The Human and the Computer

Foundations of HCI: understanding human information-processing capabilities (vision, hearing, touch, memory, reasoning, emotion) and computer characteristics (devices, memory, processing, networks) in order to design usable, effective interactive systems. The central idea (the "golden rule of design") is: understand your materials — both the machine and the person.

---

The Human

We must understand the human as an information-processing system: stimuli are received through the senses, held and processed in memory, and acted upon through reasoning and motor response. Emotion and individual differences modulate every stage.

Vision

• Two stages of visual processing:
  1. Physical reception of the stimulus — light reflects from objects, is focused (upside-down) on the retina. The retina contains rods (low-light / monochrome vision) and cones (colour vision); ganglion cells in the brain detect pattern and movement.
  2. Processing and interpretation of the stimulus — the visual system resolves ambiguity using context.
• Visual angle indicates how much of the view an object occupies (a function of its size and distance from the eye). Visual acuity (ability to perceive detail) is limited and increases with luminance; flicker perception also improves with luminance.
• Perceived size constancy: familiar objects are perceived as constant size despite changes in visual angle with distance. Cues such as overlapping help perception of size and depth.
• Colour = hue + intensity + saturation. Cones are sensitive to colour wavelengths, but blue acuity is lowest. About 8% of males and 1% of females are colour-blind → never use blue for important detail, and never rely on colour alone (combine with shape/text).
• The visual system compensates automatically for movement and changes in luminance. Optical illusions (Ponzo, Müller-Lyer) arise from over-compensation; context resolves ambiguity.
• Reading: involves saccades (rapid eye jumps) and fixations (where perception occurs). Word shape matters for recognition. Negative contrast (light text on dark background) improves reading from a computer screen.

Hearing

• Physical apparatus: outer ear (protects and amplifies) → middle ear (transmits sound as vibrations) → inner ear (chemical transmitters released → impulses in auditory nerve).
• Sound parameters: pitch (frequency), loudness (amplitude), timbre (quality/type).
• Humans hear roughly 20 Hz–15 kHz, and are less accurate at distinguishing high frequencies than low.
• The auditory system filters sounds to attend to one over background noise — the cocktail party phenomenon.

Touch and Movement

• Skin receptors: thermoreceptors (heat/cold), nociceptors (pain), mechanoreceptors (pressure — some instant, some continuous). Some body areas (e.g. fingers) are far more sensitive.
• Kinesthesia = awareness of body position; affects comfort and performance.
• Reaction time (response to stimulus) + movement time (depends on age/fitness):

  Stimulus	Approx. reaction time
  Visual	~200 ms
  Auditory	~150 ms
  Pain	~700 ms

  • Increasing reaction time decreases accuracy for unskilled operators but not for skilled ones (skill absorbs the cost).
• Fitts' Law — time to hit a screen target:

  Mt = a + b · log₂(D / S + 1)

  where Mt = movement time, D = distance to target, S = size of target, and a, b are empirically determined constants. Design implication: make targets as large as possible and distances as small as possible (large buttons, short travel).

Memory

Three functional memory systems; selection of stimuli is governed by level of arousal (attention).

Type	Capacity	Access	Decay
Sensory (iconic/echoic/haptic)	Buffer for stimuli	—	Continuously overwritten
Short-term / working (STM)	7 ± 2 chunks	~70 ms	~200 ms
Long-term (LTM)	Huge / unlimited	~0.1 s	Very slow (if at all)

