Good managers often make unethical decisions-and don't even know it.

How {Un)ethical Are You?

by Mahzarin R. Banaji, Max H. Bazerman, and Dolly Chugh

Answer true or false: "I am an ethical man­ lence of these biases suggests that even the ager." most well-meaning person unwittingly allows If you answered "true," here's an uncom­ unconscious thoughts and feelings to influence ~ fortable fact: You're probably not. Most of us seemingly objective decisions. These flawed ~ believe that we are ethical and unbiased. We judgments are ethically problematic and un­ '"~ r imagine we're good decision makers, able to dermine managers' fundamental work-to re­ <.J 2 objectively size up a job candidate or a venture cruit and retain superior talent, boost the per­ -' «-' deal and reach a fair and rational conclusion formance of individuals and teams, and :i o that's in our, and ourorganization's, best inter­ collaborate effectively with partners. ~ 2 ests. But more than two decades of research This article explores four related sources of 8'" confirms that, in reality, most of us fall woe­ unintentional unethical decision making: im­ <.J Z fully short of our inflated self-perception. plicit forms ofprejudice, bias that favors one's I V> ::J We're deluded by what Yale psychologist own group, conflict ofinterest, and a tendency => '"Q. David Armor calls the illusion of objectivity, to overclaim credit. Because we are not con­ the notion that we're free of the very biases sciously aware of these sources of bias, they 8r ~ we're so quick to recognize in others. What's often cannot be addressed by penalizing peo­ V> V> W Z more, these unconscious, or implicit, biases ple for their bad decisions. Nor are they likely Vi => can be contrary to our consciously held, ex­ to be corrected through conventional ethics '"Cl plicit beliefs. We may believe with confidence training. Rather, managers must bring a new '"~ « and conviction that a job candidate's race has type of vigilance to bear. To begin, this re­ r ~ no bearing on our hiring decisions or that quires letting go of the notion that our con­ 19 we're immune to conflicts ofinterest. But psy­ scious attitudes always represent what we f0­ r <.J chological research routinely exposes counter­ think they do. It also demands that we aban­ 2 ~ intentional, unconscious biases. The preva- don our faith in our own objectivity and our u

HARVARD BUSINESS REVIEW· DECEMBER 2003 PAGE 3 How (Un)ethical Are You?

ability to be fair; In the following pages, we "pain," and "sorrow" and at the same time sort will offer strategies that can help managers images offaces that are (depending on the bias recognize these pervasive, corrosive, uncon­ in question) black or white, youngorold, fat or scious biases and reduce their impact. thin, and so on. The test exposes implicit biases by detecting subtle shifts in reaction time that Implicit Prejudice: can occur when test takers are required to pair Bias That Emerges/rom different sets ofwords and faces. Subjects who Unconscious Beliefs consciously believe that they have no negative Most fair-minded people strive to judge others feelings toward, say, black Americans or the according to their merits, but our research elderly are nevertheless likely to be slower to shows how often people instead judge accord­ associate elderly orblack faces with the "good" ing to unconscious stereotypes and attitudes, words than they are to associate youthful or or "implicit prejudice." What makes implicit white faces with "good" words. prejudice so common and persistent is that it Since 1998, when Greenwald, Brian Nosek, is rooted in the fundamental mechanics of and Mahzarin Banaji put the IAT online, peo­ thought. Early on, we learn to associate things ple from around the world have taken over 2.5 that commonly go together and expect them million tests, confirming and extending the to inevitably coexist: thunder and rain, for in­ findings ofmore traditional laboratory experi­ stance, or gray hair and old age. This skill-to ments. Both show implicit biases to be strong perceive and learn from associations-often and pervasive. (For more information on the serves us well. IAT, see the sidebar "Are You Biased?"). But, of course, our associations only reflect Biases are also likely to be costly. In con­ approximations of the truth; they are rarely trolled experiments, psychologists Laurie Rud­ applicable to every encounter. Rain doesn't al­ man at Rutgers and Peter Glick at Lawrence ways accompany thunder, and the young can University have studied how implicit biases also go gray. Nonetheless, because we auto­ may work to exclude qualified people from matically make such associations to help us or­ certain roles. One set ofexperiments examined ganize our world, we grow to trust them, and the relationship between participants' implicit they can blind us to those instances in which gender stereotypes and their hiring decisions. the associations are not accurate-when they Those holding stronger implicit biases were don't align with our expectations. less likely to select a qualified woman who ex­ Because implicit prejudice arises from the hibited stereotypically ''masculine'' personality ordinary and unconscious tendency to make qualities, such as ambition or independence, associations, it is distinct from conscious forms for a job requiring stereotypically "feminine" of prejudice, such as overt racism or sexism. qualities, such as interpersonal skills. Yet they This distinction explains why people who are would select a qualified man exhibiting these free from conscious prejudice may still harbor same qualities. The hirers' biased perception biases and act accordingly. Exposed to images was that the woman was less likely to be s0­ thatjuxtapose black men and violence, portray cially skilled than the man, though their quali­ Mahzarin R. Banaji is the Richard women as sex objects, imply that the physi­ fications were in fact the same. These results Clarke Cabot Professor of Social Ethics cally disabled are mentally weak and the poor suggest that implicit biases may exact costs by in the department of psychology at are lazy, even the most consciously unbiased subtly excluding qualified people from the Harvard University and the Carol K person is bound to make biased associations: very organizations that seek their talents. Pforzheimer Professor at Harvard's Rad­ These associations play out in the workplace Legal cases also reveal the real costs of im­ cliffe Institute for Advanced Study in just as they do anywhere else. plicit biases, both economic and social. Con­ Cambridge, Massachusetts. Max H. Inthe mid-199OS, Tony Greenwald, a profes­ sider Price Waterhouse v. Hopkins. Despite log­ Bazerman is the Jesse Isidor Straus sor ofpsychology at the University ofWashing­ ging more billable hours than her peers, Professor of Business Administration at ton, developed an experimental tool called the bringing in $25 million to the company, and Harvard Business School in Boston. Implicit Association Test (IAT) to study uncon­ earning the praise ofher clients, Ann Hopkins Dolly Chugh, a Harvard Business scious bias. A computerized version ofthe test was turned down for partner, and she sued. School MBA, is now a doctoral candi­ requires subjects to rapidly classify words and The details ofthe case reveal that her evalua­ date in Harvard University's joint pro­ images as "good" or "bad." Using a keyboard, tors were explicitly prejudiced in their atti­ gram in organizational behavior and test takers must make split-second "good/bad" tudes. For example, they had commented that social psychology. distinctions between words like ''love,'' "joy," Ann "overcompensated for being a woman"

