Exam cheatsheet

L1 HCI, usability, rules

HCI (SIGCHI 1992): design, evaluation, implementation of interactive systems for human use + study of major phenomena around them. 4 reasons for studies: understand good interaction and why it works; design good UIs; evaluate UI quality; improve / invent interaction. Disciplines: human factors, psychology/cognition, CS, engineering, ergonomics, social science. Multi-disciplinary team: + knowledge, more ideas; − communication. Book: Sharp, Preece, Rogers, Interaction Design 5e.

Usability (ISO 9241-11:2018): specified users, goals, context → effectiveness, efficiency, satisfaction. UX (ISO 9241-210): perceptions and responses from use of a product, system or service. Utility = right functions. Usefulness = needs usability AND utility. Easy but missing features = usable, not useful. Powerful but hard = high utility, low usability, potentially useful. Usability ≠ looks; not only final-stage.

6 usability goals: effectiveness (does what it should); efficiency (productivity AFTER learning); safety (errors, recovery); utility; learnability (first encounter, self-exploration); memorability (return after time away). Cannot max all: trade-offs, prioritise, quantitative indicators. Shortcuts = efficiency, not learnability. Missing function = utility gap; no undo = safety.

Rule types: standards (high authority, low generality, e.g. ISO); guidelines (best practice, in-company consistency; iOS HIG, Material); principles (abstract, low authority); patterns (common practice, NOT necessarily good); anti-pattern (poor practice); dark pattern (tricks users, stakeholder over user value). Principles: visibility, consistency, affordances, feedback, constraints, mapping.

Shneiderman 8: consistency; shortcuts for frequent users; informative feedback; dialogs yield closure; error prevention + simple handling; easy reversal; internal locus of control; reduce STM load.

Nielsen 10: 1 visibility of system status; 2 match system/real world; 3 user control and freedom; 4 consistency and standards; 5 error prevention; 6 recognition not recall; 7 flexibility and efficiency; 8 aesthetic and minimalist; 9 help recognise, diagnose, recover from errors; 10 help and documentation. Overlaps: consistency, feedback/visibility, STM/recognition, shortcuts/flexibility, error prevention. "Closure" and "locus of control" = Shneiderman only; "minimalist" = Nielsen only.

L2 Perception, memory, attention

Vision: fovea = cones, central, sharp, ~5 to 18 deg, reading. Periphery = rods, blurred, movement and brightness change. Pop-out needs a feature contrast the periphery detects (many similar red items do not pop). Put errors where the user looks.

Gestalt (7): proximity (close = group); similarity (colour/shape/size); common fate (moves together); continuity (continuous forms); closure (fills gaps); symmetry (reduces complexity); figure/ground (pop-ups). Implication: visual hierarchy. Trap: each option is true of a different law, so read the law in the stem.

Sound: 20 Hz to 20 kHz, most sensitive 1 to 4 kHz. Intensity = loudness (dB), frequency = pitch (Hz), timbre = instrument. Whisper 15, talk 60, horn 110, concert 120+ dB. Alerts +: less visual attention, multitask, less screen info, hard to ignore. −: low resolution, absolute data hard, need 2 sounds for up/down, annoying. Phoneme; phonemic restoration (Warren 1970); sine-wave speech (Remez 1981). Artificial voices: accentuate intonation.

Haptics (ISO 9241-910:2011): skin, muscles, joints, tendons. Touch/tactile = skin; kinaesthesis = muscles/joints (proprioception, NOT touch). 3 modalities: discriminative touch (mechanoreceptors), temperature + pain (free nerve endings), kinesthetic. Fingertip acuity 1 to 2 mm (two-point threshold).

Working memory: central executive (decisions), visuo-spatial sketchpad (visual), articulatory loop (auditory); rehearsal or decay. Long-term: semantic (meaning), procedural (how-to: typing, driving), episodic/autobiographical (events), permastore (lifetime). Recognition (cue shown, easy) vs recall (no cue, hard). Prefer recognition: less LTM effort, less WM load, supports decisions and error recovery. Recall is NOT faster or more satisfying.

