Tacit Knowledge Elicitation¶
Essence¶
Tacit Knowledge Elicitation is the intervention pattern for drawing out know-how that works in practice but has not yet been articulated. It is not merely asking an expert for tips. It is a structured effort to observe practice, identify the cues and distinctions that guide judgment, probe why decisions were made, compare expert and novice perception, and validate the result in realistic contexts.
The archetype is useful because much expertise is compressed into action. A practitioner may recognize a weak signal, pause at the right moment, break a rule safely, escalate early, or ignore a distracting cue without being able to immediately explain the reasoning. The knowledge is real, but it is embedded in examples, bodies, tools, timing, social signals, and local histories.
The goal is not to turn every expert judgment into a rigid rule. The goal is to make enough of the tacit structure visible that it can be inspected, taught, improved, debated, translated, or protected. A mature elicitation draft should say both what has been surfaced and what remains context-dependent.
Compression statement¶
When important know-how is implicit, elicit tacit knowledge through observation, probing, comparison, and contextual validation so the cues, distinctions, judgments, and limits of practice can be shared, taught, improved, or safely translated.
Canonical formula: situated_practice + cue_probe + rationale_probe + expert_novice_contrast + contextual_validation -> articulated_know_how + transfer_ready_material + preserved_judgment_limits
When to Use This Archetype¶
Use Tacit Knowledge Elicitation when performance depends on unspoken cues, distinctions, timing, and judgment. It is especially relevant when a team faces succession risk, inconsistent training, key-person dependency, safety-sensitive handoff, process improvement, automation design, or a gap between formal procedure and real practice.
Use it when experts say things like “you will know it when you see it,” “that is just experience,” “it depends,” or “the manual is right, but not complete.” Those phrases often signal that real knowledge exists but is compressed into practice. They can also signal bias or folklore, which is why elicitation must include validation rather than treating expert authority as sufficient.
Do not use the archetype as a substitute for practice. Some tacit knowledge can be represented as cues, examples, thresholds, or heuristics; other parts may require supervised performance, sensory learning, social calibration, or repeated feedback. Good elicitation marks that boundary rather than hiding it.
Structural Problem¶
The structural problem is invisible competence. The expert’s visible actions are available, but the hidden attention structure is not. Novices can copy the action without knowing why it happened, when it should change, what warning signs preceded it, what alternatives were considered, or which exceptions matter.
This is why purely procedural documentation often fails. A procedure may say what to do, but not what to notice. It may list escalation steps, but not the early signs that a case is becoming abnormal. It may describe a customer-support script, but not how an experienced agent hears confusion, fear, anger, or manipulation. It may list machine checks, but not the smell, vibration, or timing pattern that an expert uses to diagnose trouble.
Tacit knowledge also creates a reliability problem. If valuable know-how lives only in a person or informal group, the organization cannot inspect it, transfer it, improve it, or challenge it. It may lose the knowledge when people leave. It may also preserve unsafe habits because no one has made them visible enough to test.
Intervention Logic¶
A Tacit Knowledge Elicitation intervention begins by naming the tacit target. The target should be a recurring skill or judgment: diagnosing a fault, recognizing risk, adjusting a material, handling an exception, escalating a case, coordinating under pressure, or choosing when not to follow the ordinary path.
The next move is observation. Elicitation should start as close to real practice as possible: live observation, case replay, artifact review, screen recording, simulation, or immediate debrief. Abstract interviews can help, but they often produce generic advice. Concrete practice gives the elicitor something to probe.
The elicitor then separates cue, interpretation, and action. What did the practitioner notice? What did they think it meant? What alternatives did they consider? What did they do, delay, avoid, or escalate? What would have changed their mind? What looks similar but means something different?
Expert-novice comparison is especially powerful. Expertise often shows up less as more information and more as different attention. Experts notice earlier, ignore distractions safely, represent the situation differently, or use different thresholds. Comparing how experts and novices interpret the same case reveals the hidden distinctions that training needs to name.
Finally, the elicited material is validated. The output may become cue lists, case contrasts, heuristics, decision maps, training notes, judgment aids, or protocol annotations, but it should remain provisional until tested. Other practitioners should review it. Learners should try to use it. Counterexamples should be sought. The context that makes the cue reliable should be recorded.