PAGE 4 HARVARD BUSINESS REVIEW· DECEMBER 2003 How (Un)ethical Are You?

and needed a "course at charm school." But ity that such patterns ofbehavior could unwit­ perhaps more damning from a legal stand­ tingly create. point was blunt testimony from experimental research. Testifying as an expert witness for In-Group Favoritism: the defense, psychology professor Susan Fiske, Bias That Favors Your Group now at Princeton University, argued that the Think about some ofthe favors you have done potential for biased decision makingis inherent in recent years, whether for a friend, a rela­ in a system in which a person has "solo" sta­ tive, or a colleague. Have you helped someone tus-that is, a system in which the person is get a useful introduction, admission to a the only one of a kind (the only woman, the school, or a job? Most of us are glad to help only African-American, the only person with a out with such favors. Not surprisingly, we disability, and the like). Judge Gerhard Gesell tendto domore favors for those we know, and concluded that "a far more subtle process those we know tend to be like ourselves: peo­ [than the usual discriminatory intent] is in­ ple who share ournationality, social class, and volved" in the assessments made ofAnn Hop­ perhaps religion, race, employer, or alma kins, and she won both in a lower court and in mater. This all sounds rather innocent. What's the Supreme Court in what is now a landmark wrong with asking your neighbor, the univer­ case in discrimination law. sity dean, to meet with a coworker's son? Isn't Likewise, the 1999 case of Thomas v.Kodak it just being helpful to recommend a former demonstrates that implicit biases can be the sorority sister for a job or to talk to your basis for rulings. Here, the court posed the banker cousin when a friend from church gets question of "whether the employer con­ turned down for a home loan? sciously intended to base the evaluations on Few people set out to exclude anyone race or simply did so because of unthinking through such acts ofkindness. But when those stereotypes or bias." The court concluded that in the majority or those in power allocate plaintiffs can indeed challenge "subjective scarce resources (such as jobs, promotions, and evaluations which could easily mask covert or mortgages) to people justlike them,they effec­ unconscious race discrimination." Although tively discriminate against those who are dif­ courts are careful not to assign responsibility ferent from them. Such "in-group favoritism" easily for unintentional biases, these cases amounts to giving extra credit for group mem­ demonstrate the potential for corporate liabil- bership. Yet while discriminating against those