Attention: selective = Broadbent early selection; divided (multitasking) = Kahneman capacity allocation. Stroop effect. Response time = reaction + movement. Fitts: T = a + b log2(D/W + 1); D up = slower, W up = faster; a = start/stop, b = device speed. Big, near targets.

Emotion: Maslow (basic, psychological, self-fulfilment). Ortony et al. 2005: visceral (look, feel, sound), behavioural (use = usability), reflective (meaning, personal value, e.g. MoonSwatch). Affective computing; Galvactivator (skin conductance).

L3 Frameworks, process, PACT

Internal: info-processing model, mental models, gulfs. External: distributed cognition (propagation across representations), external cognition (tools). Model Human Processor: Card, Moran, Newell 1986. Gulf of execution: goal to system means ("What do I do now?"). Gulf of evaluation: effort to read system state ("What happened?", re-clicking Submit). Norman 7 stages: Exec 1 goal, 2 intention, 3 specify action, 4 execute; Eval 5 perceive, 6 interpret, 7 evaluate. Goal = stage 1 of execution.

Interaction design: goal-directed, creative problem-solving; Saffer 2010 "facilitating interactions between humans through products and services". UCD (Gould and Lewis 1985): early focus on users and tasks; empirical measurement; iterative design. Involve users: usable and used, ownership, expectations. Too much participation: conflict, rework. Inspections can inform revisions (sample key: iteration = real users).

Double Diamond: Discover, Define, Develop, Deliver. 4 basic activities: discover requirements, design alternatives, prototype, evaluate. Lifecycles: waterfall, spiral, agile (1 to 3 week timeboxes; XP, Scrum, Kanban, DSDM). Parallel tracks: vision first, design 1 iteration ahead. Docs: minimal, not none. Stakeholders = direct or indirect influence. Usability engineering = measurable criteria. A/B testing.

PACT: People, Activities, Contexts, Technologies. People: physical, psychological, language/culture, experience/personality, goals. Activity: temporal, cooperative, complexity, safety-critical, data and content. Context (3 only): organisational, social, physical. Tech: input (alphanumeric, numeric, indirect/direct pointing, speech, sensors, eye, brain); output (screen, multi-touch, printer, sound/earcons/TTS, haptics). Earcon = audible alert.

L4 Requirements, data gathering

6 requirement types: functional (what it does); data (type, volatility, size, persistence, accuracy, value); environment (physical, social, organisational, technical); user (profile: background, language, education, disability; novice/expert, frequent/casual); usability; UX. Story: As a <role>, I want <behaviour> so that <benefit>. MoSCoW: Must, Should, Could, Won't (this time). Others: Kano, 100 dollar test, cost of delay.

5 data-gathering issues: goals; whom; relationship; triangulation (more than one method or data type); pilot. Sampling: probabilistic = known chance (census, random, stratified); non-probabilistic = unknown (snowball, self-selected, volunteer). PIS + consent form.

Survey: attitudes, awareness, intent; NOT precise measures or usability problems. + many, cheap, non-intrusive; − shallow, no follow-up. Open: own words, rich qualitative, effort. Closed: ordered (ranking, Likert) / unordered; easier stats. Validated: SUS, CSUQ, QUIS, WAMMI.

Interview: unstructured (topic guide); semi (probe); fully structured (same Qs, same order, easy to compare, hard to follow up). Unfamiliar domain = unstructured/semi. Phases: intro, warm-up, main, cool-off, closure. No double-barrelled / double-negative Qs; break at least every 2 h. Prestige bias. Focus group 3 to 10, facilitator: + sensitive/diverse issues, consensus, interaction; − controversial topics, group-dependent, high effort.

Observation: direct vs indirect (diaries, logs, video). Silent; think aloud; constructive interaction (2 people). Frames: person/place/thing; Robson 2014: space, actors, activities, objects, acts, events, time, goals, feelings. Logging: unobtrusive, privacy. Diary: feedback vs elicitation; structured/unstructured. Issues: presence alters behaviour, overt/covert, recording, timing; not only labs.

