Your Unlearning Report: The Trouble with Empathy, Implicit Bias, and Believing in Luck
January 27, 2017
For a skeptic, there is nothing more satisfying than discovering that some previously cherished truth has been overturned by new evidence. It is in that spirit that I offer the following Unlearning Report.
Empathy is Bad
Everybody loves empathy. Former President Barack Obama often spoke about our “empathy deficit” and the need to “see the world through the eyes of those who are different from us.” Amazon.com lists over 1,500 books with “empathy” in their titles or subtitles, and the Internet is replete with blogs and YouTube videos on the subject.1 Yes, everybody loves empathy—everybody except Yale University psychologist Paul Bloom.
It seems wrong to be against something so kind and well meaning as empathy, but in his new book, Against Empathy: The Case for Rational Compassion, Bloom builds a strong argument for empathy as a destructive emotion. He suggests that empathy is innumerate and myopic and that more good can be done by adopting “rational compassion,” a more detached form of caring.
For example, Bloom cites the research of C. Daniel Batson and colleagues who told study participants about a ten-year-old girl who had a fatal disease and was waiting in line for a treatment that would relieve her pain. The participants were told that they could move her ahead in the line, and when simply asked what to do, most said that the girl must wait because there were others ahead of her in line. However, when they were asked to imagine what she felt, they were more likely to choose to move her up the line. In this case, empathy made participants bend the rules unfairly.
Bloom spends considerable time discussing the innumeracy of empathy. This passage sums up the problem:
Stalin has been quoted as saying, “One death is a tragedy; one million is a statistic.” And Mother Teresa once said, “If I look at the mass, I will never act. If I look at the one, I will.” To the extent that we can recognize that the numbers are significant when it comes to moral decisions, it’s because of reason, not sentiments.2
In one study, participants in two separate groups were asked to give money to a drug that would save one child or a drug that would save eight children. They tended to give equal amounts, but when a name and a picture where associated with the child, donations were larger for the one child than the eight. Similarly, psychologist Paul Slovic points to the case of eighteen-year-old U.S. high school student Natalee Holloway, who went missing while on vacation in Aruba. Coverage of this case dominated the cable news for weeks, despite an on-going genocide in Darfur.
In addition, caregivers who are empathetic may also be less effective than those who are more detached. As a recent article in Aeon suggested, doctors who attempt to feel their patients’ pain will be ill suited to concentrate on their work. Similarly, when seeking the help of therapists, we don’t really want them to experience what we are experiencing. We are looking for a compassionate listener who can provide suggestions and support. Finally, there is evidence that, in contrast with a more reserved sense of compassion, sustained empathy can lead to burnout and negative emotions.3
As an alternative to empathy, Bloom advocates rational compassion, an approach consistent with the utilitarianism of philosopher Peter Singer. For Bloom, being a good person involves caring for others combined with an appreciation of how best to distribute that care. This is a view embodied by a growing Effective Altruism movement, which Singer also supports. Effective altruists donate money and time to others, but they do so after calculating where they can do the most good. When it comes to financial contributions, this approach often leads people to direct their funds to Africa and other parts of the developing world, where economic differences allow their dollars to go further. Effective altruists also consult rigorous evaluators of charities, such as GiveWell. It’s less about making the giver feel good and more about doing the most good you can.
You’re Not As Racist As You Thought
Since the 1990s, social psychologists have been studying unconscious biases—particularly those surrounding race, gender, and other socially relevant variables—using the Implicit Association Test (IAT). The test uses reaction times as an indirect measure of cognitive processes. In a typical arrangement, participants are asked to sort images or words by pressing either the “e” key of a keyboard with their left hand or the “i” key with their right hand. At first the task is simple. You might be asked to sort words that are either negative sounding (“bad”) or positive (“good”). In a similarly easy task, you might be asked to sort pictures of African American or European American men. The words in the upper left-hand and upper right-hand corners of the panels in Figure 1 indicate the left and right choices that participants are to use in sorting.
The IAT gets more difficult when trials combine the two kinds of stimuli— pictures and words—to be sorted. As shown in panels 3 and 4 of Figure 1, participants are told that either a word or a face might appear in the center of the screen, and they should sort the stimulus accordingly. These combined sorting trials provide the crucial test for presence of bias.
Typically, when the left and right responses combine categories that are inconsistent with the participant’s biases, then responses slow down. For example, if the participant has an unconscious bias against African Americans, it will take longer to decide the word “Happy” requires a right-hand response when it is arranged as shown in Figure 1 panel 4. The most common theory behind this effect is that certain mental operations are automatic and others require deliberation. When the categories are inconsistent with our implicit biases, slower deliberative processing is required to sort the stimuli.
By now implicit bias is a well-established concept, so much so that Carl Bialik of the FiveThirtyEight blog recently cited unconscious bias against women as a possible explanation for the results of the 2016 United States presidential election. Over two million people have taken the IAT by visiting the Harvard University Project Implicit website,4 and in 2013 Mahzarin Banaji and Anthony Greenwald, the researchers most closely associated with the IAT, published a bestselling book called Blind Spot: Hidden Biases of Good People. The IAT is often used in industry, education, and nonprofit organizations to spur an appreciation for the kinds of biases most of us harbor.
