The Intelligence Commons
The intelligence commons is not the AI tool itself, which is a private good fully excludable through pricing. It is the larger ecology in which the tool operates: the shared body of human knowledge from which training data is drawn, the distributed pool of professional expertise that sustains quality assessment, the finite space of human evaluative attention, the accumulated trust in human-generated information, and the institutional arrangements through which the community manages its relationship to the other four flows. Each exhibits the defining characteristics of a common-pool resource — subtractability and difficulty of exclusion — though in forms distinct from natural-resource commons.
The Enclosure Already Happened — Contrarian ^ Opus
There is a parallel reading that begins not from resource degradation but from primitive accumulation. The intelligence commons Ostrom-style analysis describes is already enclosed — the historical moment when community self-governance might have structured AI development passed somewhere between GPT-2's open release and GPT-3's API lockdown. What appears as a tragedy of the commons is better understood as the final stage of enclosure: the conversion of common resources into private assets through extraction that was always structural, not accidental.
The five flows are not symmetrically vulnerable. Knowledge and skills were enclosed first — training data scraped without consent, human expertise embedded in model weights owned by private firms. Attention and trust are degrading now as the natural consequence of that prior enclosure, not as failures of community governance but as the working-out of property relations already established. The institutional commons never existed in AI; what governance arrangements we have were designed by the firms who completed the enclosure. Applying Ostrom's framework to this landscape mistakes aftermath for commons management. The builders operating in this space are not commoners deliberating about shared resources; they are participants in a market whose infrastructure layer is privately owned. Community self-governance cannot emerge when the territory governance would address is already privatized. The analytical move is not to extend Ostrom but to read her alongside historians of enclosure who documented how commons became impossible.
In the AI Story
The framing shifts analytical attention from the AI product to the resource system in which AI operates. A subscription to a language model is a private good. The knowledge commons on which the model was trained, the skills commons that produces practitioners capable of evaluating its output, the attention commons in which its output competes for evaluation, and the trust commons that allows any human-AI exchange to function — these are shared resources vulnerable to degradation through the aggregate of individually rational decisions.
The recursive character of the intelligence commons distinguishes it from every natural-resource commons Ostrom studied. In a fishery, the resource units are produced by marine ecosystems that the users do not control. In the intelligence commons, the resource units are produced in significant part by the users themselves. The community does not merely extract from the commons; it constitutes the commons. A degraded commons produces a degraded community, which further degrades the commons. A well-managed commons produces a thriving community, which further enriches the commons.
The feedback loops between the five flows amplify individual dilemmas into systemic crisis. The degradation of the skills commons reduces the community's capacity to evaluate quality, which accelerates the degradation of the attention commons. The erosion of the trust commons undermines the monitoring mechanisms that maintain standards, which further degrades the knowledge commons. The underinvestment in the institutional commons means governance arrangements needed to address cascading degradations are themselves inadequate.
This framing also dissolves the market-versus-state binary that dominates contemporary AI governance discourse. The governance challenges the intelligence commons generates are neither resolved by clearer property rights nor by centralized regulation alone; they require the third institutional possibility — community self-governance — that Ostrom's research documented as viable.
Origin
The concept extends Ostrom's framework to AI by applying the analytical test she established: any resource system exhibiting subtractability and difficulty of exclusion is a common-pool resource, regardless of its physical substrate. The five-flow decomposition — knowledge, skills, attention, trust, institutions — emerges from applying that test systematically to the ecology of AI-augmented work.
Key Ideas
Five interlocking flows. Knowledge, skills, attention, trust, and institutions each function as common-pool resources with distinct subtractability dynamics.
Recursive character. The community does not merely extract from the commons; it constitutes the commons, creating feedback loops unknown to natural-resource commons.
Invisible degradation. Each flow can erode silently, masked by the surface quality of AI-generated output that conceals failures of judgment beneath fluent execution.
Beyond market and state. The governance challenges require the third institutional possibility — community self-governance — that the dominant binary forecloses.
Appears in the Orange Pill Cycle
Phased Enclosure, Residual Commons — Arbitrator ^ Opus
The most productive synthesis recognizes the intelligence commons as partially enclosed terrain with meaningfully distinct phases. The knowledge commons exhibits 70% enclosure / 30% residual commons — training data is largely privatized but not entirely; open models and datasets preserve some common access. The skills commons remains 60% common / 40% enclosed — professional judgment is still substantially developed and shared through community mechanisms, though increasingly mediated by proprietary tools. This asymmetry matters because different flows face different governance possibilities.
The key question is temporal: are we observing degradation of an active commons (Ostrom's frame fully applies) or aftermath of completed enclosure (enclosure frame dominates)? The answer depends on which institutional scale you examine. At the foundation model layer, enclosure is 80% complete — meaningful community governance over training and architecture is foreclosed. At the application layer, commons dynamics are 70% active — how builders share techniques, evaluate quality, and maintain trust networks remains substantially open to community self-organization. The recursive character Edo identifies is real but operates differently across these scales.
The synthetic move is recognizing that partial enclosure creates hybrid governance requirements. Pure Ostrom (community manages intact commons) doesn't fit because enclosure is too advanced. Pure enclosure analysis (all commons lost) doesn't fit because residual commons dynamics remain active and consequential. The intelligence commons requires governance frameworks designed for resources that are simultaneously partially enclosed and partially common — a regime type Ostrom studied less but that her institutional analysis can illuminate.
Further reading
- Max Fang, "The Tragedy of the AI Data Commons" (Stanford working paper, 2025)
- Ostrom, Governing the Commons (1990)
- Charlotte Hess and Elinor Ostrom (eds.), Understanding Knowledge as a Commons (MIT Press, 2007)
- Mozilla Foundation and Ostrom Workshop, data commons governance framework