Behavioural Sciencethe minimum vocabulary every designer needs

Behavioural science is fashionable in design, more cited than understood. When a designer says “we should reduce the cognitive load here,” they might be reaching for a real model of working memory, or just gesturing at a feeling. That difference is what decides whether behavioural science is a Heart tool or just decoration.

Four frames carry most of the work.

• Fast and slow thinking (System 1 and System 2): Kahneman’s dual-process model splits fast, automatic thinking (System 1) from slow, effortful thinking (System 2).¹ Most of what people do, including most product use, runs on System 1: you recognise a pattern and act, the way a familiar checkout slides past on autopilot. System 2 wakes up when the pattern breaks, when something is unfamiliar, high-stakes, or not where you expected it. The question for you is which system your interface asks a person to work in, and whether that suits the moment they are in. A consent flow built for System 1 fails. So does an expense form that demands System 2 for every field.
• Loss aversion and framing: From prospect theory: people feel a loss about twice as hard as they feel an equal gain.² The frame you put on a choice, gain or loss, changes what people do without changing a single fact. “Save 20%” and “Don’t lose 20%” pull differently. Loss-framing is fair when it matches what the person actually wants, like showing them what they risk by skipping a security step. It turns manipulative when it uses that asymmetry against them, which is the engine behind most dark patterns.
• The Fogg Behavior Model: A behaviour happens only when motivation, ability, and a prompt arrive at the same moment.³ When it is not happening, one of the three is missing: the person does not see enough reason (the benefit is unclear or the cost too high), cannot manage the step right now, or never got a prompt at the moment it would have helped. The usual mistake is to assume the problem is motivation and pile on banners, reminders, and incentives, when really the step is too hard or the timing is wrong.
• Mental models: Norman’s point: a person runs on their own model of how a system works, and it usually differs both from how the system really works and from what the designer imagines they think.⁴ Most usability failures live in that gap. Someone who believes the app saves automatically will not save by hand. Someone who believes it does not will save compulsively. Someone who has no idea will behave in ways you cannot predict. Designing well means knowing which model the person brings, then confirming it, correcting it, or designing around it. Research is how you find that out. Synthetic personas, model-summarised behaviour, and AI-generated “user research” turn up in a lot of recent tooling. Whatever the label, they are pattern-matching on training data that under-represents the long tail of how real people behave. Lean on them instead of real contact and you get models that are confident, and wrong about exactly the people at the edges. This vocabulary is how you tell what a model claims about people from what an observation actually shows.