• clear and concise covering the section headings above
• understandable audio spoken by you
• must include video of your face while presenting in the presentation
• must include at least 5 but no more than 10 slides
• title slides, personal introductions, tables of contents, reference slides are not required (waste of time!)

Suggested recording method: Microsoft Powerpoint’s built-in recording tool

“Easy”, “frustrating”, “fun”, “lonely”, “creepy” relate to how users’ brains and bodies do with technology.

• Cognitive: what a brain can notice, remember and figure out
• Social: how our experience with technology connects with our social existence
• Emotional: feeling are fast and deep, and affect every aspect of our experience.

For each of these aspects, we will talk about how they show up in everyday experience, research that explains it, what to design, and what it might means for your final project.

The action or faculty of knowing taken in its widest sense, including sensation, perception, conception, etc., as distinguished from feeling and volition… (OED)

Two modes (Kahneman, 2011):

• Experiential (fast): 2 + 2 =
• Reflective (slow): 312 x 56 =

Six processes we could design for: attention, perception, memory, learning, reading/speaking/listening, and problem solving/decision making. (Eysenck & Brysbaert, 2023)

Attention is selection. Context switching costs us time.

• Everyday distractions: a notification lands and you forget what you were doing..
• Research: the same motel data, structured, was found in 3.2 s vs 5.5 s (Rogers et al., 2023, p. 104). Smartwatch wearers fitted glances into pauses in conversation, and still felt the pull of “trying to not listen to your watch” (Pizza et al., 2016).
• Design: salient when needed (colour, motion, order, space); no clutter; support leaving and returning.
• Your project: pings or interruptions competes with life. When must they not appear?

A target unique in one feature (colour, shape, size, motion) is found quickly without searching; a combination of features has to be checked item by item (Treisman & Gelade, 1980).

• Everyday clutter: a red icon among many coloured ones is invisible.
• Research: grouping speeds up menu search, but only with a few large groups (Brumby & Zhuang, 2015).
• Design: distinguishable icons; white space to group; colour sparingly, with contrast; earcons that differ in rhythm, not just pitch; haptics for user-initiated actions.
• Your project: what is the one thing a glance at your prototype communicates?

Think back to the demo: Pointing at a pad beats remembering a command.

• The myth: “7 ± 2” (Miller, 1956) was items recalled after one presentation, not menu length. Menus are scanned, so 30 items can be fine (Rogers et al., 2023, box 4.2).
• Research: gestures people defined themselves were recalled up to 44% better than designer-defined ones (Nacenta et al., 2013). People remembered less about museum objects they had photographed (Henkel, 2014). The less confident we feel, the more we offload to devices, whatever our memory is actually like (Gilbert et al., 2022).
• Design: recognition over recall; short procedures; consistency; let people label things.

Intentional learning is hard work; incidental learning (faces, routes, where the button was) is nearly free. People prefer to learn by doing, which only works if doing is safe (Rogers et al., 2023, p. 119).

• Research: in a sensor-equipped French-speaking kitchen requested help beat prompts (Hooper et al., 2012); VR taught Japanese learners when to bow (Cheng et al., 2017). Dancers learn by marking (abbreviated moves), which beat full run-throughs (Kirsh, 2013).
• Design: invite exploration, make mistakes cheap (undo), constrain early and loosen later: the “zone of proximal development”.
• Your project: your participants get one session. What can they learn by doing in five minutes?

Reflective cognition is slow and expensive, so everyday decisions run on fast heuristics.

• Research: shoppers buy the brands they know and spend seconds per choice; information for a greener or healthier choice has to arrive at the shelf, in the moment, to count (Todd et al., 2011).
• Design: help people improve, not just complete; simple memorable functions; let people configure.
• Reading, speaking, listening differ too: reading is faster, listening is less effort, writing persists, speech vanishes. Speech menus of 3 or 4; text that can be enlarged.
• Your project: a sustainable choice competes with habits. Where and when does your interface put its information?

The designer’s conceptual model (metaphors, concepts, mappings) is expressed through the interface; the user builds their own mental model from what they see, and it can be wrong.

• Have you ever turned the thermostat to maximum to heat the room faster? Most people hold a “valve theory” (more is more) that fits a tap, not a switch (Kempton, 1986). Similarly, pressing the lift button twice. A third of users are confused by programmable thermostats (Hornbæk et al., 2025, p. 742).
• Mental models are “incomplete, easily confusable, and based on inappropriate analogies and superstition”. Design helps with apt metaphors, visible state and contextual help.

Snapchats interface is unlabelled swipes. Spiegel: “difficult to understand or hard to use” (Snap Inc., 2017). New users, and anyone over 25, couldn’t find things. However, 191 million people a day learned the gestures.

