Bibliography &
Reference Shelf

The foundational text for the project. Polanyi's central claim — "we can know more than we can tell" — names the transmission problem more precisely than anything written before or since. The book distinguishes between focal and subsidiary awareness, arguing that all knowing involves a tacit dimension that cannot be fully made explicit. Essential for the Transmission Mechanism argument.

Ryle's distinction between "knowing that" (propositional knowledge) and "knowing how" (practical competence) is the philosophical grounding for why wisdom resists verbal transmission. Language can encode the first. It cannot encode the second. His critique of the "intellectualist legend" — the idea that intelligent action always follows from propositional knowledge — is directly relevant to the micro-scale problem.

Wittgenstein's arguments about rule-following and the limits of language as a transmission medium complement Polanyi and Ryle. His concept of "forms of life" — that meaning is embedded in shared practice, not just in words — supports the argument that wisdom requires lived participation, not merely verbal instruction. The private language argument implies that the most important knowing is irreducibly social and embodied.

The most influential business-facing treatment of the tacit/explicit distinction. Nonaka and Takeuchi's SECI model (Socialisation, Externalisation, Combination, Internalisation) attempts to describe how tacit knowledge circulates inside organisations. Useful for the business moats application — and also as a foil, since the model arguably overstates how much tacit knowledge can be externalised and recaptured.

Henrich's core argument: human intelligence is a collective property, not an individual one. The evidence that isolated populations lose critical technologies within generations — the Tasmania case is the most striking — directly supports the Forgetting Machine thesis. "The collective is the genius; the individual is the vessel." Also documents how tacit knowledge embedded in cultural practices (food preparation, tool use) is lost catastrophically when transmission chains break.

Tomasello's account of the "ratchet effect" — how human culture accumulates by building on previous generations' achievements — is the mechanism underlying the knowledge-compounding side of the central asymmetry. His comparison with other primates who cannot sustain cumulative culture clarifies what is distinctively human about the problem.

The foundational academic text on cultural evolution and gene-culture co-evolution. Boyd and Richerson model how cultural transmission works at the population level — and where it fails. Background scaffolding for the anthropology section; less likely to be cited directly in accessible writing but important for conceptual rigour.

The original empirical documentation that scientific publications double every 10–15 years — the quantitative backbone of the macro-scale argument. de Solla Price was the first to make the exponential growth of collective knowledge visible as a structural fact with consequences, not merely a sign of progress. His warnings about where this leads were largely unheeded. Core reference for the civilisational scale section.

Durkheim's foundational argument that complex societies distribute knowledge and function across members the way a body distributes function across organs. His concept of organic solidarity — where interdependence replaces shared beliefs as the social glue — maps directly onto the macro-scale fragmentation problem. Also introduces the idea that society constitutes a higher-order intelligence that transcends any individual mind.

Snow's Rede Lecture named the growing gulf between scientific and humanistic knowledge — a concrete early instance of what the project calls the Specialisation Trap. The inability of specialists in each domain to talk across the divide is an early documented symptom of the widening gap. Still widely cited; useful as an accessible historical anchor for the specialisation argument.

Simon's most accessible treatment of bounded rationality — the argument that human decision-making takes place within hard cognitive limits, producing "satisficing" rather than optimising behaviour. As the information environment grows more complex, bounded rationality becomes a more severe constraint. Also relevant for the specialisation trap: experts are bounded rational agents who can only master a narrow slice.

The landmark paper establishing working memory capacity as approximately seven items — the biological floor beneath all the project's arguments. This constraint has not changed in 50,000 years. It is the hardest of the hard limits: not a cultural or educational failure but an architectural fact about the thinking organ. Useful as a concrete anchor when discussing why cognitive capacity cannot scale with knowledge.

The accessible synthesis of decades of cognitive bias research. Kahneman's System 1/System 2 framework maps interestingly onto tacit vs explicit knowledge — intuitive, fast, embodied knowing vs deliberate, slow, propositional reasoning. Also relevant for the business application: what we call "good judgment" often involves System 1 expertise that is extremely difficult to audit or transmit.

Piaget's stage theory of cognitive development establishes that the ~20-year bootstrapping timeline is not merely cultural or educational but reflects genuine developmental stages that cannot be fast-forwarded. Children are not small adults processing less information — they have qualitatively different cognitive structures that must unfold in sequence. This supports the argument that the bootstrapping cost is biological and irreducible.

Draws a productive analogy between human tacit knowledge and AI "hidden knowledge" — the weights in deep learning models that produce outputs nobody can fully explain. Key contribution: AI hidden knowledge is frictionlessly transferable across model instances, unlike human tacit knowledge which is ineradicably personal. Useful as a foil: the distinction between AI performing tacit expertise and AI transmitting tacit expertise is exactly where the Forgetting Machine critique of AI optimism must be made.

Anthropic's interpretability research — cited by Grant — attempts to probe what is actually happening inside large language models. Relevant as evidence that AI hidden knowledge is not fully analogous to human tacit knowledge: researchers can partially interrogate model weights in ways that have no equivalent for human expertise. But the researchers themselves acknowledge they can only capture a fraction of what is happening. Useful in the AI section for nuancing the optimist/sceptic debate.

The foundational strategy paper arguing that tacit knowledge is the primary source of sustainable competitive advantage — precisely because it cannot be easily imitated or transferred. Directly relevant to the business moats application: organisations built on embedded tacit knowledge are structurally harder to replicate than those built on explicit processes. Gives academic grounding to the commercial case for tacit knowledge investment.

Collins updates and complicates Polanyi — usefully for the project — by distinguishing three types of tacit knowledge: relational (could be made explicit but isn't), somatic (embodied in the body, harder to make explicit), and collective (embedded in social groups and practices, the hardest of all). The collective category is particularly relevant: some tacit knowledge doesn't live in any individual but in the group's shared practice. When that group disperses, the knowledge is gone.