Errors

Errors

Most "accidents" attributed to human error are actually design errors. The user took the action the design invited; the action's consequence wasn't what the user intended. Reframing errors as a design property — rather than a user failing — opens the design space dramatically: errors can be prevented, made impossible, made obvious, or made recoverable. The right strategy depends on the kind of error.

Definition (in our own words)

An error is any action (or omission) that produces an unintended result. Errors come in two distinct kinds, and they need different design responses. Slips are skill-based failures: the user knew what to do but their action didn't match their intent — a typo, a wrong-button click, a slip of attention. Mistakes are knowledge-based failures: the user did exactly what they intended, but the intent was wrong because their model of the situation was wrong — they misread an alarm, misunderstood a diagram, made a wrong choice based on incomplete information. The two kinds happen for different reasons; they're prevented by different design moves.

Origins and research lineage

• Donald Norman, "Categorization of Action Slips" (Psychological Review, 1981, vol. 88) and The Design of Everyday Things (1988). Norman introduced the slip/mistake distinction to design vocabulary and argued that error is a property of design, not user.
• James Reason, Human Error (Cambridge University Press, 1990). The standard reference work in human-error research, originally from industrial-safety domains (aviation, nuclear, medical). Reason expanded Norman's framework with detailed taxonomies (skill-based / rule-based / knowledge-based errors) and the "Swiss cheese" model of cascading failures.
• Lidwell, Holden & Butler (2003) compactly distinguished slips (errors of execution, automatic/unconscious) from mistakes (errors of intention, conscious mental processes). The book identifies two slip subtypes (Action, Attention) and three mistake subtypes (Perception, Decision, Knowledge).
• Charles Perrow, Normal Accidents (1984). Argued that complex tightly-coupled systems produce errors as an emergent property of system design, not as failures of individual operators. Influential in industrial-safety and software-system design.
• Atul Gawande, The Checklist Manifesto (2009). Documented how checklists prevent slip-type errors in complex high-stakes domains (surgery, aviation).

Why the slip/mistake distinction matters

The two error types are caused by different mechanisms and prevented by different design moves:

A confirmation dialog can prevent slips ("Did you really mean to click delete?") but doesn't prevent mistakes (the user intended to delete; they just had the wrong file selected). Conversely, better explanations help with mistakes but don't help with slips (the user already knows what they meant).

Designs that treat all errors the same fail on at least one type.

The book's slip and mistake subtypes

Slip subtypes

Action slips

Wrong action despite correct intent. Examples:

• Typing the wrong word despite knowing the right one.
• Clicking the wrong button (intended Save, hit Submit).
• Hitting Reply All when meaning to Reply.

Design responses: confirmations for critical tasks; constraints; clear distinctive feedback; affordances and mappings that make wrong actions structurally harder.

Attention slips

Lapses in attention during procedures. Examples:

• Forgetting to attach a file before sending an email.
• Skipping a step in a multi-step process.
• Resuming an interrupted task at the wrong point.

Design responses: status cues that survive interruption; orientation aids when resuming; alarms for critical situations.

Mistake subtypes

Perception mistakes

Wrong action because of misread information. Examples:

• Nurse misreading a temperature curve as stable when it's trending up.
• Pilot misjudging altitude from a single-point readout.
• User misinterpreting an icon's meaning.

Design responses: clearer information presentation; trend / historical displays vs. point-in-time; reduce ambiguity in icons and labels.

Decision mistakes

Wrong choice under stress, bias, or overconfidence. Examples:

• Software engineer choosing a quick-fix that creates bigger problems.
• Medical professional anchoring on an initial diagnosis despite evidence.
• Designer ignoring user research that contradicts their intuition.

Design responses: decision trees, checklists, second-opinion mechanisms, training in error recovery.

Knowledge mistakes

Wrong action because of missing knowledge. Examples:

• New user not knowing how to use an unfamiliar tool.
• Operator unfamiliar with edge-case procedures.
• Visitor lost in a building they've never been in.

Design responses: memory aids, conventions and standards, training, simulations, accessible documentation, mnemonic devices.

When to apply

• Always, on every design that involves user actions with consequences.
• Especially on destructive, irreversible, or high-stakes flows.
• Critically in safety-relevant domains (medical, aviation, financial, security).
• In post-incident analysis — every "user error" report should be examined for the design that invited it.

Worked examples

Example 1: preventing a common slip (wrong-recipient email)

The "send to wrong person" slip happens when the user types a recipient's name and the autocomplete suggests someone with a similar name; the user accepts on autopilot.

Design responses:

• Distinct affordances — show the recipient's full name, photo, and email when adding to To/Cc/Bcc.
• Confirmation for high-stakes recipients — Gmail prompts when sending to a new external recipient: "You've never emailed person@external.com before. Are you sure?"
• Undo Send — covers slips after they happen.
• Reply-All caution — extra prompt when replying to large groups.

<!-- After clicking Send -->
<div class="undo-toast" role="status">
  Sending...
  <button onclick="cancelSend()">Undo</button>
</div>
<!-- After 5 seconds with no Undo, the message actually transmits. -->

Example 2: preventing an attention slip (forgot the attachment)

Common slip: writing "see attached" in an email, then sending without attaching the file.

Design response: Gmail and most modern email clients scan the message body for words like "attached" or "attachment"; if found and no attachment exists, prompt before sending: "You wrote 'see attached' but didn't attach a file. Send anyway?"

A simple heuristic that catches a common slip before it commits.

Example 3: preventing a perception mistake (misread chart)

A monitoring dashboard shows current CPU usage as a single number: "23%."

The user perceives "23%, fine" — but doesn't see that CPU has been climbing from 5% to 23% over the last hour. By the time they notice, the system is overloaded.

Design response: show trends, not just point-in-time values. A sparkline next to the number reveals the trajectory.

<div class="metric">
  <p class="label">CPU usage</p>
  <p class="value">23%</p>
  <canvas class="sparkline" data-trend="up"></canvas>
  <p class="status status--warning">Trending up over last hour</p>
</div>

The same number; the perception is now of a trend, not an instant.