L5 Analysis, tasks, personas

7 techniques: survey, interviews, focus groups, observation, contextual inquiry, ethnography, documentation. CI = interview + observation; 4 principles: context, partnership, interpretation, focus. Ethnography: immersion, full participant to full observer. Documentation: background/legislation, never alone. Choose by: focus, participants, nature of technique, resources. Combine: observation = context; interview = specific users; survey = reach; focus group = consensus.

Quantitative: numbers; mean, median, mode, SD; significance test compares means; Pearson correlation. Qualitative: words/images; themes, patterns, stories. Clean data (age 223); coding = to numeric. Inductive = bottom-up (themes, affinity diagram); deductive = top-down (preset categories). Critical incidents. Interaction analysis (video, inductive) vs content analysis (preferences). Excel, SPSS, R; nVivo, Atlas.ti.

Task analysis: HTA (logic of task) vs cognitive (cognitive walkthrough, GOMS). HTA: goal on top, 4 to 8 subtasks, plans = order/conditions (not tasks); outputs manuals, interface design. GOMS: Goals, Operators, Methods, Selection rules.

Use case: essential (user intention vs system responsibility, no implementation) vs normal + alternative courses. Diagram: ellipse = use case; stick figure = actor (person, org, external system); box = boundary; line = association. Persona: synthesised, typical not idealised; fixes elastic user. Types: primary, secondary, supplementary, customer, served, negative. Scenario: persona's task story: who, trigger, what happens, outcome. Courage and Baxter 2005.

L6 Models, interaction, prototypes

Conceptual model (Johnson and Henderson 2002): high-level how it is organised and operates. Metaphor jobs: conceptualise activity, instantiate model (desktop), visualise operation (cart). Choose (4): understand function; spot problems; generate; evaluate. Evaluate (5): structure? relevant? easy to represent? understood? extensible? + easy to learn, helps model, diverse users, innovation; − constrains design, limits understanding, poor designs, clashes with new models, breaks cultural rules (recycle bin on desktop).

5 interaction types: instructing (commands, repetitive); conversing (dialogue: phone, disability, children, elderly); manipulating (drag, zoom; affordances); exploring (VR, AR); responding (system initiates). Direct manipulation (3): continuous representation; rapid, reversible, incremental actions with feedback; physical actions not syntax. Interaction type ≠ interface type. Object/operation: nouns + attributes + verbs; exclude buttons, menus.

Sketch: quick, timely, inexpensive, disposable, plentiful. Storyboard = context of use, 3 to 7 scenes; wireflow = user view through a task; wireframe = one screen, structure. Lo-fi: early, no commitment (sketches, cards, Wizard of Oz). Hi-fi: final look/feel (tool, GUI builder). Paper: 3x5 cards, non-technical input, fast. Wizard of Oz = human fakes system; video prototype = techniques, not fidelity levels. Horizontal = breadth, little detail; vertical = depth, few functions (Hartson and Pyla 2019).

L7 Evaluation planning, inspection

Tools: clickthrough linked by hotspots; Balsamiq; POP (iPhone), Prott (Android). Physical: Arduino, micro:bit, MaKey MaKey; 3D print, laser cut, printed electronics. Why evaluate (4): features/function; effects on users; unforeseen problems; vs competitors. Where: usability lab (controlled, one aspect, equipment; unnatural, hard to generalise), naturalistic (realistic, long-term, UX; hard to run, uncontrolled), research lab, remote. When: iterative; pre-deployment acceptance; post ("permanent beta"). 7 steps: aims; methods; non-participant first; plan participant tests; recruit/setup; run; analyse/report.

Ethics (approval before study; RCOS storage). Before: pilot; system tested, not user; can stop any time; anonymise; informed consent, no coercion. During: one task at a time, early success, breaks, no displeasure, no outsiders. After: thank, answer questions without biasing, share media only with permission.

Heuristic evaluation: few experts check whole UI vs Nielsen 10 (from 249 problems); each alone, then merge; briefing, 1 to 2 h (2 passes), debrief. 5 evaluators = 75 to 80% (Nielsen and Mack 1994). + no users; − trivial/false alarms, costly experts, bias. Cognitive walkthrough: ease of learning, one task + scenario, Norman gulfs. Qs: 0 need for sub-task understood? 1 correct action obvious? 2 label understood? 3 result interpreted right? + no users, cheap; − labour-heavy. WCAG usable as heuristics.