• Sensory memory: buffers — iconic (visual), echoic (aural), haptic (tactile). Examples: the "sparkler" trail, stereo sound. Continuously overwritten.
• STM / working memory: a scratch-pad for temporary recall (e.g. a phone number). Capacity 7 ± 2 chunks — a chunk can be a digit, a word, or a meaningful group. This is why menu/option groupings matter (and why Miller's 7±2 is about STM, not menu size).
• LTM: repository for all knowledge. Two types:
  • Episodic — serial memory of events.
  • Semantic — structured memory of facts, concepts, skills (semantic is derived from episodic).
• LTM models of structure:
  • Semantic networks — nodes with inheritance (child nodes inherit parent properties); explicit relationships; inference through inheritance.
  • Frames — data structures with slots: fixed (legs: 4 for DOG), default (diet: carnivorous), variable (size/colour). Type–subtype relationships.
  • Scripts — stereotypical information to interpret a situation (entry conditions, results, props, roles, scenes, tracks), e.g. a visit to the vet.
  • Production rules — condition → action (IF dog wagging tail THEN pat dog; IF growling THEN run away).
• Storage: rehearsal moves info STM→LTM; total-time hypothesis (retention ∝ rehearsal time); distribution-of-practice effect (spread learning over time); structure/meaning/familiarity aid recall.
• Forgetting: decay (gradual, very slow) and interference — retroactive (new replaces old) and proactive (old interferes with new). Retrieval: recall (reproduce, cue-assisted) vs recognition (knowing it was seen — simpler, because the item itself is the cue). Emotion can make us "choose" to forget.

Reasoning and Problem Solving

• Deductive reasoning: derive the logically necessary conclusion from given premises. Conclusion may be logically valid yet factually false (premises can be wrong).
• Inductive reasoning: generalise from seen cases to unseen. Unreliable (can only be proven false, never proven true) — but useful. Humans are poor at using negative evidence.
• Abductive reasoning: from event to cause (e.g. "Sam drives fast when drunk → he must be drunk"). Unreliable (false explanations possible). These three — deduction, induction, abduction — are the classic modes of human inference.
• Problem-space theory: problem solving = generating states via legal operators within a problem space; heuristics (e.g. means-ends analysis) select operators. Operates within human information-processing limits (e.g. STM). Best applied to well-defined (puzzle-like) domains.
• Analogy / analogical mapping: use knowledge of a similar problem from a similar domain; difficult when domains are semantically different.
• Skill acquisition: characterised by chunking — grouping information to optimise STM; conceptual rather than superficial grouping; information structured more effectively.

Errors and Mental Models

• Slips: right intention, wrong execution (poor physical skill, inattention, change to skilled behaviour). → Better interface design fixes these.
• Mistakes: wrong intention (incorrect understanding). Humans build mental models to explain behaviour; if the model is wrong (differs from the actual system) errors occur. → Requires better user understanding of the system.
• Design must anticipate both: make correct actions easy and obvious, and make the system's true behaviour legible.

Emotion

• Theories: James–Lange (emotion = interpretation of physiological response), Cannon (emotional psychological response to stimulus), Schacter–Singer (emotion = evaluation of physiological response in light of the whole situation). Emotion involves both cognitive and physical responses; the biological response is called affect.
• Affect influences response: positive → creative problem solving; negative → narrow thinking. Implications: stress increases problem-solving difficulty; relaxed users are more forgiving; aesthetically pleasing/rewarding interfaces increase positive affect.

Individual Differences

• Long-term: sex, physical and intellectual abilities.
• Short-term: stress, fatigue.
• Changing: age.
• Ask: will this design decision exclude some section of the user population? (see Universal Design, Ch6).

---

The Computer

A computer system is made of elements — input devices, output devices, virtual reality, physical interaction, paper, memory, processing, networks — each of which affects the interaction. To understand HCI, you must understand the computer.

