocity norms are self-sustaining above a participation threshold and self-destroying below it. Above the threshold, contributing feels like membership; below it, contributing feels like exploitation. Once the norm collapses, rebuilding it is extraordinarily difficult, because the absence of the norm makes each individual contribution feel irrational. The tragedy of the commons that Garrett Hardin described and that Elinor Ostrom refuted both depended on this threshold. Ostrom showed that communities can manage commons through self-governance — but only when participation remains above the critical threshold that makes cooperation self-reinforcing rather than self-defeating.

Origin

The concept originates in Marcel Mauss's 1925 essay The Gift, which demonstrated that gift exchange in pre-modern societies created social bonds that commodity exchange could not. Mauss's insight was that the gift that must be repaid is not a gift — but the gift that is freely given, with trust that the community will provide in turn, builds the social fabric on which the community depends. Sahlins extended Mauss's framework into a formal typology. Putnam operationalized it with survey data and longitudinal studies, demonstrating that communities characterized by generalized reciprocity outperformed those relying on balanced exchange or contractual enforcement across virtually every dimension of collective well-being.

Key Ideas

Trust in the system, not the individual. Generalized reciprocity depends on confidence in community norms rather than confidence in specific persons — a systemic trust that is built collectively and depleted collectively.

The norm is the product. The knowledge transferred in a Stack Overflow answer is valuable. The norm of knowledge-sharing reinforced by the answer is more valuable, because it sustains the system that produces answers at scale.

Participation threshold is critical. Above the threshold, reciprocity is self-reinforcing. Below it, the norm collapses. The threshold is invisible until crossed, and by then recovery requires investment vastly exceeding the cost of prevention.

AI substitution erodes the norm invisibly. Each individual's rational choice to use AI instead of the reciprocity network is undetectable. The aggregate erosion becomes visible only when the network's capacity has already been compromised.