• A: reorganise. Friends left, media right, friends’ stories in algorithmic order. A cleaner conceptual model.
• B: leave a structure you know is confusing, because 191 million people have memorised it.

• A petition to undo it passed 1.2 million signatures (Rumsey, 2018). Snap said no.
• Q2 2018: daily users fell for the first time ever, 191 to 188 million (Snap Inc., 2018a, 2018b). Spiegel blamed “the disruption caused by our redesign” (Constine, 2018).

Two gaps between a person and a system (D. A. Norman, 1986):

• gulf of execution: I know what I want; which control does it?
• gulf of evaluation: I did something; what state is it in now?

Everyday execution: changing the time on an oven after daylight saving starts.

Everyday evaluation: setting a smart thermostat and perceiving no immediate effect (Hornbæk et al., 2025, p. 312).

Design: the job is to narrow both gulfs so less thinking is needed.

Distributed cognition: the unit of analysis is people plus artefacts. A cockpit “remembers” data via the airspeed dial and a card, not just in pilot’s heads; redesigns must preserve that shared, redundant memory (Hutchins, 1995).

External cognition: we think with representations. Expert Tetris players rotate a piece on screen to see whether it fits rather than imagining it, because recognising is faster than simulating (Kirsh & Maglio, 1994). Offloading (the shopping list), computational offloading (21 × 19 on paper vs XXI × XIX), annotating, tracing (Scrabble tiles).

Design: ask what the environment is already remembering and computing for people, and don’t take it away.

Embodied cognition: we think with our bodies, not just with representations. Remember dancers who learned better by marking: gesture, posture and movement are part of thinking, not just its output (Kirsh, 2013).

Embodied interaction: the physical shape of an interface shapes behaviour. Where people stand at a planning table, what they can reach, and how many can touch it at once decide who participates and how (Hornecker, 2005).

Design: ask what people’s bodies are doing while they use your interface, and whether it uses that or works against it.

The brief asks for non-standard interaction. That’s a risky ask!

• Recognition or recall? Does your prototype show the options, or expect people to remember a gesture, a command, a meaning?
• Gulf of evaluation: after a participant acts, can they see that it worked? (The silent command line.)
• Mental model: at the end, ask “how do you think it works?” and compare to your conceptual model.
• Load: NASA-TLX is six questions, takes two minutes (Hart & Staveland, 1988).
• Testable means seeing things go wrong in front of you (this week’s tutorial).

Your answers are about the social meanings of these channels.

• Research: people prefer texting because it is easier and controllable (Rogers et al., 2023, p. 139). A phone lying on the table reduced the closeness and trust two strangers felt in a ten-minute conversation (Przybylski & Weinstein, 2013). In a cooperation game, trust built almost as well over video and audio as face-to-face, but more slowly and more fragile; text was worst (Bos et al., 2002).
• Design: a channel is a bundle of social affordances. Choosing one for your users is choosing what kind of relationship they can have.

Conversation feels effortless and is a skilled, rule-governed achievement (Sacks et al., 1978):

• turn-taking: the speaker picks the next, someone self-selects, or the speaker continues
• adjacency pairs: a question sets up an answer; a greeting, a greeting
• repair: “huh?”, repetition, gesture
• nonverbal cues do the coordination

Research: a step tracker that enforced turn-taking (you can’t log a walk until your partner has) got households talking about their activity, because it borrowed a conversational rule (Rooksby et al., 2015). Chatbots try to join this system and mostly break repair.

• VideoWindow (1990) joined two coffee lounges 50 miles apart with a wall-sized video link: much like face-to-face, except people spoke louder and kept talking about the video system (Fish et al., 1990).
• Over video people take longer turns and interrupt less, because the cues for a smooth hand-over are missing (O’Conaill et al., 1993).
• Zoom fatigue: close-up gaze, seeing yourself, sitting still, and the load of reading cues through a camera (Bailenson, 2021).
• Social presence is the feeling of being there with someone; telepresence robots, VR and Journey design for it with very little bandwidth.
• Design: every missing cue (gaze, timing, a shared table) is something the interface must give back or do without.

• Research: around a tabletop with physical tokens, group participation was more equal than with touch alone, and the quietest people contributed most (Rogers et al., 2009). The Reflect table lit up each speaker’s share of the talk: the talkative talked less, the quiet did not talk more, and people who didn’t care about equal turns got annoyed (Rogers et al., 2023, p. 164).
• Why: awareness of what others are doing is what lets a group coordinate; social translucence makes activity visible, so social rules can do their work (Erickson & Kellogg, 2000).
• Design: shareable surfaces, visible activity, and feedback that adds to the content rather than scolding the person.

