Mental models at work: Why knowing the steps is not enough

A procedure tells an employee what to do in a known situation.

A mental model helps the employee understand why the procedure works, which cues matter, how the parts relate, and when the normal sequence must change.

That difference becomes visible when reality does not match the training example. An employee who memorized steps may continue the sequence even when an exception applies. An employee with a useful model can recognize the changed condition, explain the trade-off, and choose the appropriate response.

This article introduces a practical four-layer framework for designing that understanding. It is a LoreGraph synthesis of established ideas about prior knowledge, worked examples, transfer, and complex skill learning—not a validated psychological scale or a claim that every internal representation can be captured completely.

Key takeaways

• Procedural recall and system understanding are related but different outcomes.
• A useful workplace mental model connects purpose, structure, decision logic, and boundaries.
• Explanations should show why an action is selected, not only what action occurred.
• Varied examples help reveal which features matter and which are incidental.
• A new scenario and an explanation of reasoning provide better evidence than repeating the demonstrated case.

Steps are necessary but fragile

Clear steps reduce ambiguity and support consistent work.

They are especially useful when:

• the task follows a stable sequence;
• errors are costly;
• a beginner needs orientation;
• timing and handoffs matter;
• the work must be auditable;
• the employee can consult the procedure during performance.

The problem appears when training treats the sequence as the complete capability.

A learner may remember:

1. open the reporting tool;
2. select the incident type;
3. enter the event details;
4. notify the supervisor;
5. submit the report.

But the learner may still not know:

• which events qualify as incidents;
• whether an urgent action must happen before documentation;
• why particular details matter;
• which supervisor is responsible in a special case;
• what to do when the tool is unavailable;
• when the standard sequence no longer applies.

The sequence supports execution. The missing relationships support judgment.

Mental models organize relationships, not just facts

In practical terms, a mental model is the learner's internal representation of how a system or situation works.

It may include:

• important entities and roles;
• cause-and-effect relationships;
• expected sequences;
• conditions and thresholds;
• possible consequences;
• boundaries and exceptions;
• strategies for checking whether an action is appropriate.

Mental models are not necessarily complete, conscious, or accurate. Two employees can remember the same policy wording while reasoning from different models.

That is why a final answer alone can be misleading. The reasoning behind the answer may reveal whether the learner understands the governing relationship or simply recognizes a familiar pattern.

The National Academies' How People Learn II emphasizes that learning depends on how knowledge is organized and connected to context, prior knowledge, and later use. Source: National Academies, How People Learn II. A workplace mental model is one useful design lens for those organized relationships.

A practical four-layer framework

The framework has four connected layers.

1. Purpose and protected outcome

What is the process trying to achieve or protect?

Examples:

• client safety;
• accurate records;
• confidentiality;
• a reliable handoff;
• legal compliance;
• timely escalation.

Purpose does not replace the rule. It helps the learner understand why certain features matter.

2. System structure

Which people, tools, information, and process stages are connected?

The learner should understand:

• who owns each decision;
• where information comes from;
• what another role needs;
• how the current task affects later work;
• where the source of truth lives.

3. Decision logic

Which cues, conditions, and priorities determine the next action?

This includes:

• if–then relationships;
• thresholds;
• distinctions between similar cases;
• priority rules;
• reasons for selecting one path over another.

4. Boundaries and exceptions

Where does the normal model stop working?

The learner should recognize:

• exceptions;
• emergencies;
• conflicting requirements;
• missing information;
• tool failure;
• situations requiring escalation rather than independent judgment.

These layers are connected. A decision rule makes more sense when the learner understands the protected outcome and process structure. An exception becomes meaningful when the normal model is already clear.

Explain the expert’s reasoning, not only the finished answer

A worked example can show a complete task. A process-oriented worked example goes further by explaining how and why the performer chose each important action.

Compare these examples.

Answer-only example

The coordinator selects the urgent escalation path.

Reasoning example

The coordinator selects urgent escalation because the scenario includes an immediate safety threat. Documentation still matters, but the protected outcome requires the first action to address safety before the normal reporting sequence continues.

The second version exposes:

• the critical cue;
• the relevant priority;
• the relationship between two actions;
• why the normal order changes.

Research on process-oriented worked examples argues that expert “how” and “why” information can support understanding and transfer for complex cognitive tasks. Source: van Gog, Paas, and van Merriënboer, “Process-Oriented Worked Examples”.

More explanation is not always better. As learners gain expertise, process detail can become redundant. A later study by the same research group found that process information could initially help and later become less efficient, illustrating the need to fade support. Source: “Effects of Studying Sequences of Worked Examples”.

Use contrasting cases to reveal the important cues

One example can accidentally teach the learner to copy surface details.

Use two or more cases that make the governing distinction visible.

For example:

Case A

A client complains about a delayed nonessential service. There is no immediate threat.

