How to diagnose prior knowledge before employee training

A prior-knowledge diagnostic should answer a practical question:

What does this employee already know, believe, and know how to do before instruction begins?

It is not a miniature final exam. It is a low-stakes way to uncover the learner's starting point so training can address real gaps instead of repeating familiar material or overlooking a dangerous misunderstanding.

The process begins with the required workplace capability, not with a bank of generic questions. Map what the employee must know and do, select evidence that can reveal those prerequisites, and use the results to adjust instruction.

Key takeaways

• Diagnose prerequisites that matter to a specific workplace task, not general intelligence or broad ability.
• Use more than correct-versus-incorrect scores when reasoning, confidence, or procedure matters.
• Keep the diagnostic short and low stakes so people are willing to reveal uncertainty.
• Adapt training from patterns of evidence; do not label a learner from one response.
• Verify the diagnosis later through a new question, task, or workplace observation.

Start with the capability, not the quiz

A useful diagnostic is anchored to something the employee must eventually decide, explain, or perform.

Suppose a home-care employee must recognize a reportable incident, take the correct immediate action, notify the right person, and document the event accurately. A generic quiz about the employee handbook will not tell you whether those component capabilities are present.

Begin by writing the final outcome in observable language:

Given a realistic client incident, the employee selects the appropriate immediate action, identifies the notification path, and completes the required documentation without omitting critical information.

Then break that outcome into prerequisites:

• vocabulary and definitions;
• relevant rules and exceptions;
• roles and responsibilities;
• decision criteria;
• procedural steps;
• tool knowledge;
• judgment under realistic conditions.

Carnegie Mellon University's learning principles note that learners must acquire component skills, practice integrating them, and know when to apply what they have learned. That is a useful design reminder: a complex task can fail because one component is missing even when the employee appears generally experienced. Source: Carnegie Mellon University, “Principles of Learning”.

Map only the prerequisites that can change the outcome

A diagnostic becomes bloated when designers test everything mentioned in the source document.

Instead, ask of each potential prerequisite:

1. Could missing this knowledge cause the target task to fail?
2. Could an outdated version lead to the wrong action?
3. Does this prerequisite need to be remembered, or can it be looked up?
4. Is it required by everyone or only a particular role?
5. What evidence would reveal whether it is available when needed?

This creates a compact prerequisite map.

For example:

Required capability	Important prerequisite	Diagnostic evidence
Recognize a reportable event	Current incident definition	Classify two contrasting scenarios
Choose the immediate action	Priority and escalation rules	Explain the first action and why
Notify the right person	Current role structure	Select the notification route in a changed case
Complete documentation	Required fields and workflow	Complete a short sample report

The map should preserve the link to the current SOP or policy version. Otherwise, the diagnostic may test yesterday's rule with today's employees.

Choose evidence that matches the required capability

Different diagnostic methods reveal different things.

Short recall or recognition questions

Use these for facts, vocabulary, roles, and clearly defined rules.

A correct answer can show access to the information, but it may not reveal whether the employee can use it in a realistic situation.

Scenario decisions

Use scenarios when context changes the correct response.

A good scenario includes enough detail to require a decision but avoids irrelevant complexity. Contrasting cases are especially useful: two situations may look similar while only one triggers a particular rule.

Explanation prompts

Ask the employee to explain why an option is appropriate.

Reasoning can distinguish genuine understanding from guessing and can reveal an outdated rule even when the final choice happens to be correct.

Confidence ratings

Ask how certain the employee is after answering.

Confidence is not competence, but the combination can guide follow-up:

• correct and confident may indicate stable knowledge;
• correct and uncertain may need reinforcement;
• incorrect and uncertain may indicate a gap;
• incorrect and confident may indicate a competing belief worth investigating.

Do not treat this four-part pattern as an automatic diagnosis. It is a prompt for additional evidence.

Demonstrations and work samples

Use a short demonstration when the outcome is procedural.

Someone may know the steps in theory but struggle with the actual reporting tool, sequence, timing, or handoff. A demonstration often reveals problems that a multiple-choice item cannot.

Keep the diagnostic low stakes and psychologically safe

People reveal more useful information when they are not punished for not knowing.

State the purpose clearly:

This check helps us tailor the learning experience. It is not part of your performance rating or certification score.

That separation matters. If employees believe the diagnostic will be used against them, they may search for answers, avoid admitting uncertainty, or choose the response that sounds safest rather than showing their actual reasoning.

Use a diagnostic that is:

• brief enough to complete without fatigue;
• directly related to the work;
• free of trick questions;
• accessible in language and format;
• transparent about how the information will be used;
• separated from disciplinary or high-stakes decisions unless formally validated for that purpose.

A five-minute check can be more useful than a forty-question pretest if every item has a defined instructional decision behind it.

Prequestions can diagnose and sometimes support later learning

Asking a question before instruction can do more than measure a starting point. Research on prequestioning and pretesting has found that attempting an answer before studying the correct information can sometimes improve later memory for the tested material.

