Head in the Cloud

We live in a Dunning–Kruger world. That the public is unaware of its own ignorance is a fact of life, one that needs to be taken into account by designers, marketers, and communicators. Consider the hamburger icon. You’ve seen it: it’s a stack of three rounded bars (“patties”), like a Big Mac assembly schematic from McDonald’s Hamburger University. The hamburger is an icon for summoning a menu or navigation bar. Invented as a response to the cramped screens of early smartphones, the hamburger has become ubiquitous and has migrated to desktop screens. It’s a global standard—except that, for many, the hamburger is “mystery meat,” says Web designer Eric W. Mobley. Novice users may not even recognize the hamburger as an icon. Its vaguely classical symmetry suggests a decorative bullet.

The hamburger is part of an evolving visual language that fills our real and virtual worlds. This language has its roots in twentieth-century design-school utopianism. Clean modern graphics would replace words in the multicultural, visually orientated world to come.

Exemplifying this ethos, the American Institute of Graphic Arts (AIGA) and the US Department of Transportation released a set of icons designed for airports and train stations in 1974. They include the familiar no-smoking glyph and the stylized men’s-room and women’s-room symbols that have perfectly circular heads. Many AIGA icons are models of lucidity, as long as you understand the concept of synecdoche (not a coat hanger but a cloakroom; not a martini but a bar).

Simultaneously with the AIGA effort, Xerox PARC engineers were developing the first computer interface using icons rather than typed-in commands. Xerox designers were influenced by ergonomic studies suggesting that users relate better to pictures than to words, an idea that would catch on commercially in the 1980s with the Apple Macintosh.

In today’s world of small screens, there is less room for elaborate icons, and app designers favour simple, sometimes cryptic, pictographs. This philosophy assumes that users will experiment to learn how icons work.

I reproduced several standard icons in a survey to see how many people could recognize them. I expected that the hamburger would throw a few people, and it did. I didn’t expect how poor recognition would be for some of the airport icons.

A mere five percent could correctly identify the AIGA symbol for “exit,” and less than four percent identified the dollar sign within a circle as the sign for “cashier.” The other multiple-choice options for that icon were “automatic teller machine,” “bank,” “currency exchange,” and “expensive.” The most popular guess was “currency exchange” (chosen by thirty-five percent).

My sample also failed to recognize icons that Web designers assume to be all but universally understood. A quarter of respondents didn’t understand the Wi-Fi symbol. The hamburger eluded nearly half. (As usual, everyone in my sample was a computer and Internet user. In that regard, they would have been somewhat more digital-savvy than the population as a whole.)

The magnifying-glass icon for “search” also scored poorly (fifty-two percent recognition). This is one icon that depends on a literary allusion, to Arthur Conan Doyle’s Sherlock Holmes. Notwithstanding an ongoing movie and TV franchise, many smartphone users may not be familiar with the character. Another problem is that a magnifying-glass icon is also used in graphics programs to indicate “zoom in” and “zoom out.” This ambiguity was a big factor here, for “magnify” was the second-most-popular choice (thirty-six percent).

The least recognized of all the icons was the ellipsis—the three dots meaning “click here for more options.” Barely a quarter were able to pick the right meaning. In practice, the ellipsis is often interchangeable with the hamburger. At least the hamburger looks something like a list of menu items. The ellipsis metaphor refers to punctuation—not always a strong suit of digital natives.

Icons are tools to help users do what they want to do. When people don’t understand them, it’s the designers, not the users, who deserve a flunking grade. Designers need to know what users don’t know (even if users are unaware of this themselves) and factor that into their designs. App makers have more remedies than those creating real-world signage. An app can know whether the user is a newbie or a power user. It can also know the user’s age. My survey showed, as expected, that young people are far more likely than old people to recognize icons. An app can label an icon with text for users likely to need it, then dispense with the label after that user gains experience.

Food and drug companies should also heed the public’s cluelessness. A 2006 study in Annals of Internal Medicine asked adult patients to describe how they would take five common prescription medicines based on the labels. Serious errors were typical, and they were often of the Dunning–Kruger variety. Those who misunderstood the labels thought they understood them perfectly. The errors mostly came down to two things: botching simple maths and mixing up simple units of measurement.

Asked to explain the instruction, “Take one teaspoonful by mouth three times daily,” many replaced “teaspoon” with “tablespoon.” This mix-up accounted for just over half of all the errors.

Even among those who were able to recite the label instruction “Take two tablets by mouth twice daily” word for word, one-third did not measure out the correct number of pills they were supposed to take per day (four).

I did surveys testing the public’s understanding of labels, including this nutrition label for ice cream. A large majority knew that “g” means “grams” and “Sat Fat” stands for “saturated fat”—the bad kind. But many had trouble applying label information. When the question required maths or drawing a logical inference, scores went south.