Analytics: logs actions, unobtrusive, ethics if unaware; web, learning, visual. A/B: 2 versions, 2 user groups, large scale. Model-based: formulas predict performance without users (Fitts).

L8 Testing, experiments, field

Usability test: typical users, typical tasks; timed, logged, questionnaires. 5 roles: facilitator (in lab, puts user at ease); prototype executor (lo-fi only, silent); quantitative collector; qualitative collector; supporting actors. Quant: task time, errors per task, errors per time, help navigations. Qual: notes, critical incidents, think-aloud. Tasks: representative, frequent, critical; what, not how. 3 to 5 participants (~80%); typically 5 to 10.

Consent: PIS + PCF (sign before session; 2 copies); NDA if IP. Pre/post questionnaires: SUS, USE. Session 30 to 120 min. SUS: odd = r − 1; even = 5 − r; sum x 2.5 = /100; average 68.

Experiment: tests hypothesis (relationship of variables). IV = manipulated; DV = measured outcome. Between (different participants per condition; many; pre-test). Within (same participants, all conditions; counter-balance order effects). Matched pairs (match then randomly allocate). 7 steps: goals/hypothesis; design/variables; subjects; pilot; improve; run; accept/reject. Test: improve product, few people, usually not completely replicable. Experiment: new knowledge, many people, statistical, replicable.

Field study: natural settings; "in the wild"; minutes to years. Uses: opportunities, requirements, how to introduce tech, evaluate in use. Conundrums: informing vs consent, privacy at home, prototype breakdown/security. Field studies can still have IV/DV.

L9 Interface types, social

CLI: granular control, scripts, efficient at scale; steep learning, shell differences. GUI/WIMP (Windows, Icons, Menus, Pointing): intuitive, discoverable, no training; less granular, many clicks, does not readily support scripting. Buxton: WIMP = one eye, one finger, visual channel only. Post-WIMP (van Dam 1997): at least one technique beyond 2D widgets. Reality-based (Jacob 2008): naive physics, body, environmental, social awareness. Families: mobile, speech, surfaces, tangible (objects control AND represent data), VR, wearable, BCI (brain waves), social. Speech = interface type.

Conversation: Sacks 1978, 3 rules: 1 speaker picks next; 2 another self-selects; 3 speaker continues. Cues: length ("three things"), interrupt (gesture, "Excuse me"), end (orientation, raised voice). Breakdowns must be detected to repair. Remote: no non-verbal cues; telepresence robot = attend a place; social presence = awareness of virtual others. Turkle 2015. The WELL = early community.

Co-presence: coordination (talk, schedules/rules/agenda, shared representations) vs awareness (peripheral: overhear/oversee; situational: ATC, theatre). Groupware, CSCW: time/place matrix. Social engagement: voluntary, unpaid exchange.

L10 Colour, legibility, situations

Colour: performance, affect, attention. 2.2 billion visual impairment; colour-blind ~8% men, 0.5% women (Birch 2012). Do: black and white first, add colour sparingly; brightness/outline + redundant cues; no red-green; user control; consistent. Tog: colour + secondary cue; do not drop colour. Laws: ADA, EAA, AODA, WCAG. Tools: Coblis, WebAIM. Beneficiaries: older users, low-res screens, sunlight.

Contrast = luminance difference (different hues can still fail). Blue: ~2% cones, focus falls in front of retina, poor acuity worsening with age: avoid dark/saturated blue text and thin blue lines. Mid-spectrum (green, yellow) seen best; reds look closer than blues; warm looks larger; blooming (light on dark looks larger). NASA: max 6 colours, cultural conventions, consistent, redundant, not on small elements. Black foreground: cyan, magenta, white. WCAG ratios not in lecture extract.

Situational impairments (Sears and Young 2003): cold, noise, stress, light. Encumbrance (0/0.5/1.0 kg, walking): heavier + walking worst (exact % unreliable). WICE = shoulder fatigue (UIST 2025). Voice assistants: perceived older voice raised empathy, trust, satisfaction; humour/apology raised empathy.