Ask an experienced surgeon how they know a procedure is going wrong before any instrument has shown a definitive sign. Ask a master baker how they know the dough is ready. Ask a veteran detective how they knew the suspect was lying before they had any evidence. The answer, almost invariably, involves some variation of: "You just know. You can feel it."
This is tacit knowledge — the expertise embedded in practice, perception, and skilled judgment that resists being written down. Michael Polanyi, the Hungarian-British scientist and philosopher who gave tacit knowledge its name, summarized it with a phrase that has become one of the most quoted lines in the philosophy of knowledge: "We know more than we can tell."
This gap between what we know and what we can articulate is not a minor footnote to knowledge. It is, for most domains of genuine expertise, the central feature. And the failure to understand it has produced systematic problems in how organizations transfer knowledge, how educators design training, and how technologists build systems intended to replicate human judgment.
What Tacit Knowledge Is
Tacit knowledge is knowledge that is embedded in the skilled performance of a task rather than in explicit propositions or procedures that can be stated and transmitted. It is know-how as opposed to know-that. It is knowledge in the hands, the eyes, the body, and the intuition — not in the manual.
Michael Polanyi introduced the concept in his 1958 book Personal Knowledge and developed it further in The Tacit Dimension (1966). His central argument was that all knowledge has a tacit dimension: even the most explicit scientific knowledge rests on a substrate of unspoken assumptions, perceptual skills, and practical judgments that are never fully articulated and perhaps never can be.
"We can know more than we can tell. This fact seems obvious enough; but it is not easy to say exactly what it means. Take an example. We know a person's face, and can recognize it among a thousand, indeed among a million. Yet we usually cannot tell how we recognize a face we know." — Michael Polanyi, The Tacit Dimension (1966)
Explicit vs. Tacit Knowledge: The Core Distinction
The distinction between explicit and tacit knowledge is fundamental to understanding how knowledge is created, transmitted, and lost.
Explicit knowledge can be articulated, documented, and transmitted through text, instructions, procedures, or formal education. It is the knowledge in textbooks, manuals, databases, and formal specifications. A recipe, a mathematical proof, a legal procedure, a programming language syntax guide — these are explicit knowledge.
Tacit knowledge resists articulation. It is learned through practice, absorbed through experience, and often cannot be fully transferred by instruction. The ability to ride a bicycle is the classic example: you cannot learn to ride by reading a description of bicycle riding, no matter how detailed. The balance, the reflexive micro-adjustments, the feel of the bike — these are tacit. They must be practiced.
| Dimension | Explicit Knowledge | Tacit Knowledge |
|---|---|---|
| Form | Propositions, documents, procedures | Skills, intuitions, embodied habits |
| Transmission | Teaching, reading, instruction | Apprenticeship, practice, observation |
| Codifiability | High — can be written down | Low — resists full articulation |
| Loss risk | Low — stored in documents | High — leaves with the person |
| Examples | Textbooks, recipes, legal codes | Surgical judgment, athletic skill, taste |
| Transfer speed | Fast — can be distributed at scale | Slow — requires sustained contact |
Examples of Tacit Knowledge in Practice
Medical Diagnosis
Experienced physicians develop diagnostic intuitions that operate faster and more accurately than systematic algorithmic reasoning. They recognize disease presentations from subtle configurations of symptoms, patient demeanor, and physical signs that no checklist captures. This expert intuition — studied extensively by researchers including Dreyfus and Dreyfus in their work on expertise development — is tacit: it was built through years of pattern exposure and cannot be fully transferred to a student through lectures alone.
This is why medical education involves clinical rotations. The explicit knowledge in medical textbooks is necessary but insufficient. The tacit perceptual and judgment skills that distinguish a competent practitioner from a dangerous one can only be developed through supervised practice with real patients.
Skilled Trades and Craftsmanship
A master woodworker knows, by touch and the sound of the plane, whether the surface is ready. A master chef knows by smell and color when the garlic is at the correct stage. A glassblower knows by the glow of the molten glass what temperature it has reached. These are examples of what Polanyi called subsidiary knowledge — knowledge that operates in the background, supporting the focal activity without being consciously attended to.
