011 min
What a choice screen decides before anyone chooses
Consider a product designer building the plan picker for a subscription app. The designer has to place the plans in some order. One plan has to be selected when the page loads, or none. The plans can be shown as one list or split into groups. Prices can be shown per month or per year.
None of these steps feels like a decision about what customers should buy. Each one is still a decision about it. A different order, a different pre-selected plan or a different price format leads to a different mix of purchases, with the same plans and the same prices.
Johnson and colleagues state this directly: there is no neutral architecture, because any way a choice is presented will influence how the decision-maker chooses. A designer cannot opt out of the influence. The only options are to choose the layout on purpose or to let it happen by accident.
021 min
Where choice architecture came from
The term was coined by the economist Richard Thaler and the legal scholar Cass Sunstein. They used it in their 2008 book Nudge. The idea behind it is older. Behavioural economists had already shown that people choose differently depending on how a decision is framed and what the starting option is.
The comparison the term borrows is from buildings. An architect decides where the doors, hallways, staircases and bathrooms go. That placement changes how people move through the building, even though no one is forced to take a particular route. A choice architect does the same for decisions.
The idea moved into government after the book appeared. The Royal Swedish Academy of Sciences wrote in its 2017 background document on Thaler's Nobel Prize that policymaking in several countries, in particular the USA and the UK, has been influenced by this approach. The same document explains the approach as the design of the environment where choices take place.
Thaler and Sunstein call the political position behind it libertarian paternalism. The "paternalism" part is that the designer steers people toward what is good for them. The "libertarian" part is that every option stays available. The Academy's document puts the second part this way: specifying a sensible default option does not mean people are forced to choose this option.
031 min
The tools of choice architecture
Johnson and eleven co-authors, a group of decision researchers from several universities, sorted the tools into two groups in 2012. The first group covers structuring the choice task, which is what to present. The second covers describing the choice options, which is how to present it.
| Tool | What the designer controls | Example from the research |
|---|---|---|
| Number of options | How many alternatives appear at once | Medicare drug plans, more than 100 in some US states |
| Defaults | What happens when the person does nothing | Automatic enrolment in a 401(k) retirement plan |
| Decision aids | Sorting, filtering, recommendations | Sorting alternatives by expected attractiveness |
| Partitioning | How options are grouped into categories | Savers splitting money evenly across the funds a plan lists |
| Attribute design | Which unit or scale a number uses | Miles per gallon against gallons per 100 miles |
| Evaluative labels | A good or bad anchor on a number | A 1 to 10 emissions rating on a car label |
Each tool works because people use shortcuts when a choice is large or unfamiliar. They keep what is already selected, spread effort evenly, or judge a number on the scale it arrives in. The tool changes the input to the shortcut, so it changes the output.
041 min
How many options to show
The first decision is how many alternatives to put in front of a person. Johnson and colleagues note that the number of Medicare drug benefit plans open to US seniors now exceeds 100 in some states. They suggest this may overwhelm the processing capacity of many elderly decision makers.
Too many options is one risk. Too few is another. The authors name the second one in their own words: too few options may generate context-specific preferences, where the presence or absence of one option influences what is chosen. This is why the right number is a balance and not a minimum.
For the balance, they give a starting point. One recommendation is that four or five non-dominating options may represent reasonable initial values for the choice architect. A non-dominating option is one that is better than the others on at least one thing that matters to someone, so it is worth showing. The authors add that a designer can start with the small set and offer a way to see more.
The research on too many options is covered in choice overload. The related finding that decision time grows with the number of choices is Hick's law. Choice architecture is wider than both, because the number of options is only one of its tools.
052 min
Defaults: what happens when the person does nothing
Johnson and colleagues call the default one of the most powerful and popular tools available to the choice architect. They define it plainly: defaults are settings or choices that apply to individuals who do not take active steps to change them. A pre-checked box on a form is a default. So is the plan selected on page load.
The best-known evidence is a study of one large US company by Brigitte Madrian and Dennis Shea, published as a working paper in 2001. The company moved its 401(k) retirement plan from sign-up on request to automatic enrolment. The economic terms of the plan did not change. Participation among employees with 3 to 15 months of tenure was 37% before the change and 86% after.
The mechanism, and the reasons people keep a default, are covered in the default effect. For choice architecture the question is a different one: how to choose the default.
Johnson and colleagues give a rule that is often prudent, though not a cure-all. Set the default to the alternative most people prefer when they make an active choice, without time pressure, in the absence of any default. They note that a small experiment can measure that preference quickly and cheaply.
A default is not the only way to handle a decision where people tend to stay put. The authors list forced choice as another policy. The product or service is withheld until the person makes an active choice. This fits a decision with no answer that suits most people, or one that people will resent being decided for them.
061 min
Grouping: how the same options look different in categories
Partitioning means splitting a set of options into groups. It sounds like a layout detail, but it changes how people spread their choices.
Johnson and colleagues describe a bias toward even allocation. When people divide a limited resource over a set of options, they tend to give each named group a similar share. In retirement saving, this appears as a tendency toward allocating 1/n of savings to each of the n options that are singled out in a 401(k) plan. A plan that lists one stock fund and four bond funds will therefore lead to more money in bonds than a plan that lists four stock funds and one bond fund, even for the same saver.
A designer can use this on purpose. Favoured options can sit in separate categories, and disfavoured ones can share a single category. The authors report that the effect varies by person. The influence is strongest on those who need the greatest guidance and weakest on those who need the least. In one study, wine novices spread their picks across whichever grouping the wines were shown in, and wine experts did so far less.
The same logic applies to navigation. A settings page with eight top-level categories and a settings page with three give each category a different weight. Related methods for pacing information are in progressive disclosure.