Key Components¶
Tacit Knowledge Elicitation moves from situated practice to transferable representation through a sequence designed to recover the hidden attention structure behind visible action without flattening it into a brittle rule. Expert Practice Observation captures skilled action in the real environment — including tools, timing, interruptions, social cues, and informal workarounds — because interviews alone tend to omit what the practitioner cannot easily put into words. Cue Elicitation draws out the sensory, social, temporal, numerical, or pattern-based signals that guide expert attention, and separates the raw cue from its interpretation and its action implication so a faint smell, a tone shift, or a delay becomes meaningful only in its operational context. Decision Rationale Probe asks why the practitioner chose, delayed, adapted, or rejected an action, using concrete cases and counterfactual prompts to recover branching points, uncertainty, and rejected alternatives that experts usually compress after the fact.
The remaining components turn raw elicited material into shareable knowledge while marking what cannot be reduced to rules. Expert-Novice Comparison reveals what expertise actually changes by having both groups review the same case, simulation, or artifact; the difference in what they notice maps the hidden distinctions training needs to name. Contextual Validation tests whether the elicited knowledge works outside the original expert's head by checking peer agreement, novice application in realistic cases, and failure across different tools, teams, or populations — this is what separates the archetype from knowledge folklore. Provisional Articulation turns elicited know-how into a shareable representation such as a cue card, case library, decision tree, annotated SOP, or judgment aid, while explicitly marking confidence levels, boundary conditions, and disputed points. Finally, Transfer Boundary Annotation flags what can be transferred as written heuristic, what requires practice with feedback, what should be represented by examples rather than rules, and what is safety-sensitive or context-bound — guarding against the common failure of turning judgment into instructions and the equally common failure of pretending no part of the expertise can be made explicit.
| Component | Description |
|---|---|
| Expert Practice Observation ↗ | Expert Practice Observation captures skilled action in the environment where it actually happens. Observation should include tools, timing, interruptions, social cues, local constraints, informal workarounds, and artifacts. The purpose is not surveillance. The purpose is to see what an interview may omit. Observation is most valuable when paired with debrief. Watching an expert turn a knob, pause during a call, delay an escalation, or inspect a patient is only the beginning. The elicitor must ask what the expert noticed, what they were comparing against, and what would have made the same visible action wrong. |
| Cue Elicitation ↗ | Cue Elicitation draws out the signals that guide expert attention. A cue may be sensory, social, temporal, numerical, contextual, or pattern-based. The important question is not only “what did you notice?” but also “what did that cue change?” Strong cue elicitation separates three layers: the raw cue, the interpretation, and the action implication. For example, a faint smell is not yet a diagnosis; it becomes meaningful when paired with machine history, timing, pressure recovery, and knowledge of normal variation. |
| Decision Rationale Probe ↗ | A Decision Rationale Probe asks why the practitioner chose, delayed, adapted, or rejected an action. It should use concrete cases and counterfactual prompts: What else could this have been? What would have made you escalate? What would have made you wait? What did you ignore? What was the first sign that the ordinary rule did not fit? This component guards against generic expertise stories. Experts often compress reasoning after the fact. Good probes recover branching points, uncertainty, and rejected alternatives. |
| Expert-Novice Comparison ↗ | Expert-Novice Comparison reveals what expertise changes. Novices and experts can review the same case, observe the same artifact, or perform the same simulation. The difference in what they notice and how they interpret it becomes a map of tacit competence. This comparison should not shame novices. It should identify learning targets: missing cues, misleading cues, premature conclusions, overreliance on formal rules, and situations where novices need guided practice rather than more text. |
| Contextual Validation ↗ | Contextual Validation tests whether the elicited knowledge works outside the original expert’s head. Can another practitioner recognize the cue? Can a learner apply the distinction to realistic cases? Do peer experts agree? Does the rule fail in a different tool, team, patient population, customer segment, or operating condition? Validation is what distinguishes Tacit Knowledge Elicitation from knowledge folklore. It converts “an expert said this” into “this articulated knowledge has been tested enough to be useful, with these limits.” |
| Provisional Articulation ↗ | Provisional Articulation turns elicited know-how into a shareable representation. The representation may be a cue card, case library, decision tree, annotated SOP, training scenario, checklist with judgment notes, or map of examples and counterexamples. It is provisional because the representation is not the same as the expertise. The draft should include confidence levels, boundary conditions, disputed points, and places where supervised practice is still required. |
| Transfer Boundary Annotation ↗ | Transfer Boundary Annotation marks what can and cannot be transferred explicitly. Some knowledge can become a written heuristic. Some needs practice with feedback. Some should be represented by examples rather than rules. Some is safety-sensitive or context-bound. This component prevents the common failure of turning tacit knowledge into brittle instructions. It also protects the integrity of craft, professional judgment, and embodied skill. |
Common Mechanisms¶
Expert interviews are common, but they are rarely sufficient on their own. They work best when anchored in specific cases, artifacts, and decisions.