Written instructions for these crafts can capture the explicit structure of the process. They cannot capture the felt sense of the material that the expert uses to navigate it. This is why apprenticeship dominated skilled trades for centuries and why the elimination of apprenticeship in favor of purely classroom-based training consistently fails to produce fully capable practitioners.
Software Architecture and Engineering Judgment
Experienced software architects develop judgment about system design that goes beyond the explicit principles in architecture textbooks. They can feel when a design is accumulating hidden complexity, when a seemingly simple abstraction will become a long-term liability, when a proposed solution is technically correct but fragile. Junior engineers learn this primarily by working alongside experienced ones — by seeing decisions being made in context, asking why, and gradually internalizing the pattern recognition.
Code reviews are, among other things, a mechanism for tacit knowledge transmission: experienced engineers not only identify problems but explain the reasoning, sharing some of the judgment that produced the identification even if they cannot fully articulate the whole of it.
Management and Leadership
Experienced managers develop judgment about people, timing, and organizational dynamics that is difficult to systematize. When to push and when to wait. How to read a room. When a conflict is best addressed directly and when indirect approaches are more effective. These judgments are shaped by accumulated experience, personal pattern recognition, and intuitions about human behavior that resist formulaic expression.
This is partly why management development through classroom education alone consistently disappoints: the explicit frameworks teach how to think about problems, but the judgment required to apply those frameworks in specific real situations is tacit and develops only through experience.
Why Organizations Lose Critical Tacit Knowledge
The challenge of tacit knowledge for organizations is acute. When an expert retires, leaves, or dies, the tacit knowledge they hold leaves with them — unless specific interventions have been made to transfer it.
The Retirement Problem
In industries with aging workforces — manufacturing, engineering, healthcare, utilities — the retirement of large cohorts of experienced workers represents a knowledge loss that is rarely fully appreciated until it has occurred. Organizations frequently invest in documenting retiring experts' explicit knowledge (their procedures, contacts, specifications) while failing to invest in capturing the judgment and skill that made them valuable.
The NASA "lost knowledge" problem is one of the most discussed organizational examples: when knowledge of how to build Saturn V rockets was lost through retirements and program shutdowns, NASA discovered it could not recover that capability simply by reading the original documentation. The tacit knowledge of the engineers who built and operated the systems — embedded in judgment calls that were never written down — was irretrievably gone.
The Documentation Illusion
Organizations often respond to knowledge risk by creating documentation: wikis, procedure manuals, lessons-learned databases, knowledge management systems. These are valuable for explicit knowledge. For tacit knowledge, they offer a false sense of security.
Documented procedures capture the what; they rarely capture the judgment about when procedures should be adapted, what to do when the procedure does not quite fit the situation, or how to recognize the subtle signs that indicate an exception condition. The experienced practitioner holds all of that tacit knowledge and applies it every time they perform the procedure. The documentation holds a partial skeleton.
Onboarding and Ramp Time
The time it takes for new employees to become fully productive — sometimes measured in months or years for complex roles — is largely the time required to acquire the tacit knowledge of the role. The explicit knowledge (policies, procedures, systems) can be transferred quickly. The judgment, the pattern recognition, the relational understanding of how the organization actually works — these develop slowly, through observation and experience.
Organizations that treat onboarding as primarily an information transfer problem consistently underestimate ramp time and underinvest in the relationship-based and apprenticeship-based learning that develops tacit competence.
How to Transfer Tacit Knowledge
Since tacit knowledge cannot be fully written down, organizations that wish to transfer it must create conditions for it to be observed, practiced, and absorbed. Several approaches have evidence behind them.
Structured Apprenticeship and Mentoring
The oldest and still most effective mechanism for tacit knowledge transfer is apprenticeship: sustained, close working contact between an expert and a learner, with the learner progressively taking on more responsibility for authentic work under guidance. Apprenticeship works because it exposes the learner to the expert's judgment in real situations, allows the learner to observe the tacit knowledge operating rather than just hearing about it, and provides immediate feedback on the learner's own attempts to apply it.
