The Invisible Gorilla: And Other Ways Our Intuitions Deceive Us

by Christopher Chabris

  Ronald Rensink, a vision scientist at the University of British Columbia and a leader in the study of change blindness, has made the interesting proposal that the mind works much like a Web browser. Chris’s father, a smart man born long before the digital computer was invented, has asked Chris several times over the years to explain how all the information from the Internet gets into his “set,” his quaint name for his iMac. Most of us know that the contents of the Internet are distributed across millions of computers around the world, rather than being duplicated inside every desktop computer. But if you had a fast enough Internet connection and there were fast enough servers on the network, you would not be able to see any difference between these two accounts of how the Internet works. From your perspective, the information you want arrives as soon as you request it; you follow a link with your Web browser, and the contents of the page appear almost immediately. The perception that the Web is stored locally on your computer is a reasonable misunderstanding, and in most cases, one that would make no difference to you. When your Internet connection goes down, though, your “set” no longer has access to the information you thought was inside it. Similarly, the experiments in which we don’t notice people changing into other people reveal how little information we store in our memories. We don’t need to store this information any more than our computers need to store the contents of the Web—in each case, under normal circumstances, we can obtain the information on demand, whether by looking at the person standing in front of us or by accessing sites on the Internet.30

  Neurobabble and Brain Porn

  Companies often prey on the illusion of knowledge to hawk their wares, emphasizing technical details in a way that leads people to think they understand how a product works. For example, audiophiles and audio cable manufacturers regularly wax poetic about the quality of the cables that connect different system components. Cable manufacturers tout the superior shielding on their cables, greater dynamic range, higher-quality copper, gold-plated connectors, and cleaner sound. Reviewers say that the cables make their old speakers sound like new ones, and that there is simply no comparison between the high-end cables and regular cables. In at least one informal experiment, though, audiophiles in a blind test could not distinguish one expensive set of cables from wire coat hangers used as speaker cables!31 All of the high-tech cable technology made little difference in the sound of the music. Of course, it is possible that the other components in their stereo systems might have been of insufficient quality to reveal the difference, but most people listening to music or watching movies on a home theater system wouldn’t have the sort of equipment necessary to detect the difference either.

  The hype is much funnier in the case of cables that transmit digital signals. As long as a cable is able to transmit the 0s and 1s that make up a digital signal, the quality of the wire doesn’t matter one bit. The factor that matters is the protocol used to generate and interpret those 0s and 1s. Modern stereo systems and video systems use digital standards such as HDMI to transfer information from one component to another. Yet prices on HDMI cables vary by more than a factor of ten: A cable that costs $5 will transmit the signal just as well as a cable that costs $50. Denon even sells a 1.5-meter Ethernet cable for audio systems that is priced at $500. Here is the product description at Amazon.com:

  Get the purest digital audio you’ve ever experienced from multichannel DVD and CD playback through your Denon home theater receiver with the AK-DL1 dedicated cable. Made of high-purity copper wire, it’s designed to thoroughly eliminate adverse effects from vibration and helps stabilize the digital transmission from occurrences of jitter and ripple. A tin-bearing copper alloy is used for the cable’s shield while the insulation is made of a fluoropolymer material with superior heat resistance, weather resistance, and anti-aging properties. The connector features a rounded plug lever to prevent bending or breaking and direction marks to indicate correct direction for connecting cable.

  Apparently, some people have actually bought this product, but as reviewers on Amazon.com point out, since the signal is digital rather than analog, there is no reason to expect any difference in sound quality between this cable and an ordinary Ethernet cable you can get from your local dollar store. It’s not even clear what “jitter” and “ripple” mean, why vibration matters for a stream of 0s and 1s, or how fluoropolymers prevent aging. Most of the hundreds of reviews of this product on Amazon.com are facetious, and the five most commonly associated customer tags for it include “snake oil,” “ripoff,” “waste of money,” “throwing your money away,” and “unconscionable.”32

  A group of researchers in the Yale psychology department, including Dan’s graduate school adviser, Frank Keil, and our friend Jeremy Gray, conducted a mischievous experiment in which subjects read passages of text that included some uninformative babble like the description of Denon’s cable. Each passage began with a straightforward summary of a psychology experiment like the following:

  Researchers created a list of facts that about 50% of people knew. Subjects then read the list and noted which ones they already knew. They then judged what percentage of other people would know those facts. When subjects knew a fact, they thought that an inaccurately large percentage of others would know it, too. For example, a subject who already knew that Hartford was the capital of Connecticut might think that 80% of other people would know it, even though only 50% actually do. The researchers call this finding “the curse of knowledge.”

  After reading this passage, subjects would then read either a good or a bad explanation for the “curse of knowledge.” The “bad” explanation for the curse of knowledge was the following: “This ‘curse’ happens because subjects make more mistakes when they have to judge the knowledge of others. People are better at judging what they themselves know.” Note that this explanation doesn’t actually tell us anything about the “curse of knowledge.” The experiment showed that people judge the knowledge of others differently depending on whether they themselves have the knowledge. It said nothing about whether we are better at judging our own knowledge or the knowledge of others.

