Skip to content
Final StateTacit Knowledge: We Can Know More Than We Can Tell
VOL. I  ·  NODE 123▢  ATLAS

THE HEAD THAT WALKS OUT

Tacit Knowledge: We Can Know More Than We Can Tell

Tacit knowing is the background integration of cues that lets a person recognize, judge, and act without fully specifying every contributing detail.

THE CUE BEHIND THE ACT

Focal action rests on subsidiary cues

Diagram showing a focal judgment supported by multiple subsidiary cues integrated in the background.The figure replaces a simplistic explicit-versus-tacit iceberg with Polanyi's focal and subsidiary structure: many cues jointly support one attended judgment.SUBSIDIARY CUESSUPPORT FOCAL ACTIONFOCALJUDGMENTFEATURESENSATIONCONTEXTRELATIONCONTRASTMEMORYINTEGRATED IN THE BACKGROUNDCUES CAN BE DISCUSSEDTHE FULL COMBINATION MAY RESIST RULES
  • Focal awareness: the face, diagnosis, or next move receiving attention
  • Subsidiary awareness: features and sensations integrated in the background
  • Experts can often discuss cues, but may not exhaustively specify how they combine

Tacit does not mean secret or impossible to teach. It means the knowing is not exhausted by a list of propositions; demonstration, practice, and feedback still matter.

THE SCIENTIST OF KNOWING

A physical chemist asked what rules leave out

Source panel tracing Michael Polanyi from physical chemistry to The Tacit Dimension and its focal-subsidiary account of knowing.The panel establishes the author, publication, and conceptual move without turning biography into evidence for the theory.MICHAEL POLANYIA SCIENTIST OF KNOWINGPHYSICALCHEMISTBUDAPESTBERLIN1933MANCHESTER1966THE TACITDIMENSIONBOOKTHE CONCEPTUAL MOVESUBSIDIARY CUESFOCAL KNOWINGPERSONAL JUDGMENT IN KNOWING
  • Michael Polanyi moved from Budapest to Berlin and then to Manchester in 1933
  • His scientific work informed a wider account of personal judgment in knowing
  • The Tacit Dimension, 1966, develops subsidiary and focal awareness

The source is Polanyi's 1966 book: face recognition is his opening illustration, not proof that expertise can never be analyzed or taught.

MORE THAN WE CAN TELL

We can know more than we can tell.

Michael Polanyi, The Tacit Dimension, 1966, p. 4

The sentence sets a limit on exhaustive articulation, not a ban on elicitation: experts may reveal cues, contrasts, and stories without converting the whole skill into a context-free rulebook.

THE REUSE BURDEN

Reusable knowledge must fit a future user and situation

Knowledge-reuse map connecting producer, repository or intermediary, and future reuser across different reuse situations.The figure shows that author effort is only one factor; successful reuse depends on context packaging, intermediary support, and the future reuser's task.KNOWLEDGE REUSE IS A SYSTEMAUTHOR EFFORT IS ONE PARTPRODUCERTIME · INCENTIVEFORESIGHTREPOSITORYINTERMEDIARYCONTEXTFINDABILITYREUSERTASK · EXPERTISELOCAL CONTEXTREUSE SITUATION SHAPES SUCCESS
  • Lynne Markus, 2001: reuse situations differ by who reuses whose knowledge
  • Producers may lack time, incentive, or foresight to package context for unknown reusers
  • Successful reuse also depends on the repository, intermediary, and reuser

Markus's theory is broader than 'the author never benefits': it asks which reuse situation exists and what knowledge, incentives, and roles that situation requires.

An archive can preserve evidence without reproducing expertise

That is the turn: capture can reduce key-person exposure, but the retained record still needs provenance, context, practice, and a person able to test it against the world.