Input / Output Devices

• Typical system: screen/monitor, keyboard, mouse/trackpad; variations: desktop, laptop, PDA. The devices dictate the supported interaction style.
• Handwriting recognition: natural, but technical problems (capturing stroke path/pressure, segmenting joined writing, interpreting letters, coping with styles). Used in PDAs/tablets.
• Speech recognition: improving rapidly; most successful for single-user (trained) or limited-vocabulary systems; problems with noise, imprecision, large vocabularies, different speakers.
• Eye gaze: control interface by looking (e.g. look at a menu item to select). Uses low-power laser reflected off retina; mainly used for evaluation; potential for hands-free control; high accuracy needs a headset.
• 3D displays: desktop VR (perspective + motion), stereoscopic vision, VR helmets, shuttered specs.
• 3D interaction: cockpit/virtual controls, the 3D mouse (six degrees: x,y,z + roll,pitch,yaw), data gloves (fibre optics detect finger position), whole-body tracking (accelerometers / reflective dots + video).
• VR motion sickness: caused by conflicting cues — time delay (head moves, display lags); depth perception conflict (headset gives different stereo distance, but all focused in same plane) → conflict between eye angle and focus ⇒ sickness. Drives technology improvement.
• Simulators and VR caves: scenes projected on walls, realistic environments, hydraulic controls, other people.
• Physical controls, sound, touch: beeps/bongs/clonks for errors and confirmations; haptic devices (vibration, force feedback) in games and simulation (e.g. feel of surgical instruments). Texture/smell/taste technology is very limited.
• Environment and bio-sensing: ubiquitous sensors (door switches, ultrasound, RFID); body sensors — iris scanners, body temperature, heart rate, galvanic skin response (GSR), blink rate.
• Fonts: serif vs sans-serif; fixed-pitch (Courier) vs variable-pitch (Times). OCR converts bitmaps back to text (harder with varied fonts and page layout).
• Paper-based interaction: usually output, but also input (OCR, scanning, glyphs, annotated paper with special pens).

Memory and Processing

• RAM ≈ 100 ns access (volatile); disks (magnetic/optical) for persistent storage; flash for portable/non-volatile.
• Finite processing speed causes problems: cursor overshoot (buffered keypresses), "icon wars" (user clicks icon, nothing happens, clicks another, then everything responds and windows fly). Too fast is also bad — help screens may scroll too fast to read.
• Performance bottlenecks:
  • Computation-bound — computation takes ages, frustrating.
  • Storage-channel-bound — data transfer disk→memory.
  • Graphics-bound — updating displays; sometimes helped by a graphics co-processor.
  • Network capacity — shared resources, but slow networks reduce interactive performance.

Networked Computing

• Networks give access to large memory/processing, other people (groupware, email), and shared resources (especially the web).
• Issues: network delays (slow feedback), conflicts (many people update data), unpredictability.

---

Psychology and Design — Distilled

The psychology above is distilled into actionable knowledge throughout the course:

• Guidelines and standards (Ch4) — e.g. blue acuity ⇒ don't use blue for detail.
• Cognitive models (Ch6/Ch8) — GOMS, HTA, KLM predict performance.
• Evaluation techniques (Ch5) — experimental and analytic methods to test designs.
• Golden rule: understand your materials — both computers and people. A design that ignores human capabilities (limited STM, Fitts' law, colour blindness, error-proneness) will fail no matter how powerful the machine.

Human Capabilities — Design Cheat Sheet

A compact reference linking human limits to concrete design consequences:

Human capability	Key limit	Design consequence
Vision	blue acuity lowest; ~8% M / 1% F colour-blind	never rely on colour alone; pair with shape/text
Vision	acuity ↑ with luminance	ensure adequate contrast/brightness
Hearing	20 Hz–15 kHz; poorer at high freq	don't depend on high-frequency cues
Reaction	visual ~200 ms, auditory ~150 ms	give prompt feedback; avoid waiting
Movement	Fitts' law Mt = a + b·log₂(D/S+1)	large targets, short distances
STM	7 ± 2 chunks, ~200 ms decay	group options; don't overload screens
LTM	huge, slow access	support recognition over recall
Reasoning	induction/abduction unreliable	confirm, don't assume inference
Errors	slips vs mistakes	prevent (slips) + educate (mistakes)
Emotion	positive→creative, negative→narrow	aesthetically pleasing, low-stress UIs
Individual diff.	age, ability, fatigue	universal/inclusive design (Ch6)

This table is the practical bridge from psychology to the guidelines, models, and evaluation methods covered in later chapters.