  In contrast, a “good” explanation read as follows: “This ‘curse’ happens because subjects have trouble switching their point of view to consider what someone else might know, mistakenly projecting their own knowledge onto others.” This explanation is good because it explains the curse of knowledge in terms of a broader principle about our minds—the difficulty we have in adopting another person’s perspective. The explanation may or may not be scientifically correct, but at least it is logically relevant.

  Each subject read a series of these passages and explanations and rated how satisfying each explanation was. Generally, people rated the good explanations as more satisfying—they recognized that the good explanations actually said something to explain the experimental result, and the bad ones were largely irrelevant.

  The twist in the experiment came from a third condition, in which irrelevant information about the brain was added to the bad explanation: “Brain scans indicate that this ‘curse’ happens because of the frontal lobe brain circuitry known to be involved in self-knowledge. Subjects make more mistakes when they have to judge the knowledge of others. People are much better at judging what they themselves know.”

  Much as the technobabble in the cable description on Amazon.com doesn’t turn a $2 bundle of wires into a $500 gadget, this superfluous brain-talk, which we like to call “neurobabble,” does nothing to rescue the validity of the bad psychological explanation. But the subjects rated the bad explanations that included neurobabble as more satisfying than those that did not. The neurobabble induced an illusion of knowledge; it made the bad explanations seem like they imparted more understanding than they actually did. Even students in an introductory neuroscience course were influenced. Fortunately, neuroscience graduate students had enough actual understanding to immunize them to the neurobabble.33

  The cousin of neurobabble is “brain porn,” the colorful im
ages of blobs of activity on brain scans that can seduce us into thinking we have learned more about the brain (and the mind) than we really have. Neuroscientists have recognized that these pictures can sometimes be more of a sales tool for their research than a true aid to understanding. In one clever experiment, David McCabe and Alan Castel had subjects read one of two descriptions of a fictitious research study. The text was identical, but one description was accompanied by a typical three-dimensional brain image with activated areas drawn in color, while the other included only an ordinary bar graph of the same data. Subjects who read the version with the brain porn thought that the article was significantly better written and made more sense. The kicker is that none of the fictitious studies actually made any sense—they all described dubious claims that were not at all improved by the decorative brain scans.34

  Neurobabble has crept into advertising, alongside technobabble and other irrelevant information that makes consumers feel that they understand something better than they really do. In a ubiquitous magazine ad, Allstate Insurance asks, “Why do most 16-year-olds drive like they’re missing a part of their brain?” and answers, “Because they are.” The company attributes their risky driving to an immature dorsal lateral prefrontal cortex, a region critical for “decision making, problem solving and understanding future consequences of today’s actions.” Beneath the headline, a cartoon depicts a brain with a car-shaped hole right in this location.35 The ad copy might be right about the science, but the information about the brain is entirely irrelevant to its point. Teenagers are indeed risky drivers, but that’s all you need to know to be persuaded that parents should talk more to their children about road safety, which is the point of Allstate’s ad. If you’re more likely to talk to your kids (or to buy Allstate’s insurance) because you know what part of the brain is responsible for risk-taking, you are a victim of the illusion of knowledge—courtesy of neurobabble and brain porn.

  There’s a 50 Percent Chance the Weather Will Be Great, Sort of Wish You Were Here

  In the 2005 comedy-drama The Weather Man, the title character (played by Nicolas Cage) is paid well but receives little respect for his job, which consists entirely of acting authoritative while reading forecasts prepared by others. It’s easy to mock a class of professionals whose work comes to mind mainly when a game is rained out or a flight is delayed. There are some places, though, where the weather really is important news, and accurate weather forecasts can make millions or even billions of dollars of difference in people’s lives. Dan lives in Champaign, a college town in east-central Illinois. The University of Illinois, where he teaches, is the largest employer in the area, but the dominant economic force in the region is large-scale farming of corn and soybeans. Illinois produces a larger soybean crop than any other state and is the second-largest corn producer.36 The weather influences all of the important decisions a farmer makes, including when to plant and harvest, what to plant, and how to plan ahead for future supply and demand. Farmers in Illinois monitor conditions far outside their own region. A bumper corn crop during Argentina’s summer can affect which crops Illinois farmers plant in the spring. Even the world markets for oil and other forms of energy affect planting decisions, since Illinois corn is responsible for 40 percent of the ethanol produced in the United States.

  Few National Public Radio stations have more than one weather forecaster on staff, and even fewer have one with a meteorology degree. The Champaign NPR station, WILL, has one full-time meteorologist, two part-time meteorologists, and another weather forecaster on staff. WILL gives detailed weather forecasts throughout the day, devoting as much time to the weather as any station in the United States. It has to, because farmers depend on weather forecasts for their livelihood.37 If weather forecasters really know how much they know—in technical terms, if they are “well calibrated”—then farmers can rely on their predictions when making major decisions.