PROBE A REAL INCIDENT

Ask for a difficult decision, then reconstruct the cues

Critical Decision Method interview reconstructing an incident timeline, cues, goals, options, and counterfactuals.The figure presents a structured cognitive interview rather than the overclaim that experts cannot state rules and can only tell stories.PROBE ONE NONROUTINE INCIDENTCRITICAL DECISION METHODSTARTSHIFTDECISIONOUTCOMEPROBE THE DECISION POINTCUESGOALSOPTIONSUNCERTAINTYCOUNTERFACTUALRICH EVIDENCE FROM ONE EVENTNOT A COMPLETE EXTRACTION OF THE EXPERT
  • Critical Decision Method: a structured interview around a nonroutine incident
  • Probe timeline, cues, goals, options, uncertainty, and counterfactuals
  • The output is a richer account of one decision — not a complete extraction of the expert

Klein, Calderwood and MacGregor's Critical Decision Method gives elicitation a disciplined protocol while respecting the limit Polanyi names.

NINETY-FIVE PERCENT, IN SIMULATION

An agent recovered distributed facts in synthetic companies

Synthetic-company simulation with an agent questioning employees across 864 runs and reaching 94.9 percent full-knowledge recall.The exhibit makes the simulation, synthetic structure, run count, and factual-retrieval boundary visible.ITERATIVE QUESTIONINGRECONSTRUCTS DISTRIBUTED FACTSEMPLOYEE FRAGMENTS94.9%FULL-KNOWLEDGERECALL864 SIMULATIONSSYNTHETIC COMPANIESFACTUAL RECALLNOT TACIT SKILL TRANSFER
1Zuin et al., IJCNN 2025: 864 simulations, synthetic company structures, 94.9% full-knowledge recall.DOI 10.1109/IJCNN64981.2025.11227259

The conference paper supports a method for navigating distributed organizational information; it does not show that an agent acquired live experts' judgment.

  • Zuin et al., IJCNN 2025: 864 simulations across synthetic company structures
  • An LLM agent iteratively questioned simulated employees to reconstruct dataset descriptions
  • Reported 94.9% full-knowledge recall; this was factual retrieval, not tacit skill transfer

WHAT THE ARCHIVE CAN HOLD

Store traces; reconstruct answers; keep practice outside the archive

Three-layer knowledge handoff separating captured traces, query-time reconstruction, and human practice.The figure shows what an archive can carry and what remains outside it: embodied rehearsal, feedback, and accountable judgment.WHAT THE ARCHIVE CAN HOLDTRACES INSIDE · PRACTICE OUTSIDE01CAPTURE TRACESINCIDENTS · DEMONSTRATIONSOUTCOMES · SOURCES02RECONSTRUCT ANSWERSSCOPED QUESTION · SOURCE TRAILFRESHNESS03RETAIN PRACTICEREHEARSAL · SUPERVISIONFEEDBACK · CHALLENGETHE ARCHIVE CAN SUPPORT PERFORMANCEIT CANNOT CERTIFY SKILL
  • Capture: incidents, demonstrations, decisions, outcomes, and source trail
  • Retrieve: assemble a scoped answer for a real question
  • Retain: rehearsal, supervised practice, and authority to challenge the record

This is the boundary for fixing the wiki's clock: retrieval can become timely without pretending that generated packaging is the tacit skill itself or allowing deskilling to hollow the stand-in.

WHILE THE PERSON IS STILL HERE

Test the handoff before departure makes the test impossible

  • Choose one costly exception the key person handles
  • Have a stand-in act using the archive while the expert observes
  • Record missing cues, correct the source trail, and schedule another unaided trial

If only one person can diagnose the case, the exposure is live. Reduce it through a witnessed handoff, then continue to When a Company Wiki Becomes a Graveyard, Fix the Clock.

Read the transcript

01 · THE HEAD THAT WALKS OUT

The manual stays on the shelf. The technician leaves. The next morning, a machine makes a faint new rattle. Every documented step is still available, but the person who could tell harmless vibration from an early failure is not. This is the key-person risk inside tacit knowledge. The operation can retain procedures and lose the practiced noticing that made those procedures work in unfamiliar cases. Departure removes the judgment and, just as importantly, the chance to ask how that judgment was formed.

02 · THE CUE BEHIND THE ACT

Michael Polanyi described a structure of focal and subsidiary awareness. Focally, you attend to the face you recognize, the diagnosis you make, or the next move you choose. Subsidiarily, you rely on features, sensations, and relationships integrated in the background. Shift attention to every component and the skilled act may become harder, not easier. Tacit does not mean secret, magical, or impossible to teach. It means the knowing is not exhausted by a list of propositions. Experts can discuss cues and demonstrate action without fully specifying how every cue combines in the moment.