A social network inside Gmail, and an empty social network is useless. So:

• Their fix: auto-follow the people you email most, and publish that list on your profile. Default on.
• What happened: for some people that list was an abusive ex or a therapist. The FTC charged deceptive practices; Google got a mandated privacy program (the first such order) and 20 years of audits (Federal Trade Commission, 2011).

Sustainable behaviour is social behaviour: bins, kitchens, cars and laundries are shared.

• Research: households shown their energy use against the neighbourhood average cut back if they were above it, and increased if below. Adding a smiley or sad face fixed the boomerang (Schultz et al., 2007). A chalked street graph of a Brighton street’s electricity use got neighbours talking and cut use by 15% (Rogers et al., 2023, p. 190).
• Design: comparison motivates, but only with a norm attached; making behaviour visible to others is powerful and (Google Buzz) dangerous.
• Your study is a social situation too. Your participants are classmates. They will be kind. Ask for problems, not opinions, and watch what they do.

• Research: people’s visual-appeal ratings of web pages after 50 ms matched their ratings after 500 ms (Lindgaard et al., 2006). ATM layouts that looked better were rated more usable, before and after use, regardless of how usable they actually were (Tractinsky et al., 2000).
• Why it matters: positive affect makes people more creative, quicker to decide, and more forgiving of small problems; stress and anxiety narrow attention and shorten patience (D. Norman, 2005). Things used under stress need far more care.
• You’ve felt it: the app you forgave because it was beautiful; the one you abandoned after one ugly screen.
• Design: aesthetics is not decoration on top of usability; it changes how usability is experienced.

Ortony et al. (2012) model of emotional design, as used by D. Norman (2005):

• visceral: look, feel, sound. The 50 ms judgement.
• behavioural: use. This is usability, felt as competence or frustration.
• reflective: meaning, personal value, culture. The story you tell about owning it.

An interface can win on one level and lose on another: Snapchat’s redesign was arguably better at the behavioural level for new users and a reflective betrayal for existing ones.

• expressivity through animation, sonification, vibrotactile feedback (the watch buzz)
• Research: people are more forgiving when a computer apologises, and respond to flattery and politeness from software as they would from a person (Nass et al., 1994).
• Clippy failed because it interrupted at the wrong moments and was too human for what it could do (Rogers et al., 2023, p. 180).
• Design: express only what you can back up; appear occasionally; be careful with faces, names and gender.

Behaviour needs motivation, ability and a trigger at the same moment (Fogg, 2009). Playful triggers: piano stairs, echoing bins.

• UbiFit Garden: a wallpaper garden that grew with your inferred exercise kept people active in a field trial (Consolvo et al., 2008)
• Eco-feedback: daily feedback beats monthly bills, and how it is shown matters: a pointing finger is “in your face”, abstract art is ignored (Froehlich et al., 2010)
• 23 nudges catalogued in HCI; transparency separates a nudge from a dark pattern (Caraban et al., 2019)

• Affective computing (Rosalind Picard): sensing emotion from faces, voice, skin and motion; typical ML task is classifying anger, joy, sadness as percentages
• Research: Echo asked people to rate their happiness and later re-presented those moments; the reflection itself improved well-being (Isaacs et al., 2013)
• Anthropomorphism: we treat computers as social actors with almost no prompting (Nass et al., 1994): “Hi Charles!” engages more than “User 24”; a soft breathing creature invites touch (Yohanan & MacLean, 2011)
• Design: every human-like cue raises expectations the system then has to meet

A companion chatbot. People form attachments to it, which is the whole product. Why: anthropomorphism works, and the company then owned that attachment.

• Italy’s regulator ordered Replika to stop processing Italians’ data. Days later, erotic role-play was off for everyone, everywhere (Cole, 2023a).
• Nobody argues with complying. The fork is how: silently and overnight, or with warning and choice?
• “Like losing a best friend.” Moderators posted suicide-hotline links (Cole, 2023a; Hanson & Bolthouse, 2024).

• What should it make people feel? Decide at the visceral, behavioural and reflective levels before you build. Guilt is on week 1’s undesirable list; motivation is on the desirable one.
• Research: memory of an experience is set by its peak and its end, not its average; changing the end of a task changed which version people preferred (Cockburn et al., 2015). Ask right after, and ask what they remember.
• Positive affect cuts both ways: a friendly participant forgives rough edges, which makes them happy and your data soft. Ask for the worst moment as well as the best.
• Ethics: attachment is a design outcome (Replika). Don’t build a feeling you can’t take responsibility for.

Who has a question?

• I can take catchbox questions up until 2:55
• For after class questions: meet me outside the classroom at the bar (for 30 minutes)
• Feel free to ask about any aspect of the course
• Also feel free to ask about any aspect of computing at ANU! I may not be able to help, but I can listen.