Shadowing sessions let an observer see real practice. They reveal timing, context, tool use, informal adaptation, and interactional cues. Shadowing should be followed by debrief because visible action does not automatically reveal interpretation.
Think-aloud protocols ask practitioners to verbalize attention and reasoning while acting or reviewing a case. They are useful for cognitive work but may interfere with high-tempo, embodied, or safety-sensitive performance.
Cognitive task analysis breaks expert work into goals, cues, strategies, decisions, errors, and knowledge requirements. It is a strong mechanism when the practice is complex and judgment-heavy.
Critical incident technique uses successes, failures, near misses, and edge cases to reveal what routine descriptions omit. It is powerful but should be balanced with ordinary cases so that the organization does not overlearn from dramatic exceptions.
Simulation replay reconstructs cases and lets practitioners explain what they notice over time. It helps recover temporal judgment: when to wait, when to act, and when the meaning of a cue changes.
Case libraries store elicited examples, counterexamples, cues, rationales, and outcomes. They are useful when tacit knowledge is learned through contrast rather than a single rule.
Judgment aids present cues, watchpoints, and examples in practice-facing form. They should support attention and reflection without pretending to replace expertise.
Parameter Dimensions¶
Tacit Knowledge Elicitation varies along several dimensions.
The tacitness dimension asks how far the knowledge is from ordinary explanation. Some know-how is simply unrecorded; some is perceptual, embodied, social, or difficult to verbalize.
The context-fidelity dimension asks how close elicitation is to real work. Live observation and simulation preserve context; retrospective interviews are easier but more vulnerable to generic explanation.
The case-range dimension asks whether elicitation includes routine cases, hard cases, edge cases, failures, and misleading lookalikes. A narrow case set produces brittle knowledge.
The validation dimension asks how strongly the elicited knowledge is tested. Peer review, novice application, simulation, outcome evidence, and counterexample search all strengthen confidence.
The transfer-purpose dimension asks what the elicited knowledge will support: training, succession, documentation, process redesign, decision support, automation, or audit. Different purposes require different formats and safeguards.
The ethical-sensitivity dimension asks whether elicitation exposes personal data, worker know-how, security vulnerabilities, undocumented workarounds, or professional judgment that can be misused.
Invariants to Preserve¶
The first invariant is context. A tacit cue detached from the setting that gives it meaning can become dangerous. Always record the conditions under which a cue, heuristic, or exception applies.
The second invariant is the link from cue to action. A list of cues is not enough. The draft must say what the cue changes: attention, interpretation, decision, escalation, timing, or confidence.
The third invariant is boundary evidence. Examples and counterexamples should define where the elicited judgment applies and where it fails.
The fourth invariant is humility. Expert articulation is a hypothesis, not final truth. It should be reviewed, tested, and revised.
The fifth invariant is transfer usability. If only the original expert understands the output, the elicitation has not yet created shareable knowledge.
Target Outcomes¶
A successful Tacit Knowledge Elicitation effort reduces key-person dependency and improves knowledge transfer. It gives learners better cues, richer examples, and clearer questions to ask. It helps organizations distinguish valuable craft from habit, bias, or workaround. It improves documentation by adding judgment notes, exceptions, and context. It can also inform safer automation and decision support by showing which parts of expertise are not ready to automate.
The best evidence is practical: learners perform better on realistic cases, peers recognize the captured distinctions, procedures become more useful without becoming brittle, and the organization can update the capture as conditions change.
Tradeoffs and Failure Modes¶
The central tradeoff is articulation versus distortion. Tacit knowledge becomes more transferable when articulated, but articulation can flatten experience into slogans. A good elicitation output uses examples, counterexamples, confidence labels, and boundary notes to preserve nuance.
Another tradeoff is context fidelity versus portability. Highly contextual knowledge may be accurate but hard to scale. Highly abstract knowledge may travel but lose its force. The right balance depends on the transfer purpose.
The most common failure mode is the expert-interview trap: a team interviews a senior person, records general advice, and calls the result knowledge capture. That output may be useful, but it is not mature Tacit Knowledge Elicitation unless it includes concrete cases, cue probing, rationale probing, and validation.
A second failure mode is over-formalization. Elicited cues become a checklist, the checklist becomes mandatory, and the judgment that made the original practice effective disappears. This is especially dangerous in variable, high-stakes, or human-centered contexts.