03 · THE SCIENTIST OF KNOWING

Polanyi reached philosophy through science. Born in Budapest, he became a distinguished physical chemist in Berlin and moved to Manchester in 1933. His wider account of knowing grew from a resistance to the idea that rules could mechanically produce judgment. In The Tacit Dimension, published in 1966, face recognition opens the argument. We identify a familiar face among many, yet struggle to enumerate the particulars by which we know it. That illustration establishes a limit on exhaustive articulation. It does not prove expertise can never be analyzed, demonstrated, practiced, or taught.

04 · MORE THAN WE CAN TELL

Polanyi's exact sentence is: We can know more than we can tell. The word can matters. It describes a capacity and a limit, not a mystical ban on explanation. An expert may reveal cues, contrasts, stories, and corrections while still failing to convert the entire skill into a context-free rulebook. The practical consequence is not to abandon documentation. It is to stop treating documentation as a complete substitute for demonstration, feedback, and supervised practice.

05 · THE REUSE BURDEN

A second problem appears when knowledge must travel. In 2001, Lynne Markus proposed a theory of knowledge reuse built around different reuse situations. Someone reusing their own knowledge is not in the same position as a novice reusing an expert's record, or a team mining records created for another purpose. Producers may lack time, incentive, and foresight to package context for unknown future users. But author effort is not the whole system. Repository quality, intermediaries, and the reuser's ability also shape success. The question is not simply who pays. It is what this reuse situation requires.

06 · CAPTURE IS NOT COMPLETE

Now the turn. An archive can preserve evidence without reproducing expertise. That is still valuable. A dated incident, a demonstration, a decision and its outcome can keep the next person from starting blind. But the record still needs provenance, context, freshness, and someone able to test it against the world. Capture reduces exposure. It does not turn stored traces into the practiced perception that created them.

07 · PROBE A REAL INCIDENT

One disciplined elicitation method starts with a nonroutine incident. Klein, Calderwood, and MacGregor's Critical Decision Method uses a structured interview to reconstruct the timeline, then probes cues, goals, options, uncertainty, and counterfactuals. What did you notice? What alternatives did you consider? What would a less experienced person have missed? The result is a richer account of one difficult decision, not a complete extraction of the expert. That modest claim is the method's strength. It turns a blank request for everything you know into evidence anchored to an event.

08 · Advertisement · 11x AI Growth Workers

Pipeline is not one task. It is a chain of small handoffs: find the buyer, research the account, respond fast, qualify cleanly, follow up on time. 11x gives sales, marketing, and RevOps teams digital workers for that background motion. The system keeps work moving, while people focus on the conversations that deserve them.

09 · NINETY-FIVE PERCENT, IN SIMULATION

A 2025 conference paper by Gianlucca Zuin and colleagues tested an automated questioning method. Across eight hundred sixty-four simulations of synthetic company structures, a language-model agent iteratively questioned simulated employees to reconstruct dataset descriptions distributed through the organization. The authors report 94.9 percent full-knowledge recall. Keep the boundary attached. This was synthetic organizational information and factual retrieval, not a live company, an employee interview study, or transfer of tacit skill. The result supports navigation and reconstruction as a research direction. It does not show that the agent learned expert judgment.

10 · WHAT THE ARCHIVE CAN HOLD

A useful archive therefore has three layers. Capture traces while work happens: incidents, demonstrations, decisions, outcomes, and sources. At question time, reconstruct a scoped answer and show where it came from. Outside the archive, retain practice: rehearsal, supervised action, feedback, and authority to challenge the record. The model can package traces. It cannot certify that the stand-in has acquired the skill. Timely retrieval fixes one part of the old wiki problem. Tested performance carries the rest.

11 · WHILE THE PERSON IS STILL HERE

Do not wait for departure to discover whether the handoff works. Choose one costly exception the key person normally handles. Have a stand-in act from the archive while the expert observes without taking over. Record the missing cues, repair the source trail, and schedule another trial without help. If only one person can still diagnose the case, the exposure remains live. The cheapest time to test and improve that transfer is while the person is still present, when failure is a lesson rather than an outage.

01 / 11 · THE HEAD THAT WALKS OUT0:00 / 7:33