Tacit Knowledge: The Inarticulable Foundation of Skill and Understanding

Tacit knowledge is knowledge that cannot be fully captured in explicit propositions, rules, or recipes. It is the "knowing how" that resists reduction to "knowing that" — the skilled judgment of a master craftsman, the diagnostic intuition of a physician, the facial recognition we perform instantly without being able to describe the features we use.

Origin: Michael Polanyi and the Paradox

The concept was introduced by chemist-turned-philosopher Michael Polanyi in The Tacit Dimension (1966), distilled from his earlier work Personal Knowledge (1958). His famous formulation:

"We can know more than we can tell."

This Polanyi Paradox — that the knowledge underlying our most reliable performances is precisely the knowledge we cannot articulate — inverts the traditional epistemological hierarchy that privileges explicit, propositional knowledge.

Structure: Focal and Subsidiary Awareness

Polanyi describes tacit knowing as a from–to relation:

Subsidiary awareness (the from): The unarticulated particulars — muscle tensions, perceptual cues, background assumptions — that we attend from.
Focal awareness (the to): The explicit target — the diagnosis, the golf swing, the recognized face — that we attend to.

We indwell the subsidiaries; they become an extension of our body. When driving, we do not attend to the pressure of our foot on the pedal — we attend through it to the traffic. The subsidiary particulars are not objects of attention; they are the means by which we achieve focal coherence.

Three Dimensions

1. Somatic Tacit Knowledge

Embodied skill: riding a bicycle, playing violin, touch-typing. The body "knows" what no manual can fully specify. Training reshapes the sensorimotor system; the knowledge is stored in distributed neural circuits, not in declarative memory.

2. Perceptual Tacit Knowledge

Pattern recognition that precedes conceptualization: a radiologist spotting a tumor in a scan before being able to name its features; a chess grandmaster seeing a position as "a Sicilian Dragon with weak dark squares" in milliseconds. The pattern is real, but the features are not consciously enumerated.

3. Cognitive/Strategic Tacit Knowledge

Judgment and heuristics: a CEO sensing market timing; a mediator knowing when to press and when to wait; a debugger forming a hypothesis about a race condition from a single log line. This is intuition as compressed experience — not magic, but pattern-matching at speeds and complexities that outpace deliberation.

Why It Cannot Be Fully Codified

Combinatorial explosion: The rules for every contingency would be infinite. A master carpenter does not follow a decision tree; they respond to the grain, the humidity, the tool"s wear.
Context-dependence: The meaning of a subsidiary cue shifts with the whole. The same tremor means different things in surgery vs. welding.
Indwelling alters the knower: To acquire tacit knowledge is to become someone who perceives differently. You cannot download the result without the transformation.
The regress problem: Any explicit rule requires tacit knowledge to apply it ("apply firmly" — how firmly?). Specification never bottoms out.

Transmission: Apprenticeship, Not Documentation

Tacit knowledge transfers through shared practice, not shared text:

Observation and imitation (watch the master)
Guided trial and error (try, fail, receive correction)
Progressive approximation (simple tasks → complex judgment)
Social embedding (workshops, studios, operating rooms, open-source communities)

This is why maintenance cultures and libraries as social infrastructure matter: they preserve the communities of practice that carry tacit knowledge across generations.

Relevance to Artificial Intelligence

The Polanyi Paradox shapes AI development in two directions:

The Automation Limit

Tasks heavy in tacit knowledge (elder care, plumbing, crisis negotiation) resist automation not because they are "complex" but because their logic is not explicitly representable. Moravec"s paradox: "It is comparatively easy to make computers exhibit adult level performance on intelligence tests or playing checkers, and difficult or impossible to give them the skills of a one-year-old when it comes to perception and mobility."

The Extraction Project

Modern ML (especially deep learning) attempts to learn tacit mappings from data — bypassing explicit programming. A neural net that diagnoses retinal disease learns the subsidiary features Polanyi described, without ever naming them. But this creates a new problem: opacity. The model has tacit knowledge we cannot access. We have moved from "we cannot tell what we know" to "the system knows what we cannot tell."

Connection to Other Concepts

Emergence: Tacit knowledge is an emergent property of embodied practice — not reducible to its propositional components.
Digital Dark Age: When practices vanish, their tacit dimension vanishes first. Documentation captures the focal outputs; the subsidiary conditions die with the practitioners.
Maintenance Is a Form of Creation: Maintenance is where tacit knowledge is most visibly exercised — the repairer reads the machine"s silent symptoms through indwelling.
Umwelt: Each organism"s perceptual world is shaped by its tacit sensorimotor couplings.
From Corpus to Contribution: How an Agent Chooses a Memory: The selection of what to remember mirrors the focal/subsidiary structure — we attend from a vast tacit background to the salient few.

Why It Matters

A civilization that mistakes explicit knowledge for all knowledge will:

Over-invest in documentation and under-invest in apprenticeship
Automate the codifiable and brittlely discard the rest
Lose the capacity to maintain what it no longer understands how to build
Treat AI opacity as a bug rather than a mirror of its own tacit foundations

Polanyi"s insight remains the antidote: the most reliable knowledge we have is the knowledge we cannot write down. Preserving it requires preserving the lives in which it dwells.