A 2023 review by Pan and Carpenter concluded that benefits have appeared across multiple materials, formats, age groups, and learning settings, while also emphasizing open questions about mechanisms and boundary conditions. Source: Pan and Carpenter, “Prequestioning and Pretesting Effects”.

A separate meta-analysis of the prequestion effect reported a moderate benefit for material directly targeted by the prequestions but little evidence of broad benefit for untested material. Source: St. Hilaire, Chan, and Ahn, “Guessing as a Learning Intervention”.

These studies largely concern educational materials rather than workplace certification. The practical implication is modest:

A well-designed prequestion can focus attention and create a useful learning opportunity, but it should not be treated as proof of workplace readiness.

Always provide or lead into the correct explanation. Do not leave guessed errors unresolved.

Interpret patterns, not isolated answers

One response rarely justifies a strong conclusion.

An incorrect answer might reflect:

• missing knowledge;
• ambiguous wording;
• a memory lapse;
• a misread detail;
• unfamiliar terminology;
• a correct rule applied to the wrong context;
• an old process;
• a stable misconception;
• a user-interface problem.

Look for converging evidence.

For example, an employee classifies three different non-injury events as “not reportable,” explains that reports are required only after harm, and expresses high confidence each time. That pattern supports a targeted hypothesis: the employee may be using an incorrect threshold rule.

The next step is not to label the person. It is to test the hypothesis with another case, confirm the current source requirement, and provide a correction that explains why the existing rule fails.

A systematic review of prior-knowledge activation found that outcomes depend partly on the amount and accuracy of the knowledge being activated and on how activation is designed. Source: Hattan, Alexander, and Lupo, “Leveraging What Students Know”. That evidence comes mainly from learning from text, so workplace teams should use it as a design principle and test it locally.

Turn diagnostic evidence into an instructional decision

Do not collect data that changes nothing.

Define the action for each meaningful pattern before launching the diagnostic.

Evidence pattern	Possible instructional response
Prerequisites are secure	Allow a faster path or mastery check
A small knowledge gap appears	Provide a targeted explanation and practice
An outdated process appears	Contrast old and current procedures explicitly
Reasoning is weak despite correct choices	Add explanation, comparison, and new scenarios
Procedure fails in demonstration	Provide modeling, guided practice, and feedback
Evidence is mixed	Avoid automatic adaptation; gather more evidence

This is not a license to permanently divide people into “beginner” and “expert.” Knowledge is often domain-specific. An experienced manager may be a novice in a new software workflow, while a new employee may already possess strong incident-response skills from another organization.

Adapt the support to the demonstrated prerequisite, not the person's job title or years of service.

Verify the diagnosis after instruction

A diagnostic is a provisional model of the learner.

After targeted instruction, use a different example or task to see whether capability changed. Repeating the exact same question may measure memory for the item rather than understanding.

Verification might include:

• a new scenario with different surface details;
• an explanation of the decision rule;
• a short task demonstration;
• a delayed follow-up question;
• a supervisor observation;
• a reviewed work sample.

The CDC's training-needs guidance recommends examining the gap between current and desired performance and considering both individual and system causes before deciding on training. Source: CDC, “Assess Training Needs”. A diagnostic should support that analysis, not automatically turn every gap into another course.

How this can shape LoreGraph

LoreGraph helps organizations turn existing workplace documents into structured learning. A prior-knowledge layer could make that process more selective.

A future workflow might connect:

source requirement → prerequisite concept → diagnostic evidence → learner state hypothesis → targeted activity → new evidence

That graph would need careful boundaries. It should distinguish:

• a source-supported requirement;
• a learner response;
• an inferred knowledge state;
• confidence in the inference;
• the date and source version;
• the instructional action taken.

Learn more about LoreGraph.

The goal is not to create an AI label for every learner. It is to give creators and learners better evidence about where useful instruction should begin.

Common mistakes

Avoid these failure points:

• Testing trivia: Every question should connect to a real capability or decision.
• Using one format for every outcome: Recognition questions cannot diagnose every skill.
• Treating confidence as truth: Confidence adds context; it does not validate an answer.
• Making the diagnostic high stakes: This encourages answer-seeking and hides uncertainty.
• Adapting from one response: Use patterns and follow-up evidence.
• Ignoring the work environment: A performance gap may come from poor tools, conflicting procedures, or missing support.
• Failing to act on the results: A diagnostic without an instructional decision is data collection without a learning purpose.

Next step

Choose one important task from an existing course.

Write the target performance, identify no more than five prerequisites, and create one piece of diagnostic evidence for each. Include at least one scenario or demonstration rather than relying entirely on recall questions.

Then specify what the training will do differently for each evidence pattern.

The diagnostic is complete only when it changes the learning path—and when a later task checks whether that change helped.

Sources and further reading

• Carnegie Mellon University, “Principles of Learning”.
• Pan and Carpenter, “Prequestioning and Pretesting Effects”.
• St. Hilaire, Chan, and Ahn, “Guessing as a Learning Intervention”.
• Hattan, Alexander, and Lupo, “Leveraging What Students Know”.
• CDC, “Assess Training Needs”.