Tacit Knowledge: Learning What Experts Cannot Tell You

Michael Polanyi began The Tacit Dimension (1966) "from the fact that we can know more than we can tell." A cyclist cannot state the rules that keep the bicycle upright. A senior engineer glances at a design and says it will cause trouble later, and is usually right. This page is about that kind of knowledge: how to recognise it, why writing it down mostly fails, and how to acquire it anyway.

What tacit knowledge is

Tacit knowledge is knowledge that cannot be passed on through words alone. Polanyi's account of its structure, in Meaning (with Harry Prosch, 1975), is that "tacit knowing is a from-to knowing." You attend from many small particulars to a whole. The rider attends from countless small corrections to staying upright, and could not list the corrections if asked.

Herbert Simon gave the working mechanism in one line, which Daniel Kahneman quotes in Thinking, Fast and Slow (2011): "Intuition is nothing more and nothing less than recognition." A situation provides a cue, the cue calls up stored experience, and the experience supplies the answer. An expert's fast judgement is recognition built from many cases, not magic and not a rule being consulted quickly.

This matters well beyond sport. Cedric Chin, in a 2020 Commoncog essay, argues that much of what separates strong practitioners in knowledge work is of this kind: designing software, reading a business, choosing what to do first in a messy project.

How to recognise it

Chin describes asking his technical lead, Hieu, why one design was right. Every answer was a rule with an exception: focus on the riskiest integration first, except this one, which he had used before; that one too, except the client might change it, so do the basics first. Pushed far enough, experts end in the same place. A senior engineer Chin asked said simply: "Well, it just felt right."

That pattern is the signal. When an explanation comes as a rule with a long tail of caveats, you are probably looking at tacit knowledge. Atul Gawande's account of a routine appendectomy in "Personal Best" (The New Yorker, 2011) has the same shape: a short standard procedure, then a page of judgements about anatomy, tools, and a diagnosis that turns out to be wrong. His summary of eight years of operating is about judgement, not rules: "you arrive at solutions with less fumbling and more assurance."

The caveats are not a failure to explain. They are the visible edge of thousands of cases the expert has seen and you have not.

Why rules do not capture it

In principle, tacit knowledge might be written down. In practice the attempts have mostly failed. The expert-system projects Chin describes, from the 1970s, ran into what came to be called the knowledge acquisition problem: the rules would not come out of the experts' heads.

Chin relates one case from Gary Klein. Klein was asked to study aircraft maintenance logbooks for a system that might replace maintenance officers. He could fill in most entries by formula. For the rest, the officers reasoned past the logbook: this damage, back in service in two days, so it cannot have cut those lines, so it must have spread this way. Klein's report set out what was formulaic and what was not. The Air Force decided it could not build the system.

Sönke Ahrens, in How to Take Smart Notes (2017), summarises a five-grade scale from Hubert and Stuart Dreyfus: correct use of teachable rules can make you a competent performer, grade three, but not a master or an expert. Rules get you to competence. Past that point, the knowledge is in the pattern library.

Learning it on purpose

The traditional route is apprenticeship: work beside someone good for years and absorb it. You can compress that by arranging the same three ingredients, exposure, prediction and feedback, deliberately.

1. Find judgement that holds up. Choose people whose calls turn out right on cases you can check, not the most confident voice in the room.
2. Watch real cases, and ask what they noticed. "What did you look at first?" and "What would have changed your mind?" draw out cues. "What is the rule?" draws out the caveat list.
3. Predict before they decide. Make your own call on a change review, a packet capture or a design, write it down, then compare. Each mismatch is a cue you missed.
4. Copy whole performances, then get feedback. Chin learned Hieu's design sense by producing designs and asking for his reaction, not by collecting his principles.
5. Keep a case file. Record each mismatch as the situation, your call, their call and the cue that separated them. Review it later.

This is the loop in Deliberate Practice, applied where no textbook sets the targets: the expert's judgement becomes the standard you check against. Chin's sharper claim is that deliberate practice in Ericsson's strict sense needs a field with an established teaching method, such as music or chess, so most professions must learn this way instead. That is a reading of Ericsson, and it fits the modest effects of practice in professions reported on the deliberate-practice page.

Where this could be wrong

The strongest objection: calling something tacit can excuse not writing down what could be written down, and "it feels right" can hide plain error. Kahneman and Klein, in a joint paper ("Conditions for intuitive expertise", American Psychologist, 2009), agree on the limit. An intuition deserves trust only when the environment is regular enough to learn and the person has had the chance to learn it. They add that subjective confidence is not a reliable guide to accuracy.

The reply is to use both. Write down every rule that can be written, as procedures and checklists. Treat an expert's call as evidence, then test it, as in Evidence-Based Troubleshooting. Learn tacitly only where rules run out, and only from people whose judgement has been checked against outcomes. What would settle it: whether the prediction-and-feedback method above shortens the time to sound judgement compared with reading alone. That has not been tested here, which is why this page is Provisional.

For the wider debate, see Polanyi's place in Judgment Under Uncertainty and what text-trained machines leave out in Technology, AI and Human Judgment. Skills of this kind are among the Capabilities Over Tools.

Try this

1. Ask someone skilled to explain one judgement they make often. Count the caveats. Three or more suggests the skill is mostly tacit.
2. On the next five real cases in your field, write your call before you see the expert's. Record each mismatch and the cue that explains it.
3. Rewrite one caveat as a question you can ask yourself on the next case, then check in a month whether it changed your calls.

Further reading

• Cedric Chin, "Why Tacit Knowledge is More Important Than Deliberate Practice" (Commoncog, 2020), and the rest of his tacit knowledge series.
• Michael Polanyi, The Tacit Dimension (1966), a short book and the origin of the term.
• Daniel Kahneman and Gary Klein, "Conditions for intuitive expertise: a failure to disagree" (2009), on when to trust an expert's gut.
• Gary Klein, Sources of Power (1998), on how experienced people decide under pressure.