Case B

A client reports being unable to access essential medication and is showing signs of distress.

Both cases involve a client complaint. Only one includes cues that may require an urgent response.

Ask learners:

• What is similar?
• What is different?
• Which difference changes the action?
• Which detail is irrelevant?
• What principle explains the choice?

This helps move attention from the story's surface to the decision structure.

Do not create artificial trick questions. The distinction should reflect a verified source requirement or a real performance judgment.

Vary the situation without changing the governing principle

Transfer requires applying knowledge beyond the exact training example.

Variation can change:

• the people involved;
• the order in which information appears;
• the communication channel;
• the tool available;
• the severity or timing;
• one irrelevant surface detail;
• the combination of conditions.

Keep the underlying principle stable at first. Then introduce cases where a different rule or exception applies.

A meta-analysis of workplace training transfer found substantial variability and identified relationships between transfer and factors involving trainees, training design, and the work environment. Source: Blume and colleagues, “Transfer of Training: A Meta-Analytic Review”. Mental-model-oriented training cannot overcome a workplace that blocks application, but varied practice can make the learned structure more usable across relevant cases.

Ask learners to predict and explain

Passive agreement can hide an incomplete model.

Use prompts such as:

• What will happen next?
• Which detail matters most?
• Why does this condition change the normal sequence?
• What would make another option correct?
• Where would you verify this?
• When should you stop and escalate?
• How is this case different from the previous one?

Predictions expose the learner's current model before the answer is shown. Explanations reveal relationships that a selected option cannot.

Feedback should address the reasoning:

You identified the correct role, but the notification should occur after the immediate safety action because this condition changes the normal priority.

That is more useful than:

Incorrect. The answer is B.

Test the model with a new case

Do not assess mental-model quality by asking learners to repeat the explanation they just read.

Use a new situation and look for evidence that the learner can:

• identify the relevant cues;
• ignore irrelevant details;
• select an appropriate principle;
• explain the relationship;
• recognize an exception;
• adapt the procedure;
• identify uncertainty and seek support.

A correct action with weak reasoning may need another case. A reasonable explanation with an execution error may reveal a separate tool or procedure gap.

The assessment should match the risk. For a low-stakes administrative task, one or two scenarios may be enough for formative feedback. For safety-critical certification, the organization may require validated assessment, supervised demonstration, and workplace observation.

Represent the model in connected knowledge

A mental-model-oriented knowledge structure needs relationships, not only topic labels.

Useful node types include:

• purpose;
• protected outcome;
• concept;
• role;
• resource;
• condition;
• decision;
• action;
• consequence;
• exception;
• evidence;
• source requirement.

Useful relationships include:

• protects;
• causes;
• requires;
• changesDecision;
• ownedByRole;
• validUnderCondition;
• createsConsequence;
• exceptionTo;
• supportedBy;
• derivedFrom.

For example:

Immediate safety threat
changesDecision
Normal documentation sequence

Urgent escalation
protects
Client safety

Notification rule
derivedFrom
Current incident-response SOP

This makes the logic inspectable and supports traceability when a source changes.

How this can shape LoreGraph

LoreGraph helps organizations turn workplace documents into structured learning. A mental-model-aware workflow would not stop at extracting headings, facts, and steps.

It could help creators identify:

• the outcome a requirement protects;
• the roles and resources involved;
• conditions that change the path;
• meaningful contrasts;
• likely exceptions;
• scenarios that reveal reasoning;
• relationships that need source verification.

Learn more about LoreGraph.

AI can propose these relationships, but it should not invent operational logic. The creator must confirm that each rule, condition, and exception is supported by an approved source or qualified subject-matter expertise.

Limits and common mistakes

A mental model is a design concept, not a perfect picture of what exists in someone's mind.

Avoid these mistakes:

• Replacing procedures with vague principles: Employees still need accurate steps.
• Explaining purpose as permission to improvise: Boundaries and escalation rules remain essential.
• Adding unlimited rationale: Guidance can become redundant or distracting.
• Using only one example: Learners may copy the surface pattern.
• Inventing scenarios unsupported by the source: Hypothetical cases must preserve verified rules.
• Testing verbal explanation only: Real performance may require tools, timing, and coordination.
• Ignoring the work environment: Transfer also depends on opportunity, support, and compatible systems.
• Treating the framework as validated measurement: The four layers are a practical synthesis to guide design and review.

Next step

Choose one procedure that contains several conditions or exceptions.

Write its visible steps. Then add four columns:

1. What outcome does each step protect?
2. Which people, information, and tools does it connect?
3. Which cue determines whether the step applies?
4. What boundary or exception changes it?

Create two contrasting cases and one new transfer case. Ask the learner to choose an action and explain which cue drove the decision.

If the learner can only repeat the original sequence, the steps were learned.

If the learner can reason through the changed case, a more useful model may be forming.