One of my questions asked, “If you ate the entire container, how many calories would that be?” Sixteen percent gave a wrong answer, and most of the wrong answers were way off. Another question: “Suppose you’re on a low-carb diet that allows 15g of carbohydrates for a snack. How much of this ice cream could you have?” One in three respondents gave a wrong answer. “Pretend that you’re allergic to penicillin, latex gloves, peanuts, and bee stings. Is it safe to eat this ice cream?” The answer is “no,” as the label lists peanut oil as one of the ingredients. Eleven percent got it wrong.

I asked respondents about a medicine label containing the common warning, “Avoid alcohol.” This laconic instruction means, “Don’t drink any alcoholic beverage when taking this medicine.” But eleven percent felt that “avoid alcohol” permits wiggle room. They wrongly interpreted it to mean, “It’s okay to have a drink or two, as long as you aren’t driving or operating machinery,” or, “Don’t drink to excess while taking this medicine” or admitted that they didn’t know what the instruction meant.

Lists of side effects are another minefield of confusion. In the name of protecting consumers, US law requires drug companies to report any “temporally associated symptoms” that turned up in testing a new medicine. That means anything that happened to anyone while taking the drug or shortly thereafter. Side effects must be listed even if there was an equal or greater number of reports in the control group that got a placebo instead of the drug.

The law thus throws correlation, causation, and common sense under the bus. There is a rationale for this. The law properly recognizes that a lot of money is at stake in testing new drugs. Doctors and drug companies have financial incentives to downplay side effects, and disclosing every potential side effect is a step towards transparency.

But transparency has side effects of its own. The disclosures are worthwhile only to the extent that average people can use them to make better decisions. Unfortunately human decision-making falters when confronted with small and hard-to-gauge risks of terrible outcomes. There will be people who won’t take an antidepressant that could help them because suicide is listed as a side effect. To make any kind of an informed decision you would need to know the probabilities. Is the chance of suicide one in ten…or one in one hundred million? Is it actually less likely with the drug than without because the suicide rate was higher in the group getting a placebo? The label doesn’t give that information.

Package inserts helped fuel the anti-vaccination movement. Anti-vaxxers tell parents to look at the measles, mumps, and rubella (MMR) vaccine insert, which does indeed list autism among the possible side effects, but not for the reason the crusaders claim. The recommended age for the vaccines happens to coincide with the age at which autism is most often diagnosed. That makes autism a “temporally associated symptom” that has nothing to do with causality. Not on the label is the fact that measles still kills about 145,000 people a year, nearly all of them in developing nations where children aren’t vaccinated.

“If it takes 5 machines 5 minutes to make 5 widgets, how long would it take 100 machines to make 100 widgets?”

This brain-teaser appears in the Cognitive Reflection Test that marketing professor Shane Frederick, then of MIT, published in 2005. Frederick found that most people, including students at prestigious universities, get it wrong. It’s not a hard question in any normal sense. The difficulty with this question and the others on Frederick’s test is that they suggest an impulsive answer. The question’s parallel wording all but ensures that “100 minutes” pops to mind. Many go with that first impulse…which in this case is wrong. The correct answer is five minutes. A machine takes five minutes to make a widget. More machines mean more widgets—but it still takes five minutes to make one.

Many medicine-label instructions are confusing in the same way. Tell someone to “take two tablets twice a day,” then ask, “How many pills should you take a day?” The impulse answer is “two,” and that’s the usual wrong answer on surveys.

Frederick’s research holds out little hope that better education is the solution. MIT students get an excellent education that includes two semesters of calculus…but many of them botched the widgets arithmetic. The only realistic remedy is to avoid medicine instructions that pose cognitive difficulties. This can be accomplished by following a couple of simple guidelines:

• Minimize the need to do even simple arithmetic. Whenever practical, a dose should be a single pill.

• Avoid ambiguity. In Britain, the “avoid alcohol” language has been replaced with “Do not drink alcohol while taking this medicine.” This leaves no room for rationalizing.

For a good example of the Dunning–Kruger effect, look at practical knowledge of cooking and household management. All of us have a little knowledge of these things, and those who are truly misinformed tend not to know how misinformed they are. I asked a survey sample how long it takes to hard-boil a large egg. More than two-thirds failed to give an answer that was even in the ballpark (nine to thirteen minutes).

Now, of course the answer depends on whether you put the egg in cold or already boiling water. Either way, it takes about eleven minutes at boiling heat to cook an egg yolk all the way through. Someone who puts the egg in cold water and counts the time it takes to boil the water could come up with a figure like twenty minutes. But a third of the sample believed they could cook a hard-boiled egg in six minutes or less.