072 min
Describing attributes: the unit changes the judgment
The second group of tools changes how a fact about an option is described. The best-documented case is fuel economy, studied by Richard Larrick and Jack Soll and published in Science in 2008.
In the United States, fuel efficiency is quoted in miles per gallon (MPG). The amount of fuel used for a fixed distance is not a straight line in MPG. A car that gets 12 MPG uses 833 gallons over 10,000 miles. A car that gets 14 MPG uses 714 gallons, a saving of about 120 gallons. The saving from going from 28 MPG to 40 MPG over the same distance is smaller, about 107 gallons.
Larrick and Soll found that people reason as if the line were straight. People falsely believe that the amount of gas consumed decreases as a linear function of a car's MPG, they wrote, but the actual relationship is curvilinear. As a result, people underestimate the value of removing the most fuel-inefficient vehicles. In their first study, 77 college students ranked five pairs of old and new cars by how much each change would cut gas use. Sixty percent ranked them by the change in MPG, and 1% ranked them by actual gas saved.
The fix is a change of unit and no change to the car. Expressing efficiency as gallons per 100 miles (GPM) makes the saving equal to the difference in the numbers. Johnson and colleagues list this under the principle of linearity: describe a quantity on a scale where a bigger difference means a bigger real effect. Their other principles are parsimony, which means fewer attributes, comparability, which means the same scale across options, and evaluability, which means labels such as a 1 to 10 rating that tell a reader which end is good.
082 min
How to apply choice architecture in a product
Treat the layout of every decision as something that has an owner and a reason. These steps work for a designer, a product manager or an engineer building a form.
List the decisions.
For each screen where a person chooses, write down the order of the options, what is pre-selected, how options are grouped, and what unit each number uses.
Name the person's goal.
The layout should serve what the person would pick with time and information. If you cannot say what that is, you do not yet know which layout is the helpful one.
Set defaults from active-choice data.
Ask a sample of people to choose with no default, set the default to the most common answer, and keep every other option one step away. A short test is usually enough.
Limit options, and let people see more.
Start with four or five that do not duplicate each other. Offer a link to the full list.
Convert numbers to the scale of the person's real decision.
Show price per month if the bill arrives monthly. Show fuel use over a distance, not efficiency per fuel unit.
Test with the people you will serve.
Run the layout with a group that includes novices, experts and people with different priorities.
Here is another role-based example: a founder choosing which pricing plan to select by default should compare the plan that most trial users pick with the plan that earns the most. If the two differ, the choice is not a layout question. It is a decision about whether the product serves the customer or the quarter's revenue, and the team should settle it openly.
092 min
When choice architecture fails or goes wrong
No layout is neutral, but the effect varies by person. Johnson and colleagues describe a case where a nudge went wrong. A message that told households their energy use relative to neighbours led to an average 2% fall in use. The change depended on political affiliation: households identified as liberal reduced their use, and households identified as Republican increased theirs, in the study by Costa and Kahn from 2010. The same message was a nudge for one group and a different push for another.
The designer's own knowledge gets in the way. The authors point out a risk that choice architects may anchor first on what they themselves know or want. The design then fits the designer and not the user. This is a form of the curse of knowledge. The authors' conclusion is that the intuitions of choice architects will not always be enough and that choice architectures should be tested in diverse populations of interest.
A default can take away a real choice. Johnson and colleagues set the ethical line here: if defaults have an effect because consumers are not aware that they have choices, or because the transaction costs of changing from the default are too high, defaults impinge upon liberty. A pre-selected plan that takes five screens to change fails this test.
It does not replace the traditional tools. The authors note that changing prices, giving information and placing restrictions can all be ineffective on their own. Choice architecture works alongside those tools and does not replace them.
101 min
Choice architecture, nudges and the default effect
Three terms are often used as if they meant the same thing. Each answers a different question.
- Choice architecture is the design of the setting. It always exists, whether a designer made it on purpose or not. The question it answers is how the choice is presented.
- [Nudge theory](/concept/nudge-theory) is the idea that deliberate changes to that setting can steer behaviour while leaving every option open. A nudge is something a designer does to a choice architecture. The architecture is what the nudge changes.
- The default effect is one behaviour, the tendency to keep the pre-selected option. A default is one tool of the architecture, and the default effect is the reason it works.
A short test separates them. If the sentence is about the layout that exists, it is about choice architecture. If it is about a change made to steer people, it is about a nudge. If it is about people keeping what was selected, it is about the default effect.
?6 questions
Questions people ask
What is choice architecture?
Who coined the term choice architecture?
What is the difference between choice architecture and a nudge?
Can a choice be presented in a neutral way?
How many options should a choice architect show?
Is choice architecture manipulation?
§4 sources
Sources
Show all 4 sourcesHide sources
Johnson, E. J., Shu, S. B., Dellaert, B. G. C., Fox, C., Goldstein, D. G., Häubl, G., Larrick, R. P., Payne, J. W., Peters, E., Schkade, D., Wansink, B. and Weber, E. U. (2012). Beyond nudges: Tools of a choice architecture. Marketing Letters, 23(2), 487-504. DOI 10.1007/s11002-012-9186-1.
Larrick, R. P. and Soll, J. B. (2008). The MPG illusion. Science, 320(5883), 1593-1594. DOI 10.1126/science.1154983.
Madrian, B. C. and Shea, D. F. (2001). The power of suggestion: Inertia in 401(k) participation and savings behavior. NBER Working Paper 7682.
The Committee for the Prize in Economic Sciences in Memory of Alfred Nobel (2017). Richard H. Thaler: Integrating economics with psychology. Scientific background.