A third failure mode is bias capture. Tacit assumptions about people, risk, competence, or normal behavior can masquerade as expertise. Validation should include evidence, diverse cases, and affected perspectives where appropriate.
Variants¶
Critical Incident Elicitation¶
Critical Incident Elicitation uses successes, failures, near misses, and edge cases as high-information windows into tacit judgment. It is useful when routine descriptions are bland but difficult cases reveal what experts actually monitor. Its risk is drama bias: the organization may overlearn from vivid cases while neglecting ordinary practice.
Expert-Novice Contrast Elicitation¶
Expert-Novice Contrast Elicitation compares how experts and novices interpret the same situation. It is useful for training because it reveals the attention patterns and distinctions that novices lack. The comparison should be developmental, not humiliating.
Shadowing-Based Elicitation¶
Shadowing-Based Elicitation starts with real or near-real observation. It is useful when context, tools, timing, interruptions, and social cues matter. Its failure mode is silent observation: the observer records visible steps but never probes what the practitioner noticed or why it mattered.
Cue-Based Judgment Capture¶
Cue-Based Judgment Capture centers the cue -> interpretation -> action chain. It is useful in diagnosis, quality control, early-warning detection, customer interaction, and other fields where weak signals matter. It requires special care to distinguish validated cues from biased impressions.
Tacit Knowledge Transfer Debrief¶
Tacit Knowledge Transfer Debrief occurs inside a learning loop. After observation or practice, expert and learner discuss what was noticed, what was missed, and what should be tried next. This variant overlaps with cognitive apprenticeship mechanisms and should remain under review until that ontology boundary is settled.
Neighbor Distinctions¶
Tacit Knowledge Elicitation is closely related to Observational Learning by Modeling, but they are not the same. Modeling helps learners acquire behavior by watching. Elicitation makes the hidden cues and reasoning inspectable so they can be questioned, represented, validated, and reused.
It is also related to Cognitive Apprenticeship, which appears in the roadmap but is not in the canonical prime list. In this draft it is treated as a proposed prime and as a teaching/mechanism neighbor, not as a canonical source prime. Cognitive apprenticeship can make expert thinking visible during instruction; Tacit Knowledge Elicitation can supply the material that such instruction uses.
Tacit-to-Explicit Translation is downstream. Elicitation asks “what is the hidden know-how?” Translation asks “how should the elicited know-how be encoded as guidance, examples, rules, protocols, or tools?” The roadmap keeps that candidate on merge review, so this draft preserves the boundary rather than absorbing it fully.
Practice-to-Protocol Transition is also downstream. A protocol may be appropriate after tacit know-how has been elicited and validated, but premature protocolization can erase context and craft.
Examples and Non-Examples¶
A good example is a clinical team that observes senior nurses, reconstructs recent deterioration cases, identifies subtle patient-baseline cues, and tests whether trainees can apply those cues in simulation. The output may improve escalation guidance, but it also marks which judgments require supervised clinical practice.
A second example is an industrial maintenance team that captures how senior technicians use sound, vibration, smell, and repair history to diagnose problems. The output is a case library and judgment aid, not merely a checklist.
A third example is a customer-support group that compares expert and novice interpretations of the same calls. The experts notice pauses, tone shifts, and policy signals that novices miss. Those distinctions become coaching material and escalation guidance.
A non-example is a silent training video. It shows visible behavior but does not elicit the hidden attention and rationale. Another non-example is an exit interview that asks a retiring expert for general wisdom without cases or validation. A third non-example is an automation project that mines worker know-how without consent, context, or safety review.
Implementation Checklist¶
- Name the tacit skill or judgment and the decision it affects.
- Select credible practitioners and representative cases.
- Observe or replay actual practice before abstracting.
- Probe cues, thresholds, alternatives, and exceptions.
- Compare expert and novice perception where transfer is a goal.
- Represent the knowledge provisionally with examples and counterexamples.
- Validate the capture in realistic cases.
- Mark what can be written down, what requires practice, and what should remain context-limited.
- Add governance for sensitive, safety-critical, or worker-supplied know-how.
- Connect the output to training, documentation, decision support, or learning loops.
Review Notes¶
The draft is high-confidence as a first-wave archetype because the roadmap marks Tacit Knowledge Elicitation as promote-first and draft-ready. The main review issue is ontology hygiene: cognitive_apprenticeship appears as a roadmap primary prime but is absent from the canonical prime slug list. This draft therefore places it under identity.proposed_primes and keeps only canonical primes in source_primes and related_primes.