AI Product Psychology · Finale

Sixteen Effects on One Table, Plus Fourteen Pre-Launch Questions

Every psychological effect in this chapter maps to an engineering switch you already learned. The finale lays them out in one table, then hands you a checklist you can paste straight into the launch process. Psychology isn’t mystical: it just turns “why users react this way” from guesswork into pattern.

Chapter table · Psychological effects × Engineering levers
EffectOne-line ruleEngineering leverRevisit
See what this table has in common? Nothing in the right column needs new tech: streaming is an API parameter, visible progress is an intermediate state in the Agent Loop, model routing and semantic cache are the same cast from the cost chapter, HITL and permission confirms showed up in security. Psychology didn’t give you new blocks—it gave you reasons to place the ones you have: when to flip these switches purely for feel. That’s the kind of rationale a PM can say out loud in design review and an engineer can recognize.
Fourteen pre-launch questions · Check them against your product

Hold up the AI product you’re building (or using) and self-audit—how many can you check? Not filling every box isn’t embarrassing; not knowing which ones you’re missing is.

0 / 14
Further reading · Whose shoulders this chapter stands on
Waiting
David Maister — The Psychology of Waiting Lines(1985)

The source paper on waiting-line psychology; its eight propositions still guide airports, banks, and loading animations.

Waiting
Jakob Nielsen — Response Times: The 3 Important Limits

Source of the 0.1s / 1s / 10s thresholds, proposed in 1993—still precise in the streaming era.

Labor illusion
Buell & Norton — The Labor Illusion(2011)

Harvard Business School travel-site experiment: waits that show labor beat instant results on satisfaction.

Peak-end
Kahneman et al. — When More Pain Is Preferred to Less(1993)

Cold-water experiment original: remembered experience is set by peak and end; duration barely matters.

Trust
Lee & See — Trust in Automation(2004)

Classic review of the trust-calibration framework: trust must match real system capability—too much or too little both cost.

Anthropomorphism
Reeves & Nass — The Media Equation(1996)

Source of the CASA paradigm: people automatically apply real social rules to media and computers.

Mental model
Don Norman — The Design of Everyday Things

Mental models, affordance, feedback: the foundation of interaction design—worth rereading every chapter in the AI era.

Service recovery
McCollough & Bharadwaj — Service Recovery Paradox(1992)

Well-recovered failure customers can be more loyal than no-failure ones: theoretical backing for AI error-recovery design.

Algorithm aversion
Dietvorst, Simmons & Massey — Algorithm Aversion(2015 / 2018)

See an algorithm err and abandon it—even when it’s more accurate than people; the 2018 follow-up shows tweak permission can reverse it.

Cognitive offloading
Sparrow, Liu & Wegner — Google Effects on Memory(2011)

Science original: people who expect information to be saved remember less—the starting point of cognitive-offloading research.

Paying
Prelec & Loewenstein — The Red and the Black(1998)

Coupling model of pain of paying and mental accounting: the tighter payment and consumption bind, the more it hurts—theoretical basis for flat-rate pricing.

✅ What this chapter wants to share