Are Yon Biased? Are you willing to betthatyou feel the same 2.5 million online tests and further research their group on explicit measures but way toward European-Americans as you do tells us that unconscious biases are:· show relatively less implicit preference in toward African-Americans? How about widely prevalent. At least 75% oftest tak­ the tests. Conversely, white Americans re­ women versus men? Or older people versus ers show an implicit bias favoring the port a low explicit bias fortheir group but younger ones? Think twice before you take young, the rich, and whites. a higher implicit bias. that bet. Visit implicit.harvard.edu or www. robust. The mere conscious desi re not to • consequential. Those who show higher lev­ tolerance.org/hidden_bias to examine your be biased does not eliminate implicit bias. els of bias on the IAT are also likely to be­ unconscious attitudes. contrary to conscious intention. Al­ have in ways that are more biased in face­ The Implicit Association Tests available though people tend to report little or no to-face interactions with members ofthe on these sites reveal unconscious beliefs by conscious bias against African-Americans, group they are biased against and in the asking takers to make split-second associa­ Arabs, Arab-Americans, Jews, gay men, choices they make, such as hiring decisions. tions between words with positive or nega­ lesbians, or the poor, they show substan­ • costly. Research currently under way in tive connotations and images representing tial biases on implicit measures. our lab suggests that implicit bias gener­ different types of people. The various tests different in degree depending on group ates a "stereotype tax"-negotiators on these sites expose the differences~rthe status. Minority group members tend to leave money on the table because biases alignment-between test takers' conscious show less implicit preference for their cause them to miss opportunities to and unconscious attitudes toward people of own groupthan majority group members learn about their opponent and thus cre­ different races, sexual orientation, or physi­ show for theirs. For example, African­ ate additional value through mutually cal characteristics. Data gathered from over Americans report strong preference for beneficial trade-offs.

HARVARD BUSINESS REVIEW· DECEMBER 2003 PAGES How (Un)ethical Are You?

who are different is considered unethical, help­ Overclaiming Credit ing people close to us is often viewed favor­ Bias That Favors You ably. Think about the number of companies It's only natural for successful people to hold that explicitly encourage this by offering hir­ positive views about themselves. But many ing bonuses to employees who refer their studies show that the majority ofpeople con­ friends for job opportunities. sider themselves above average on a host of But consider the finding that banks in the measures, from intelligence to driving ability. United States are more likely to deny a mort­ Business executives are no exception. We tend gage application from a black person than from to overrate our individual contribution to a white person, even when the applicants are groups, which, bluntly put, tends to lead to an equally qualified The common view has been overblown sense of entitlement. We become that banks are hostile to African-Americans. the unabashed, repeated beneficiaries of this While this may be true ofsome banks and some unconscious bias, and the more we think only loan officers, social psychologist David Messick of our own contributions, the less fairly we has argued that in-group favoritism is more judge others with whom we work. likely to be at the root of such discriminatory Lab research demonstrates this most per­ lending. A white loan officer may feel hopeful sonal of biases. At Harvard, Eugene Caruso, or lenient toward an unqualified white appli­ Nick Epley, and Max Bazerman recently asked cant while following the bank's lending stan­ MBA students in study groups to estimate dards strictly with an unqualified black appli­ what portion of their group's work each had cant. Indenying the black applicant's mortgage, done. The sum ofthe contribution by all mem­ the loan officer may not be expressing hostility bers, of course, must add up to 100%. But the toward blacks so much as favoritism toward researchers found that the totals for each Would you be willing to whites. It's a subtle but crucial distinction. study group averaged 139%. In a related study, The ethical cost is clear and should be rea­ Caruso and his colleagues uncovered rampant risk being in the group son enough to address the problem. But such overestimates by academic authors of their disadvantaged by your inadvertent bias produces an additional effect: contribution to shared research projects. It erodes the bottom line. Lenders who dis­ Sadly, but not surprisingly, the more the sum own decision? criminate in this way, for example, incur bad­ of the total estimated group effort exceeded debt costs they could have avoided if their 100% (in other words, the more credit each lending decisions were more objective. They person claimed), the less the parties wanted to also may find themselves exposed to damaging collaborate in the future. publicity or discrimination lawsuits if the Likewise in business, claiming too much skewed lending pattern is publicly revealed. In credit can destabilize alliances. When each a different context, companies may pay a real party in a strategic partnership claims too cost for marginal hires who wouldn't have much credit for its own contribution and be­ made the grade but for the sympathetic hiring comes skeptical about whether the other is manager swayed by in-group favoritism. doing its fair share, they both tend to reduce In-group favoritism is tenacious whenmem­ their contributionsto compensate. This has ob­ bership confers clear advantages, as it does, for vious repercussions for the joint venture's per­ instance, among whites and other dominant formance. social groups. (It may be weaker or absent Unconscious overclaiming can be expected among people whose group membership of­ to reduce the performance and longevity of fers little societal advantage.) Thus for a wide groups.within organizations, just as it dimin­ array ofmanagerial tasks-from hiring, firing, ished the academic authors' willingness to col­ and promoting to contracting services and laborate. It can also take a toll on employee forming partnerships-qualified minority can­ commitment. Think about how employees didates are subtly and unconsciously discrimi­ perceive raises. Most are not so different from nated against, sometimes simply because they the children at Lake Wobegon, believing that are in the minority: There are not enough of they, too, rank in the upper half of their peer themto counterthe propensityfor in-group fa­ group. But many necessarily get pay increases voritism in the majority. that are below the average. If an employee learns ofa colleague's greater compensation­ while honestly believing that he himself is