Another question asked how long it takes to grill each side of a one-inch-thick steak until it’s medium done. Five minutes would be a reasonable answer. Again the majority gave unreasonable answers. With steaks as well as eggs, more undershot rather than overshot the reasonable answer—expectations shaped by extensive experience with a microwave oven, perhaps?

Name the ingredients in bread. This came with eleven options, among them flour, water, and yeast—without which it would be a challenge to make anything resembling bread. Eight percent thought they could make bread without flour; an equal number omitted yeast, and twenty-six percent left out water.

Everybody’s heard that you should only eat oysters in a month with an R in it. Or is it a month without an R? Nearly half said they didn’t know, and only thirty percent picked the correct answer, with an R. The rule of thumb skips the four warm months May through August, in which shellfish may accumulate toxins from red tides.

It’s a well-to-do crowd that eats oysters. The income difference between those who knew and didn’t know this rule was $24,000 (£16,000).

We’re told that British and American cooks resist going metric because they don’t know a gram from a millilitre. Guess what? They don’t know a teaspoon from a tablespoon, either.

I asked a nationwide US sample how many teaspoons there are in a tablespoon. Barely half (forty-nine percent) gave the correct answer, three.

Asked how many tablespoons are in a fluid ounce, only twenty-four percent got it right (two tablespoons = one ounce), and answers ranged from one to sixteen. For the record, these right answers only apply in the US and UK. In Australia four teaspoons make a tablespoon, and a tablespoon is two-thirds of a fluid ounce.

We all know “bad” cooks who swear they follow recipes to the letter. I wonder how many culinary failures can be traced back to the measuring-spoon confusion.

How big is a two-by-four? I asked a survey sample whether that familiar lumber size was actually two by four inches, smaller than two by four inches, or larger than two by four inches. Anyone who’s handy knows that it’s thinner and narrower than claimed, 1.5 by 3.5 inches.

Forty-three percent of the sample took the question to be a case of, “Who’s buried in Grant’s Tomb?” They said a two-by-four was just that. Only thirty-eight percent said “smaller,” and fourteen percent said “larger.”

The sample did better with questions about tools that would be used in light indoor repairs. Shown a picture of a screw, ninety-one percent could identify the type of screwdriver needed (Phillips head). Sixty-eight percent could identify an adjustable wrench from a picture.

“About how often should you change the oil in your car?” This survey question was multiple choice, offering answers ranging from “every 500 miles” to “every 200,000 miles.” By far the most popular answer was “every 3,000 miles,” picked by thirty-eight percent.

It used to be a rule of thumb that oil should be changed every three thousand miles. Around that time, it was also a rule of thumb that every computer needs two floppy-disk drives. In the past thirty years, motor oils and motors have improved, greatly increasing the recommended service interval. Most car manufacturers recommend oil changes at intervals of 7,500 or 10,000 miles (or else that you defer to on-board monitoring systems that calculate the recommended interval based on usage). One of the most popular motor oils, Mobil 1 Extended Performance, is guaranteed for 15,000 miles.

If the survey results are any indication of behaviour, consumers are changing their oil two or three times as much as they need to. Not only is that billions of dollars down the drain, it’s also tonnes of oil down the drain, placing a needless burden on waterways and wildlife.

Ignorance doesn’t just happen. Sometimes it’s manufactured. That the three-thousand-miles credo is a profitable scam is an open secret in the industry. The trade publication National Oil and Lube News brims with advice about persuading customers to change their oil early and often. “Many people…know when to have their oil changed but don’t pay that much attention to it. Take advantage of that by using a window sticker system [and] customers will be making their way back to you in a few short months.” In other words, even the minority that knows to change the oil every ten thousand miles may not remember when the last change was. They defer to a window-sticker or e-mail reminders.

David Langness, a former automotive service adviser, called the three-thousand-mile rule “a marketing tactic that dealers use to get you into the service bay on a regular basis. Unless you go to the drag strip on weekends, you don’t need it.”

It’s not hard to learn the truth about oil changes. It’s in your car’s manual. Otherwise, Google, “How often should I change my car’s oil?” You’ll find plenty of good information and even a Wikipedia entry entitled “3,000 Mile Myth”—along with propaganda, of course. The thing is, you’re not going to ask the question unless you already know that you don’t know the answer. In my survey, less than six percent chose “don’t know”—a fraction of the number who confessed ignorance on the boiled egg and grilled steak questions.

The Not-So-Smart Home

Not a joke: Do you know how to screw in a light bulb? Fifteen percent of my sample didn’t. The question asked which way to turn an Edison screw bulb into an empty socket (the correct answer is clockwise).