Formal mentoring programs in organizations often fall short of apprenticeship because the contact is less intensive, the work is less authentic, and the feedback is less immediate. Effective mentoring for tacit knowledge transfer requires mentors and mentees to work on real problems together, not just to have periodic career conversations.
Communities of Practice
Etienne Wenger's concept of communities of practice — groups of practitioners who share a domain of interest and regularly interact to learn from each other — provides a mechanism for tacit knowledge to circulate through a professional community rather than being held by isolated individuals.
In communities of practice, tacit knowledge surfaces through stories, discussions of specific cases, demonstrations, and collaborative problem-solving. The knowledge does not fully articulate — but it becomes more visible, more shared, and more accessible than it would be in isolated individual practice.
Structured Reflection and Storytelling
One approach to bridging the gap between tacit and explicit knowledge is to prompt experts to tell stories about specific cases rather than to describe general principles. Case-based learning — widely used in medical, legal, and business education — works partly on this principle: experts discussing real situations reveal tacit knowledge through their commentary that they would not express if asked to describe their general approach.
Cognitive task analysis methods, developed in applied cognitive psychology, use structured interviews to elicit expert decision-making processes in ways that reveal more of the tacit dimension than standard knowledge-elicitation techniques.
Learning by Doing
For learners, the primary implication of tacit knowledge is that practice is irreplaceable. No amount of reading, lectures, or conceptual understanding substitutes for doing the actual work in realistic conditions. Deliberate practice — practice structured to develop specific components of skill, with feedback — is the most effective path to developing tacit expertise.
Research by K. Anders Ericsson on expert performance found that the distinguishing characteristic of experts across many domains was not natural talent but accumulated deliberate practice. The tacit knowledge of expertise is not innate; it is constructed through thousands of hours of practicing the specific perceptual and judgment skills the domain requires.
Tacit Knowledge and Artificial Intelligence
Tacit knowledge poses one of the most fundamental challenges for artificial intelligence systems. Large language models and other AI systems are trained primarily on explicit knowledge — documented text, structured data, written procedures. They have processed more explicit knowledge than any human could read in a lifetime.
What they lack is the tacit knowledge that comes from embodied practice, physical interaction with the world, and the kind of experience-based pattern recognition that humans develop through years of performing real tasks in real contexts.
This is why AI systems that perform impressively on language tasks — where explicit knowledge is the primary substrate — still struggle with tasks that require physical skill, practical judgment in novel situations, or the kind of expertise that cannot be fully articulated. The boundary between what current AI systems can and cannot do maps surprisingly well onto the boundary between explicit and tacit knowledge.
For organizations evaluating AI automation, this suggests a useful framework: tasks that can be fully documented as explicit procedures are candidates for automation; tasks that depend heavily on tacit judgment, perceptual skill, or embodied expertise are likely to remain human for the foreseeable future.
Implications for Education and Training
Education systems are optimized almost entirely for explicit knowledge transmission: lectures, textbooks, written assessments. Tacit knowledge acquisition requires a different approach that most formal education provides inadequately.
Medical education has been the most deliberate in designing for tacit knowledge transfer — the extended clinical training that follows medical school exists precisely to develop tacit diagnostic and procedural skill. Other professional fields have been slower to recognize that their educational models leave critical competencies undeveloped.
The most effective training programs for tacit knowledge development share common features:
- Real work, not simulations: When simulations are used, they are high-fidelity and consequential enough to generate genuine engagement
- Expert observation: Learners have sustained access to watching experts perform the relevant tasks
- Progressive responsibility: Learners take on increasing responsibility for real work rather than only observing
- Feedback tied to specific judgments: Feedback addresses the judgment calls made, not just the outcomes
The Spectrum of Tacitness
It is worth noting that tacit and explicit knowledge are not a binary distinction but a spectrum. Some knowledge is almost entirely explicit: mathematical proofs, legal statutes, programming syntax. Some knowledge is almost entirely tacit: the balance used in riding a bicycle, the intuition of an experienced nurse recognizing a deteriorating patient before the monitors signal anything.
