tools-only HCI Theory Reference
The theoretical foundations and predictive models behind why interfaces work—or don't.
Foundational Frameworks
Norman's Seven Stages of Action
Every interaction follows this cycle:
- Goal — forming the intention
- Plan — deciding on action sequence
- Specify — translating plan to physical actions
- Perform — executing the action
- Perceive — sensing the state of the world
- Interpret — making sense of perception
- Compare — evaluating outcome against goal
Gulf of Execution: Gap between user's intentions and available actions Gulf of Evaluation: Gap between system state and user's perception of it
Design implication: Minimize both gulfs through clear affordances and feedback.
Affordances and Signifiers (Gibson → Norman)
- Affordance: What an object allows you to do (exists whether perceived or not)
- Perceived Affordance: What the user thinks they can do
- Signifier: What communicates the affordance
Common confusion: A button's affordance is "pressability." The signifier is its raised appearance, shadow, or label.
Design implication: Affordances are designed; signifiers must be visible.
Mental Models
- User's mental model: How the user thinks the system works
- Designer's model: How the designer thinks the system works
- System image: What the system actually presents
When these diverge, errors occur. The system image is the only communication channel.
Design implication: Design the system image to align user and designer models.
Predictive Laws
Fitts's Law
Movement time to a target:
MT = a + b × log₂(2D/W)
Where D = distance, W = target width
Implications:
- Larger targets are faster to hit
- Closer targets are faster to reach
- Corners and edges are infinitely wide (screen bounds)
- Why menus at screen edges are fast (Mac menu bar)
- Why pie menus outperform linear menus
Hick-Hyman Law
Decision time increases with choices:
RT = a + b × log₂(n + 1)
Where n = number of alternatives
Implications:
- More choices = slower decisions
- But only for unpredictable choices
- Expert users bypass through memory
- Organize options to reduce effective choices
- Progressive disclosure limits cognitive load
Steering Law
Time to navigate through a constrained path:
T = a + b × (A/W)
Where A = path length, W = path width
Implications:
- Cascading menus are slow (narrow paths)
- Wider channels allow faster movement
- Explains difficulty of hover menus
- Why tunnel interfaces feel laborious
Power Law of Practice
Performance improves with repetition:
T = a × N^(-b)
Where N = number of trials
Implications:
- Novice and expert performance differ dramatically
- Design for learnability, not just initial use
- Shortcuts reward repeated use
- Muscle memory is real and valuable
Cognitive Foundations in HCI
Working Memory (Miller, Baddeley)
- Capacity: 7±2 chunks (Miller) or ~4 chunks (Cowan)
- Duration: ~20 seconds without rehearsal
- Easily disrupted by interference
Implications:
- Don't require users to remember across screens
- Chunk information meaningfully
- Keep related information visible together
- Minimize interruptions during complex tasks
Attention (Treisman, Kahneman)
- Preattentive processing: Color, size, orientation, motion detected instantly
- Selective attention: Limited capacity, can only focus on one complex task
- Change blindness: Unattended changes go unnoticed
- Inattentional blindness: Unexpected objects missed entirely
Implications:
- Use preattentive features for critical information
- Don't rely on users noticing changes
- Animation draws attention (use sparingly)
- Important changes need explicit signaling
Recognition vs. Recall (Tulving)
- Recognition: Identifying something seen before (easier)
- Recall: Retrieving from memory without cues (harder)
Implications:
- Menus > command lines for novices
- Show options rather than requiring memory
- Consistent locations leverage recognition
- Icons + labels beat icons alone
Cognitive Load (Sweller)
- Intrinsic load: Complexity inherent to the task
- Extraneous load: Complexity added by poor design
- Germane load: Effort toward learning/schema building
Implications:
- Minimize extraneous load (that's design's job)
- Scaffold intrinsic load for complex tasks
- Don't simplify away essential complexity
- Reduce load during critical operations
Error Theory
Slips vs. Mistakes (Reason, Norman)
- Slips: Right intention, wrong action (execution failure)
- Capture errors (habit takes over)
- Description errors (similar objects confused)
- Mode errors (wrong system state assumed)
- Mistakes: Wrong intention (planning failure)
- Knowledge-based mistakes
- Rule-based mistakes
Implications:
- Slips need physical constraints and undo
- Mistakes need better feedback and mental model alignment
- Different error types need different solutions
Swiss Cheese Model (Reason)
Accidents happen when holes in multiple defensive layers align.
Implications:
- Single safeguards are insufficient
- Confirmation dialogs are one (weak) layer
- Undo is another layer
- Constraints prevent some errors entirely
- Multiple independent safeguards for critical actions
Interaction Paradigms
Direct Manipulation (Shneiderman)
- Continuous representation of objects of interest
- Physical actions instead of complex syntax
- Rapid, incremental, reversible actions
- Immediate, visible feedback
Why it works: Reduces gulf of execution, exploits spatial reasoning, enables exploration.
Limitations: Not everything has physical metaphor; poor for abstract operations.