But how important is implicit bias? A recent article in the Chronicle of Higher Education outlines a growing number of questions that dog the IAT. Two decades after the test was first introduced, hundreds of research studies have used it, and several meta-analyses have been conducted summarizing the combined results of many investigations. Most notably Patrick Forscher of the University of Wisconsin and colleagues recently conducted a meta-analysis looking at 426 studies incorporating a total of 72,063 research participants. Furthermore, the Forscher study employed a number of improvements over previous meta-analyses.
Earlier investigations had already raised questions about whether the IAT is reliable enough to produce consistent scores when users are retested, and even Banaji and Greenwald admitted that the correlation between implicit bias and actual prejudicial behavior was weak.5 Unfortunately, the Forscher study found even weaker associations between unconscious bias and prejudicial behavior. Furthermore, implicit bias researchers have long assumed that interventions aimed at reducing these unconscious biases would result in changes in actual behavior, but the authors of the new meta-analysis concluded, “Our findings stand in stark contrast to these predictions” (p. 33). Forscher and colleagues suggest there is still value in studying implicit bias as an indicator of the culture’s effects on us, but they also state that any “efforts to change behavior by directly changing implicit bias would be misguided” (p. 34).
The Forscher study is still under review at the prestigious journal Psychological Bulletin and has not yet been accepted for publication. Nonetheless, there are two reasons to believe the results are worthy of attention. First, in an approach advocated by the Center for Open Science (COS), the authors used a publicly stated methodology and posted the raw data of their study so that others could examine and reanalyze it if they wish. As I mentioned in my last Skeptical Inquirer column, COS is a direct response to recent revelations about the unreliability of research in social psychology, medicine, and other areas. The hope is that opening up research and inviting peer comment throughout the entire process will result in more trustworthy outcomes. The recommendations of COS are quickly gaining acceptance among many scientific journals and professional associations.
Finally, one of the coauthors of the Forscher meta-analysis is Brian Nosek, who is both a cofounder of COS and, along with Banaji and Greenwald, the person most closely associated with early use of the IAT.
Believing in Luck Will Not Help Your Golf Game
This particular unlearning lesson hits close to home for me. In 2010, a group of researchers at the University of Cologne conducted the first study to show that believing in luck could improve performance of a skilled activity.6 Research participants were invited to putt a golf ball into a cup on the carpet of a laboratory. As they were handed a golf ball, half the participants were told, “So far it has turned out to be a lucky ball,” and the other half were simply told “This is the ball everyone has used so far.” Remarkably, in this and different versions of the study, participants who had the “lucky ball” actually sunk significantly more putts on average than those who did not get a lucky instruction.
I found these results quite interesting, and given that the German researchers had published their work in a prestigious journal and had conducted several related experiments, all of which showed the luck effect, I incorporated the study into the 2013 revision of my book on superstition and began to mention it in public comments about belief in superstition. It seemed plausible that, in the case of a skilled activity, such as golf, there could be a psychological benefit to believing in luck. There was no reason to credit anything supernatural in the findings, but just as placebos can produce real health benefits, it has long been speculated that superstitions might have a psychological effect that could translate into better performance on skilled activities. Unfortunately, in this case, I was too quick to jump at an exciting new finding.
In 2014, two researchers from Dominican University in Illinois published a replication of the German golf ball study in Social Psychology.7 This was a registered replication, which means that, similar to the Forscher meta-analysis, the researchers publicly stated exactly what they intended to do before they started collecting data, and, in addition, once the experiment was complete, they made the raw data publicly available so that anyone who wished to could independently reanalyze the results. Finally, the sample of participants for the Dominican study was roughly three times that of the German study, making it a much more powerful test.
As by now you may have guessed, the results of the replication were negative. Telling people “this ball has been lucky today” did not improve their ability to putt into a cup. It is true that the original study used German participants, and the replication used U.S. participants, but the Dominion researchers measured the level of superstition in their people and found it was comparable to that of the German participants. Of course, some kind of cultural difference could be at play, but superstition and luck are popular concepts in both countries. So it looks as though the famous golf ball study has become just another victim of the reproducibility crisis. We will have to wait for further research to determine if there is any psychological benefit to believing in superstitions.
- I highly recommend Paul Bloom’s book Against Empathy: The Case for Rational Compassion. It is well reasoned, clear, and entertaining.
- If you have not already done so, you might enjoy visiting the Implicit Project website and taking one of the several implicit bias tests offered there. They are free, and although—as we now know—unconscious bias does not necessarily translate into biased actions, you might find the test interesting and thought-provoking. Do you think it really measures what it purports to measure?
- It appears the best advice about luck and golf comes from a quote so good that it has been attributed variously to Ben Hogan, Gary Player, or Arnold Palmer: “The more I practice, the luckier I get.”
- Paul Bloom, Against Empathy: The Case for Rational Compassion (New York: Ecco), 18-19.
- Bloom, Against Empathy, 89.
- Bloom, Against Empathy, 137-139.
- Damisch, Lysann, Barbara Stoberock, and Thomas Mussweiler. "Keep your fingers crossed! How superstition improves performance." Psychological Science 21, no. 7 (2010): 1014-1020.
- Calin-Jageman, Robert J., and Tracy L. Caldwell. "Replication of the superstition and performance study by Damisch, Stoberock, and Mussweiler (2010)." Social Psychology 45 no. 3 (2014), 239-245.