Much practical knowledge sits in the middle: it can be partially articulated, which is valuable for teaching and documentation, while the remaining tacit dimension can only be developed through practice. A surgical procedure can be described in a manual, and that description is useful — but the judgment about when to deviate from the standard procedure, how to handle unexpected anatomical variations, and when to stop and reassess is tacit and develops only through supervised practice.
Ikujiro Nonaka and Hirotaka Takeuchi's influential framework on organizational knowledge creation describes processes by which knowledge moves along this spectrum: socialization (tacit to tacit, through shared experience), externalization (tacit to explicit, through articulation), combination (explicit to explicit, through systematization), and internalization (explicit to tacit, through learning by doing). Organizations that manage knowledge deliberately create conditions for all four processes.
The practical implication: when trying to capture or transfer tacit knowledge, the goal is not necessarily to make it fully explicit — that may be impossible — but to articulate as much as possible, identify what remains tacit, and design learning experiences that develop the tacit remainder through practice and observation.
Summary
Tacit knowledge — the expertise embedded in skilled performance that resists full articulation — is central to most domains of genuine competence. Michael Polanyi's observation that "we know more than we can tell" captures a fundamental feature of human knowledge rather than an edge case.
For organizations, the key challenge is that tacit knowledge leaves with the people who hold it. Documentation cannot fully capture it; apprenticeship, mentoring, communities of practice, and deliberate practice are the mechanisms through which it transfers.
For learners and practitioners, the implication is that reading about a skill is necessary but insufficient. The tacit dimension of expertise develops through practice, observation of experts in real work, and the accumulated experience of making judgment calls and receiving feedback.
For AI and technology, tacit knowledge defines the frontier of what can be automated: systems trained on explicit knowledge can do much, but the judgment, perceptual skill, and embodied expertise that constitute tacit knowledge remain, for now, distinctively human.
Frequently Asked Questions
What is tacit knowledge?
Tacit knowledge is knowledge that a person holds but cannot fully articulate or transfer through explicit instruction. It is the know-how embedded in skilled performance — how an experienced surgeon feels when a procedure is going wrong, how a master craftsperson knows when a material is ready, how an expert teacher reads a classroom. Michael Polanyi, who introduced the concept, summarized it as: 'We know more than we can tell.'
What is the difference between tacit and explicit knowledge?
Explicit knowledge can be articulated, written down, and transferred through documentation: instructions, procedures, formulas, and rules. Tacit knowledge resists this — it exists in practice, perception, and embodied skill rather than in propositions. The ability to ride a bicycle is a classic example of tacit knowledge: you cannot learn to ride by reading instructions, regardless of how detailed they are.
Why does tacit knowledge matter for organizations?
Organizations lose critical tacit knowledge when experienced employees retire or leave without adequate knowledge transfer. This is sometimes called 'brain drain.' In knowledge-intensive industries, tacit knowledge — how experts diagnose problems, navigate institutional relationships, and make judgment calls — can represent the most valuable and irreplaceable organizational capability. Unlike explicit knowledge captured in documents, tacit knowledge walks out the door with the person.
How can organizations transfer tacit knowledge?
Since tacit knowledge cannot be fully written down, transferring it requires sustained interaction between the expert and learner. Proven approaches include structured apprenticeship and mentoring programs, job shadowing and co-working on real problems, communities of practice where practitioners discuss their work, storytelling and case-based learning, and deliberate reflection exercises where experts attempt to articulate their reasoning. Each method works by creating conditions for tacit knowledge to be observed and absorbed.
What does tacit knowledge mean for artificial intelligence?
Tacit knowledge poses a fundamental challenge for AI systems trained on explicit text. Large language models can process documented knowledge but cannot directly access the embodied, perceptual knowledge that experts hold. This is part of why AI excels at tasks that have been extensively documented and struggles with tasks that rely heavily on judgment, physical skill, and contextual pattern recognition developed through years of direct experience.