Cognitive Foundations

The science of how minds work, and what that means for design.

When to Use This Skill

• Explaining why a design works or fails (grounded in research, not opinion)
• Evaluating cognitive load or working memory demands
• Predicting user performance (Fitts, Hick-Hyman)
• Diagnosing mental model misalignment
• Justifying design decisions to stakeholders with evidence
• Understanding attention, perception, or memory failures

Output Contracts

For Single-Principle Analysis

## Cognitive Principle: [Name]

**Principle**: [1-sentence explanation]

**Evidence in Design**: [Where/how this applies]

**Implication**: [Specific, actionable recommendation]

**Confidence**: [High/Medium/Low] — [rationale]

For Cognitive Audit (Comprehensive)

## Cognitive Audit: [Screen/Flow Name]

### Working Memory Load
- Items requiring recall: [count]
- Cross-screen memory demands: [Y/N]
- Verdict: [Acceptable / High / Overloaded]

### Attention Demands
- Preattentive features for critical info: [Y/N]
- Competing attention demands: [list]
- Change blindness risk: [areas where changes may go unnoticed]

### Mental Model Alignment
- Expected user model: [what users likely think]
- System behavior: [what actually happens]
- Gap: [mismatch, if any]

### Predictive Laws
- Fitts's Law concerns: [target size/distance issues]
- Hick's Law concerns: [choice overload areas]

### Gulf Analysis
- Gulf of Execution: [unclear how to act?]
- Gulf of Evaluation: [unclear what happened?]

### Violations of Nielsen's Heuristics
| Heuristic | Violation | Severity |
|-----------|-----------|----------|
| ... | ... | 1-4 |

### Recommendations
1. [Highest priority fix]
2. [Second priority]
3. [Third priority]

For Explaining a Failure

## Failure Analysis: [What Went Wrong]

**Observed Behavior**: [What users did]

**Cognitive Explanation**: [Which principle explains this]

**Root Cause**: [Design element that caused it]

**Fix**: [Specific change]

---

Quick Reference: Predictive Laws

Law	Formula	Rule of Thumb
Fitts's Law	MT = a + b × log₂(2D/W)	Bigger + closer = faster. Screen edges are infinite.
Hick-Hyman	RT = a + b × log₂(n+1)	More choices = slower. Reduce or organize options.
Steering Law	T = a + b × (A/W)	Narrow paths are slow. Cascading menus are hard.
Power Law	T = a × N^(-b)	Practice helps. Design for learnability.

---

Quick Reference: Nielsen's 10 Heuristics

#	Heuristic	Quick Test
1	Visibility of system status	Can user always tell what's happening?
2	Match system ↔ real world	Language familiar? Metaphors sensible?
3	User control and freedom	Easy undo? Clear exits?
4	Consistency and standards	Same words/actions mean same things?
5	Error prevention	Constraints prevent errors before they occur?
6	Recognition over recall	Options visible? No memory required?
7	Flexibility and efficiency	Shortcuts for experts?
8	Aesthetic and minimalist	Only relevant info? No clutter?
9	Error recovery	Errors explained in plain language with fix?
10	Help and documentation	Searchable, task-focused, concise?

---

Quick Reference: Working Memory

• Capacity: ~4 chunks (not 7)
• Duration: ~20 seconds without rehearsal
• Test: Count items user must hold in mind across screens/steps

Red flags:

• "Remember this code and enter it on the next page"
• Multi-step forms without visible progress/state
• Complex comparisons requiring mental tracking

---

Quick Reference: Preattentive Features

Detected in <200ms, no focused attention required:

• Color (hue, saturation)
• Size (length, area)
• Orientation (angle)
• Motion (flicker, direction)
• Shape (curvature, enclosure)

Use for: Critical info, errors, changes, status Don't use for: Everything (loses signal value)

---

Cognitive Load Checklist

Quick assessment for any interface:

Factor	Low Load	High Load
Choices visible	2-4 options	10+ options
Memory demands	Recognition	Recall
Steps to goal	1-3 clicks	5+ clicks
Interruptions	None	Frequent modals
Novel elements	Familiar patterns	New conventions
Error recovery	Clear undo	Destructive actions
Visual complexity	Clean, grouped	Dense, undifferentiated

Scoring: Each "High Load" = +1. Score >3 = redesign needed.

---

Common Violations → Principle

Symptom	Likely Violation	Fix
Users don't notice changes	Change blindness	Animate, highlight transitions
Users can't find the button	Poor Fitts's Law	Increase size, reduce distance
Users freeze at options	Hick's Law overload	Reduce choices, progressive disclosure
Users forget mid-task	Working memory exceeded	Show state, don't require recall
Users misunderstand state	Gulf of Evaluation	Better feedback, visibility
Users click wrong thing	Poor affordance/signifier	Clearer visual treatment
Users make same error repeatedly	Mode error	Visible mode indicators
Users abandon complex forms	Cognitive load	Chunk, scaffold, save progress

---

Process

1. Identify cognitive demands — What is the interface asking the user to perceive, remember, decide, or do?
2. Match to principles — Which cognitive constraints or laws apply?
3. Evaluate alignment — Does the design respect or violate these?
4. Recommend changes — Specific modifications grounded in the principle

---

Deep Reference Files

For comprehensive principles and research:

• PSYCHOLOGY.md — Perception, memory, attention, biases, emotion, motivation
• HCI-THEORY.md — Norman's model, predictive laws, error theory, research methods, heuristics

Primary Sources

• A Feature-Integration Theory of Attention.md — Treisman & Gelade on preattentive processing (informs: Quick Reference: Preattentive Features)
• Judgment under Uncertainty- Heuristics and Biases.md — Kahneman & Tversky on cognitive biases (informs: PSYCHOLOGY.md § Decision Making)

---

Key Researchers

• Don Norman: Affordances, gulfs, emotional design
• Daniel Kahneman: Dual process theory, heuristics and biases
• Stuart Card: GOMS, information foraging, Fitts's Law
• Anne Treisman: Feature integration, preattentive processing
• Jakob Nielsen: Usability heuristics, discount usability
• Ben Shneiderman: Direct manipulation, golden rules

---

Remember

• Cognitive science explains why design principles work
• Individual differences exist—design for variability, not averages
• Lab findings may not generalize (ecological validity matters)
• Theory informs but doesn't replace observing real users
• When in doubt, reduce cognitive load—users have less capacity than you think