Propositional and Prescriptive Knowledge

In the AI Story

The distinction was developed most fully in The Gifts of Athena (2002), where Mokyr argued that the economic history of the modern world lives in the distance between these two kinds of knowledge. Propositional knowledge can be codified, printed, and distributed relatively cheaply. Prescriptive knowledge, by contrast, has historically resisted codification — it lived in the embodied skill of practitioners, transmitted through apprenticeship at the speed of physical demonstration rather than the speed of print.

Before the Industrial Enlightenment, the cost of conversion was enormous. A natural philosopher might understand in propositional terms why a particular chemical reaction produced a useful result. But transmitting that understanding to the craftsman who could apply it required a chain of translations — mathematical formulation to verbal description to practical demonstration to embodied skill — each introducing noise, delay, and loss. The Industrial Enlightenment succeeded by radically reducing conversion costs.

The distinction maps onto the AI transition with unusual precision. Large language models did not create new propositional knowledge. The training data — the vast corpus of human text, code, technical documentation — existed before the model was trained on it. What the model did was reduce the cost of converting propositional knowledge into prescriptive knowledge to near zero for a significant class of problems. The developer in Lagos, possessing extensive propositional knowledge about what users need, could now convert that understanding into working software through natural language rather than through years of formal training in specific programming languages.

The non-uniformity of the cost reduction matters enormously. Conversion costs have dropped most dramatically for problems specifiable in natural language — software development, document drafting, data analysis. They have dropped less for problems requiring tacit knowledge — the surgeon's feel for tissue, the therapist's reading of unspoken distress. And they have dropped hardly at all for what Mokyr calls epistemic base expansion — the generation of genuinely new propositional knowledge through experiment and inquiry.

Origin

Mokyr adopted Ryle's 1949 philosophical distinction (from The Concept of Mind) and gave it economic content. The 2002 formulation in The Gifts of Athena established the framework that the 2025 Nobel Committee cited explicitly when awarding Mokyr the Prize in Economic Sciences.

The Nobel Committee's Popular Science Background observed that 'Mokyr's work shows that AI could reinforce the feedback between propositional and prescriptive knowledge' — a recognition that the theoretical framework had found its most urgent contemporary application.

Key Ideas

Two kinds of useful knowledge. Propositional (knowing that) and prescriptive (knowing how) are different in kind, not merely in degree — with different cost structures for storage, transmission, and application.

Conversion cost as the variable. The economic history of the modern world is determined by the cost of converting propositional knowledge into prescriptive knowledge, not by the stock of either.

The reclassification revelation. The AI transition revealed that vast stores of knowledge treated as tacit were actually articulable — they had required a listener capable of interpreting ambiguous natural language and converting it into precise implementation.

The feedback loop accelerates. When conversion cost drops, more experiments run, more data generates, propositional knowledge expands, new prescriptive possibilities emerge, the loop self-reinforces.

Non-uniform cost reduction. Articulable prescriptive knowledge collapses in cost; tacit knowledge resists; new propositional knowledge generation remains expensive — producing a predictable distributional landscape of premium and displacement.

Debates & Critiques

Philosophers object that Mokyr's economic framing flattens Ryle's original ontological distinction. Practitioners of domains where tacit knowledge dominates — surgery, psychotherapy, elite craft — argue that Mokyr's framework understates the permanent irreducibility of embodied expertise. Mokyr's defenders respond that the framework correctly predicts which domains AI would most affect and which it would least affect.