  Although people have attempted to predict the weather for millennia, the first published forecast appeared in print less than 150 years ago, in Cincinnati on September 1, 1869: “Cloudy and warm this evening. Tomorrow clear.”38 The addition of probabilities expressed as percentages didn’t begin until 1920, when Cleve Hallenbeck, the head of the U.S. Weather Bureau office in Roswell, New Mexico, published an article advocating their use. Hallenbeck had tested his method with an informal experiment that lasted 220 days. On each day he estimated the probability of rain and then recorded whether it rained. His forecasts turned out to be remarkably well calibrated: It rained on most of his high-probability days and on few of his low-probability days. However, only in 1965 did the U.S. National Weather Service begin to regularly include percentage probabilities of rain in its forecasts. In 1980, meteorologists Jerome Charba and William Klein undertook a massive examination of more than 150,000 precipitation forecasts during the two years from 1977 to 1979. The forecasted likelihood of rain matched the actual probability of rain almost perfectly. Tellingly, the only systematic errors happened when the forecasters assigned a 100 percent chance of rain—it turned out to rain on only about 90 percent of those days. Beware of certainty!

  What makes weather forecasts, at least good ones, different from other forms of reasoning and prediction? When meteorologists say that there is a 60 percent chance of rain, they are estimating the probability that, given the existing atmospheric conditions, it actually will rain. And these estimates are highly accurate over a long series of forecasts. Meteorologists continually adjust their predictions—and the mathematical and statistical models and computer programs that generate those predictions—based on feedback from previous predictions. If a 60 percent probability of rain is attached to certain climate patterns, but it only rains 40 percent of the time, then the models are refined so that the next time those atmospheric conditions occur, the estimated probability of rain will be lower. Weather forecasting is unusual in that forecasters receive immediate and definitive feedback about their predictions, and their knowledge of probabilities accumulates over time. For example, during the period from 1966 through 1978, skill at forecasting precipitation thirty-six hours in advance nearly doubled.39

  Like weather forecasters, when we receive appropriate feedback, we can sometimes calibrate our judgments and eliminate the illusion of knowledge. In a demonstration Dan has used in an introductory psychology class, students are each given a playing card, which they proceed to hold to their forehead so that they can’t see it, but everyone else can.40 Then each person in the class tries to get the person with the highest possible card to pair up with him or her. Remember, the students can’t see their own card, but they can see everyone else’s—so they can see who rejects them.

  Initially, most people in the class will try to pair up with an Ace or King (the highest cards), but most will be rejected. Only those who have a really high card are likely to be accepted by someone who has an Ace or a King. People with an Ace or King don’t know what they have, but they know that they really can’t do better than an Ace or King and they aren’t likely to accept an invitation from someone with a 6 or 7—they hope to match with someone higher. Surprisingly, people pair off quite quickly with others who have cards comparable to their own. They are able to rapidly use the feedback they get from rejection to calibrate their expectations. The same principle can be used to explain why people of widely different attractiveness rarely end up as couples41—people reach for the best they can get, and dating allows for some calibration of your self-impressions.

  The card-matching game and the real world of dating and mating provide immediate and direct (and sometimes painful) feedback in the form of rejection. Unfortunately, for most of the judgments that we make in our lives, we never receive the precise feedback that weather forecasters do of seeing the next morning whether we were right or wrong, day after day, year after year. This is an important difference between meteorology and fields like medicine. Information about the correctness of a diagnosis, or the outcome of a surgical procedure, is available in p
rinciple. In practice, though, it is rarely collected systematically, stored, and analyzed the way data about the weather is; a doctor who diagnoses pneumonia and prescribes a treatment will have to wait awhile to learn—or may never learn—whether the treatment worked. Even then it may be difficult to distinguish the effects of the treatment from improvements that happened spontaneously. If you’ve recently switched from a film camera to a digital camera, you have experienced the benefits of instant feedback. You no longer have to wait for your film to be developed before you know what you did wrong (or right) in composing your shots. And when you do make a mistake, you can fix it right away. As any student knows, whether in photography, psychology, or business, it’s harder to improve if you don’t get immediate feedback about your mistakes.

  Why Does the Illusion of Knowledge Persist?

  Scientists, architects, and hedge fund managers are respected, but weather forecasters are parodied. Yet weather forecasters have fewer illusions about their own knowledge than do members of these other professions. In Chapter 3 we saw that doctors who consulted books and computers were underappreciated by patients, whereas a rape victim who expressed no doubt in her testimony was praised as a model witness. There we argued that our love of confidence can reward people for acting as though they are more skilled and accurate than they really are. The illusion of knowledge has similar consequences: We seem to prefer the advice of experts who act like they know more than they really do—or who honestly believe their knowledge is greater than it is.