Skip to content

Judgment Aid

Artifact — instantiates Tacit Knowledge Elicitation

A compact, practice-facing card or prompt that puts the few decisive cues, watchpoints, and exceptions in front of a practitioner at the moment of the judgment — supporting attention without replacing skill.

Elicited knowledge is worthless if it lives in a report nobody opens while working. Judgment Aid is the artifact that pushes the distilled essence of tacit expertise to the point of work: a small, glanceable representation — a cue card, a watchpoint list, a set of "if you see this, look at that" prompts with the key exceptions flagged. Its defining constraint is radical compression for the moment of use: it carries only the few most decisive cues and their exception conditions, because at the moment of judgment a practitioner can act on five watchpoints, not fifty pages. Unlike a comprehensive case library, it is not for study — it is for use under load, and its whole design discipline is deciding what to leave out so that what remains actually gets seen and applied when it counts.

Example

A wind-farm operator distils years of technicians' hard-won judgment about gearbox vibration into a laminated card that rides in every service kit. It doesn't reproduce the vibration-analysis manual. It carries a handful of cues a veteran learned to trust: a rising 1× peak that tracks with load = suspect misalignment, keep running but flag; a broadband floor that climbs when the nacelle heats = suspect bearing, stop and inspect. Beside each, a boundary note: does not apply to the older direct-drive units — different failure signature.

A newer technician on a tower, phone signal patchy, doesn't call the expert or open a PDF; she reads the two lines that match what her analyzer shows and knows this one is a stop-now. The aid worked because it was short enough to consult mid-task and honest about where its rules run out. It's also provisional: when a technician finds a case the card gets wrong, that feedback loops back and the card is revised in the next revision — the aid is a living distillation, not a monument. What the aid can't do is elicit any of this; it packages what interviews, debriefs, and incident reviews already surfaced.

How it works

  • Compress to the decisive few. Include only the highest-yield cues and watchpoints — the handful that most change the call — and ruthlessly cut the rest; the aid's power is in what it omits.
  • Design for the moment of use. Format, length, and placement are dictated by how it will actually be consulted under time pressure, poor conditions, or cognitive load — glanceable beats complete.
  • Flag the exceptions, not just the rules. Each cue carries its boundary — where it doesn't hold, what population or tool it fails on — so the aid supports judgment rather than substituting a brittle rule for it.
  • Keep it alive through feedback. When practitioners hit a case the aid mishandles, that lands in a revision loop; the aid is versioned and updated rather than frozen.

Tuning parameters

  • Compression level — how few items to include. Shorter aids get consulted and remembered but drop nuance; longer ones cover more but go unread at the moment they're needed.
  • Prescriptiveness — watchpoints-and-prompts versus hard rules. Prompts preserve judgment; rules are faster and more consistent but risk deskilling and misfiring on the exceptions.
  • Point-of-use placement — embedded in the tool, on a card in the kit, or in a reference elsewhere. Closer to the moment of judgment means more use but less room for content.
  • Revision cadence — how often the aid is updated from field feedback. Frequent revision keeps it accurate but churns the version practitioners have memorized; rare revision lets it drift stale.

When it helps, and when it misleads

Its strength is reach at the moment of decision: it delivers the elicited essentials where and when the judgment is actually made, guiding attention for a novice and offloading memory for a veteran, without pretending to contain the whole of the skill.

Its failure modes come from the compression that makes it useful. Push it too far toward hard rules and it deskills — practitioners follow the card instead of building the judgment, and when a case falls outside the aid's few cues they have no expertise to fall back on. An aid tuned for novices can even slow experts down, forcing explicit steps they'd long since automated — a real effect worth watching whenever the same aid serves both.[1] And because it is so compressed, it silently drops the exceptions unless they're deliberately flagged, turning a support into a false floor. The classic misuse is treating the aid as the knowledge rather than a pointer to it — mistaking the map for the territory. The discipline that keeps it honest is to keep it prompt-shaped rather than rule-shaped, always carry the boundary conditions, and hold it as an explicitly provisional artifact fed by a live practitioner-feedback loop, never as a finished substitute for skill.

How it implements the components

Judgment Aid fills the components concerned with packaging elicited knowledge for use:

  • provisional_articulation — renders elicited know-how as a compact, practice-facing representation, held explicitly as provisional rather than as settled doctrine.
  • transfer_boundary_annotation — each cue carries its exception and boundary, marking where the distilled rule applies and where it must not be trusted.
  • practitioner_feedback_loop — field use feeds a revision loop that keeps the aid corrected and current rather than frozen.

It packages knowledge but does not elicit it (cue_elicitation, decision_rationale_probe — the interview and analysis mechanisms), and it is not the comprehensive study corpus (critical_incident_case — Case Library Capture); the aid is the front-line distillation, the library the back-end archive.

Notes

Judgment Aid is the sharp end of the archetype and the most tempting to over-trust. Its relationship to Case Library Capture is deliberate division of labor: the library holds the full, studyable corpus with all its cases and caveats; the aid extracts the few cues that survive the moment of use. Derive the aid from the library so the two share boundary annotations and the compressed version can always be traced back to the fuller one.

References

[1] The expertise reversal effect (Kalyuga) — instructional support that helps novices can hinder experts, for whom the extra guidance becomes redundant load — is why an aid pitched at learners can slow veterans and why the same card rarely serves both audiences well. It motivates tuning compression and prescriptiveness to the intended user. ↩