Ownership of the Means of Thinking

In the [YOU] on AI Field Guide

The ownership of the means of thinking is the political dimension that underlies every other concern the cycle raises. The [YOU] on AI framework celebrates the democratization of building—the developer in Lagos who can now access execution leverage comparable to the developer in San Francisco—while simultaneously documenting the concentration of the infrastructure that makes this access possible in a small number of firms whose interests are not identical to the user's. The concentration is structural, not conspiratorial: training data, compute, and the interfaces through which AI operates require capital that only a handful of entities can supply, and the suppliers naturally shape the tools in ways that serve their purposes.

Crawford's analysis intersects with Porter's five forces framework at the supplier-power node: the AI platform providers possess the concentrated supplier power that Porter identified as the most dangerous structural feature for downstream firms to face. A firm that builds its entire workflow around a single AI provider's tools is strategically exposed in ways that Crawford adds a political dimension to: the exposure is not merely commercial but epistemic. The tool shapes the thinking, and the thinker cannot fully audit the shaping.

The democratic concern is not theoretical. Crawford's Senate testimony demonstrated what the means-of-thinking framework implies for governance: when decisions about credit, hiring, medical treatment, and information access are made by processes that cannot give an account of themselves, the principle of democratic accountability—that power must be expressible in terms the governed can scrutinize—is violated at scale. The incorruptibility of the standard that the motorcycle provides is what democratic accountability requires; algorithmic governance provides its opposite.

Origin

The essay “Ownership of the Means of Thinking” was published in December 2025, shortly before Crawford joined the American Enterprise Institute's AI Ethics Council at its launch in February 2026. The deliberately Marxist framing was chosen, Crawford explained, not to endorse a particular political program but to use the most precise vocabulary available for describing the relationship between productive infrastructure and the distribution of cognitive power. The means of production did not determine consciousness mechanically; they shaped the conditions within which thinking occurred. The means of thinking shape the conditions within which cognition can be exercised, which kinds of problems can be formulated, which solutions can be evaluated, and which forms of understanding can be transmitted across generations.

The concept is also Crawford's response to the AI discourse's tendency to frame the transition as a technical problem with technical solutions: better models, better alignment, better regulation of outputs. Crawford's argument is that the problem is political-economic at its foundation, requiring the kind of institutional analysis that addressed the concentration of industrial means of production: accountability mechanisms, distributed ownership structures, public infrastructure that is not beholden to private interests, and educational systems that maintain the capacity for independent cognitive evaluation rather than producing expert prompters of proprietary tools.

Key Ideas

Cognitive Infrastructure as Political Economy. The means of thinking are not a metaphor for individual cognition but a description of the infrastructure through which a society's collective intelligence operates: the training data that shapes what AI systems know and value, the computational capacity that determines who can run frontier models, and the interfaces that determine how humans interact with machine cognition. The ownership of this infrastructure is a political question, not merely a commercial one.

Opacity as Structural Feature. The opacity of large language models—the impossibility of reconstructing why a specific output was produced—is not a temporary engineering limitation but a structural feature of systems whose complexity exceeds the capacity of human understanding to reconstruct. This opacity is what Crawford's Senate testimony named as the foundation of a new priesthood: authority without accountability, power that need not give an account of itself.

The Democratic Stakes. Democratic self-governance requires that citizens retain the cognitive capacity to evaluate the accounts that power gives of itself. Outsourcing evaluation to tools whose operation citizens cannot audit is not merely a convenience; it is a structural erosion of the evaluative capacity on which democratic accountability depends. The discipline of the real—the formation that comes from submitting to a standard one did not set—is the cognitive foundation of that evaluative capacity.