Where does knowledge collide?

A person knows that p when p is true, they believe it, and their belief is justified (the tripartite analysis usually traced to Plato's Theaetetus and standard until 1963). Gettier's three-page paper gave two cases in which all three conditions hold and nobody would say the person knows: the justification is real but connects to the truth by luck. Every account since (reliabilism, tracking, virtue epistemology, knowledge-first) is an attempt to add or replace a condition. Knowledge here is a relation between one mind and one proposition, and truth is a requirement, not a hope.[1][2]

Unit

A single agent's relation to a proposition

Recovery means

Not applicable

Measured by

Not measured; tested by intuition on cases

The second of the Model Penal Code's four levels of culpability: a person acts knowingly as to a circumstance when aware it exists, and as to a result when aware it is practically certain to follow. Section 2.02(7) adds that awareness of a high probability suffices unless the person actually believes otherwise, which is the statutory root of willful blindness. Global-Tech held in 2011 that a defendant who subjectively believes there is a high probability a fact exists and takes deliberate action to avoid learning it can be treated as knowing it, and the Court distinguished that standard from recklessness and negligence. Knowledge here is a threshold of awareness, truth of the underlying fact is assumed by the charge, and justification is beside the point.[3][4]

Unit

A defendant's awareness as to an element of an offence

Recovery means

Not applicable, the question is proof, not restoration

Measured by

Inference from conduct, tested against a jury instruction on knowledge and deliberate ignorance

Two founding claims that pull in different directions. Hayek argued in 1945 that the knowledge that matters for an economy is dispersed among millions of people as particulars of time and place, is never given to anyone in totality, and is coordinated only by prices. Arrow argued in 1962 that knowledge as an output of invention is indivisible, cheap to reproduce, and hard to appropriate, so markets underinvest in it, the case for patents and public science. The knowledge economy literature inherits Arrow's good and Hayek's diffusion. Knowledge here is a commodity with a price and a production function, and whether anyone believes it is irrelevant.[5][6]

Unit

A stock or flow of information with economic value

Recovery means

Not applicable

Measured by

R&D spending, patents, human capital proxies, total factor productivity residuals

What an organisation knows, held partly in documents and systems and partly in the heads and hands of its people. Nonaka fixed the field's central distinction in 1994, borrowing tacit versus explicit knowledge from Polanyi and proposing that organisations create knowledge by converting one into the other in a spiral of socialisation, externalisation, combination, and internalisation. Alavi and Leidner's 2001 review organised the resulting literature around processes (creation, storage, transfer, application) and the systems built to support each. Knowledge here belongs to a firm, is an asset to be managed, and is defined by its form rather than its truth; a widely shared falsehood is still organisational knowledge.[7][8]

Unit

An organisation's stock of tacit and explicit knowledge

Recovery means

Retention and transfer when people leave; the failure mode is knowledge loss

Measured by

Knowledge-sharing self-reports, repository use, knowledge-based view proxies

Knowledge as a stored representation, divided by kind. Anderson's ACT theory of 1983 separated declarative knowledge, facts held as propositions in a network, from procedural knowledge, skills held as production rules, and gave an account of how the first is compiled into the second by practice. Squire's 1992 synthesis grounded the split in the brain: amnesic patients with hippocampal damage lose the ability to form declarative memories but learn skills and priming normally, so the two are different systems, not different labels. Knowledge here is whatever the memory system retains, truth is not a criterion, and the interesting question is which system holds it.[9][10]

Unit

A representation in an individual's memory

Recovery means

Retrieval; dissociation between systems after brain injury

Measured by

Recall and recognition for declarative; speed and accuracy of performance for procedural

Facts and their structure encoded so that a machine can store, query, and reason over them. Gruber defined an ontology in 1993 as an explicit specification of a conceptualisation, a shared vocabulary of classes, relations, and constraints designed for portability between systems. Knowledge graphs, reviewed by Hogan and colleagues in 2021, represent that content as a graph of entities and typed edges and add schema, identity, context, and inference on top. Knowledge here is whatever the graph asserts; it is knowledge by construction, its quality is measured by coverage and consistency, and a wrong triple is an error in the data rather than a failure to know.[11][12]

Unit

A machine-readable set of assertions and their schema

Recovery means

Not applicable, the failure mode is incompleteness or inconsistency

Measured by

Triple counts, completeness against a gold standard, link-prediction accuracy