Make the better choice the easier one.
A nudge changes the setting of a decision so the easy, automatic choice is the better one, while every option stays open. It works on habit rather than argument, which makes it effective, cheap, and easy to abuse.
Richard Thaler and Cass Sunstein made the idea famous in their 2008 book Nudge, though the cybernetician James Wilk had used the term earlier. A nudge, they wrote, changes behavior predictably without forbidding any option or significantly changing incentives, and it has to be easy and cheap to avoid. Putting fruit at eye level counts; banning junk food doesn’t.
Governments took it up quickly. The UK founded its Behavioural Insights Team, the original nudge unit, in 2010, and similar units followed in Australia, Canada, Germany, Japan, the World Bank, and the European Commission. The evidence is contested: a 2021 meta-analysis reported real effects, a 2022 reanalysis found no evidence for nudging after correcting for publication bias, and nudge units’ own trials showed effects far smaller than published studies.
A nudge alters the context of a choice so that fast, habitual thinking lands on the better option, without banning alternatives or changing incentives much. Defaults, placement, reminders, and showing what others do are the common levers.
Amsterdam’s Schiphol Airport etched an image of a fly into urinals to improve aim. It is the most cited nudge: a small change to the environment that works on automatic behavior without a single instruction.
In snack shops at Dutch train stations, moving fruit and healthy snacks next to the register increased their sales. In an interface, the equivalent is placing the recommended action where attention already is.
People look to others when unsure. Telling people how many neighbors pay on time, or which plan most customers choose, can move behavior, as long as it’s true and the norm is one worth following.
The Behavioural Insights Team sums up its approach as EAST: make it Easy, Attractive, Social, and Timely. Removing one step from a flow often does more than any amount of persuasion.
A reminder works when it arrives at a moment people can act on it. Suggest a backup when the device is charging, not in the middle of a task.
A study of trials run by two nudge units found effects about a sixth the size of those in academic journals, and a 2022 reanalysis found no overall evidence after correcting for publication bias. Defaults tend to hold up; many other nudges are weak.
Responses to the same nudge differ widely between people. Research suggests nudges tailored to individual differences work better than a single message for everyone, so segment and test.
Thaler calls friction that makes good choices harder sludge, like a cancellation maze. A nudge must be transparent and easy to avoid; when it serves the business against the user, it is a dark pattern.
AI can tailor a nudge to each person’s habits, and it needs nudges of its own to keep people checking its work.
AI can tailor prompts and offers to each person’s habits and moods, which may make nudges more effective and harder to notice. Keep them transparent, and give people a way to see why they were shown something.
Small prompts can counter overreliance on AI: a check-the-sources link, a confirmation before risky actions, highlighting the uncertain parts. Use nudges to slow people down where the AI is likely to be wrong.