There is no neutral way to present a choice.
How options are presented, from which one is preselected to the order they appear in, shapes what people choose. Some design is unavoidable, so design it on purpose and in people’s interest.
Richard Thaler and Cass Sunstein coined the term in their 2008 book Nudge. Anyone who arranges options, from a cafeteria manager to a form designer, is a choice architect, they argued, and can steer people toward better choices without forbidding any option. They called the philosophy libertarian paternalism.
With John Balz they listed the tools: defaults, expecting error, understanding mappings, giving feedback, structuring complex choices, and creating incentives. Other researchers split the elements into those that structure the set of choices, such as the number of options and the default, and those that describe them, such as how attributes are grouped and labeled.
Choice architecture is the design of how options are presented: how many there are, how they are ordered and described, and which one is selected by default. Because no presentation is neutral, every interface that offers a choice is already steering it. The question is only in which direction.
Most people keep whatever is preselected. In one study, organ donor registration was twice as high where people had to opt out as where they had to opt in. Pick the default that serves the user, not just the business.
When no default suits most people, require an active choice: nothing is preselected and people must pick before continuing. It costs a moment, but it avoids silently giving many people the wrong thing.
Each extra option takes time to evaluate, and too many can stall a decision. Either limit the alternatives or add decision aids such as filters, sorting, comparisons, and recommendations.
Search results, rankings, and recommendations shape what people pick, because the first and most prominent options get chosen more. Rank by what serves the user, not by what earns the most.
Convert numbers into terms people care about: gallons per 100 miles instead of miles per gallon, the total cost instead of the monthly payment. Evaluative labels such as good or poor make numbers easier to compare.
People tend to spread resources evenly across whatever categories are offered, such as the funds in a retirement plan. The number and grouping of options changes the outcome, so group with intent.
People favor the present over the future. Thaler and Benartzi’s Save More Tomorrow plan had employees commit in advance to save part of future raises, which raised savings without asking for sacrifice today.
The same tools that help people can be turned against them. When choice architecture serves the business at the user’s expense, it becomes a dark pattern. Ask whether the design points where users would go if they had time to think.
AI now sets the defaults and ranks the options: suggested replies, autofill, recommended plans, the single answer. That is choice architecture at scale.
Suggested replies, smart autofill, and predicted settings are defaults chosen by a model for each person. Make them easy to change, and show the alternatives rather than hiding them.
An assistant that returns a single recommendation removes the choice set entirely. For consequential decisions, show a few options with their tradeoffs instead of one pick.