Instrumental Interaction (Beaudouin-Lafon)
- Domain objects: What users work with
- Instruments: Tools that act on objects
- Reification: Making abstract concepts concrete
- Polymorphism: Instruments working on multiple object types
Implications:
- Clear separation of tools and materials
- Instruments should be visible and persistent
- Reify actions into manipulable objects (e.g., selections, styles)
Gulf-Bridging Approaches
| Strategy | Execution Gulf | Evaluation Gulf |
|---|---|---|
| Affordances | ✓ | |
| Constraints | ✓ | |
| Mappings | ✓ | ✓ |
| Feedback | ✓ | |
| Visibility | ✓ | ✓ |
| Consistency | ✓ | ✓ |
User Research Methods
When to Use What
Formative (early design):
- Contextual inquiry: Observe users in their environment
- Card sorting: Understand mental categories
- Participatory design: Co-create with users
- Think-aloud protocols: Expose reasoning
Summative (evaluation):
- Usability testing: Task-based observation
- A/B testing: Comparative performance
- Heuristic evaluation: Expert inspection
- Cognitive walkthrough: Step-through analysis
Quantitative Measures
- Effectiveness: Task completion rate
- Efficiency: Time on task, errors, learnability
- Satisfaction: Subjective ratings (SUS, NASA-TLX)
The ISO 9241 definition of usability: effectiveness, efficiency, satisfaction in context.
Sample Size Considerations
- Usability testing: 5 users find ~85% of problems (Nielsen)
- Quantitative studies: Power analysis required
- A/B tests: Effect size determines sample needs
- Qualitative saturation: When new themes stop emerging
Research Validity
- Internal validity: Did X cause Y?
- External validity: Does it generalize?
- Ecological validity: Does it reflect real use?
- Construct validity: Are you measuring what you think?
Lab studies sacrifice ecological validity for control. Field studies do the opposite.
Evaluation Heuristics
Nielsen's 10 Heuristics
- Visibility of system status
- Match between system and real world
- User control and freedom
- Consistency and standards
- Error prevention
- Recognition rather than recall
- Flexibility and efficiency of use
- Aesthetic and minimalist design
- Help users recognize, diagnose, recover from errors
- Help and documentation
Shneiderman's 8 Golden Rules
- Strive for consistency
- Cater to universal usability
- Offer informative feedback
- Design dialogs to yield closure
- Prevent errors
- Permit easy reversal of actions
- Support internal locus of control
- Reduce short-term memory load
Applying Heuristics
- Structured inspection with severity ratings
- Multiple evaluators find more issues
- Explain violations in terms of principles
- Distinguish opinion from evidence-based criticism
Theoretical Tensions
Novice vs. Expert
- Novices need recognition, feedback, guidance
- Experts need efficiency, shortcuts, power
- Design challenge: Serve both without compromise
- Solutions: Progressive disclosure, accelerators, adaptive interfaces
Simplicity vs. Power
- "Make it simple" vs. "make it capable"
- Removing features reduces capability
- Adding features increases complexity
- Resolution: Simplicity in common paths, power available but not required
Consistency vs. Context
- Consistency enables transfer of learning
- Context may demand different solutions
- Internal consistency > external consistency
- Breaking consistency requires strong justification
Automation vs. Control
- Automation reduces effort but also agency
- Users must maintain mental models of automated systems
- "Out of the loop" problem in automation
- Appropriate levels of automation depend on consequence and skill
Seminal Contributions
Key Researchers and Ideas
- Don Norman: Affordances, emotional design, design of everyday things
- Ben Shneiderman: Direct manipulation, information visualization
- Stuart Card: GOMS, information foraging, Fitts's Law applications
- Jef Raskin: Humane interface, modes considered harmful
- Alan Kay: Dynabook, object-oriented interfaces
- Douglas Engelbart: Mouse, hypertext, augmenting human intellect
- Ted Nelson: Hypertext, Xanadu
- Bill Buxton: Sketching, input devices, design process
Foundational Papers
- "The Psychopathology of Everyday Things" (Norman)
- "Direct Manipulation Interfaces" (Shneiderman)
- "The Psychology of Human-Computer Interaction" (Card, Moran, Newell)
- "Instrumental Interaction" (Beaudouin-Lafon)
- "As We May Think" (Bush)
- "Augmenting Human Intellect" (Engelbart)
- "A Feature Integration Theory of Attention" (Treisman)
Methodological Notes
Levels of Analysis (Marr)
- Computational: What problem is being solved?
- Algorithmic: What process solves it?
- Implementational: How is it physically realized?
Different questions, different methods.
Converging Evidence
No single method is definitive:
- Behavioral experiments: What people do
- Neuroimaging: What brains do
- Computational models: How it might work
- Lesion studies: What breaks when damaged
Triangulation strengthens conclusions.
Ecological Validity
Lab findings may not generalize to real world:
- Simplified stimuli
- Artificial tasks
- Measured awareness
- Motivated participants
Design implication: Supplement lab findings with field observation.
The Academic Lens
What Theory Provides
- Explanatory power (why does this work/fail?)
- Predictive power (what will happen if...?)
- Generative power (what should we try?)
- Evaluative criteria (how do we judge success?)
What Theory Doesn't Replace
- User research with real users
- Iteration and testing
- Domain expertise
- Craft and intuition
Bridging Research and Practice
- Research reveals principles
- Practice applies and tests them
- Practical problems inspire research
- Research findings inform design
- The cycle should be continuous