PAGE 6 HARVARD BUSINESS REVIEW· DECEMBER 2003 How (Un)ethical Are You?

more deserving-resentment may be natural. an obvious incentive to prolong and extend At best, his resentment might translate into re­ their relationships with clients. But to assume duced commitment and performance. At that during this Nasdaq free fall all brokerage worst, he may leave the organization that, it house analysts were consciously corrupt, milk­ seems, doesn't appreciate his contribution. ing their clients to exploit this incentive sys­ tem, defies common sense. Surely there were Conflict ofInterest some bad apples. But how much more likely it Bias That Favors Those Who Can is that most of these analysts believed their Benefit You recommendations were sound and in their cli­ Everyone lmows that conflict of interest can ents' best interests. What many didn't appreci­ lead to intentionally corruptbehavior. Butnu­ ate was that the built-in conflict ofinterest in merous psychological experiments show how their compensation incentives made it impos­ powerfully such conflicts can unintentionally sible for them to see the implicit bias in their skew decision making. (For an examination of own flawed recommendations. the evidence in one business arena, see Max Bazerman, George Loewenstein, and Don Trying Harder Isn't Enough Moore's November 2002 HBR article, "Why As companies keep collapsing into financial Good Accountants Do Bad Audits.") These ex­ scandal and ruin, corporations are responding periments suggest that the work world is rife with ethics-training programs for managers, with situations in which such conflicts lead and many of the world's leading business honest, ethical professionals to unconsciously schools have created new courses and chaired make unsound and unethical recommenda­ professorships in ethics. Many ofthese efforts tions. focus on teaching broad principles of moral Physicians, for instance, face conflicts ofin­ philosophy to help managers understand the terest when they accept payment for referring ethical challenges they face. patients into clinical trials. While, surely, most We applaud these efforts, but we doubt that physicians consciously believe that their refer­ a well-intentioned, just-try-harder approach rals are the patient's best clinical option, how will fundamentally improve the quality of ex­ do they lmow that the promise ofpayment did ecutives' decision making. To do that, ethics not skew their decisions? Similarly, many law­ training must be broadened to include what is yers earn fees based on their clients' awards or now lmown about how our minds work and settlements. Since going to trial is expensive must expose managers directly to the uncon­ and uncertain, settling out ofcourt is often an scious mechanisms that underlie biased deci­ attractive option for the lawyer. Attorneys sion making. And it must provide managers may consciously believe that settling is in their with exercises and interventions that can root clients' best interests. But how can they be ob­ out the biases that lead to bad decisions. jective, unbiased judges under these circum­ Managers can make wiser, more ethical de­ stances? cisions ifthey become mindful oftheir uncon­ Research done with brokerage house ana­ scious biases. Buthow can we get at something lysts demonstrates how conflict ofinterest can outside our conscious awareness? By bringing unconsciously distort decision making. A sur­ the conscious mind to bear. Just as the driver vey of analysts conducted by the financial re­ ofa misaligned car deliberately counteracts its search service First Call showed that during a pull, so canmanagers develop conscious strate­ period in 2000 when the Nasdaq dropped gies to counteract the pull oftheirunconscious 60%, fully 99% of brokerage analysts' client biases. What's required is vigilance-continual recommendations remained "strong buy," awareness ofthe forces that can cause decision ''buy,'' or ''hold.'' What accounts for this dis­ making to veer from its intended course and crepancy between what was happening and continual adjustments to counteract them. what was recommended? The answer may lie Those adjustments fall into three general cate­ in a system that fosters conflicts ofinterest. A gories: collecting data, shaping the environ­ portion of analysts' pay is based on brokerage ment, and broadening the decision-making firm revenues. Some firms even tie analysts' process. compensation to the amount of business the Collect data. The first step to reducing un­ analysts bring in from clients, giving analysts conscious bias is to collect data to reveal its

MARVARD BUSINESS REVIEW· DECEMBER 2003 PAGE 7 How (Un)ethical Are You?

presence. Often, the data will be counterintui­ knowing the magnitude and pervasiveness of tive. Consider many people's surprise to learn your own biases can help direct your attention of their own gender and racial biases on the to areas ofdecision making that are in need of lAT. Why the surprise? Because most of us careful examination and reconsideration. For trust the "statistics" our intuition provides. example, a manager whose testing reveals a Better data are easily, but rarely, collected. bias toward certain groups ought to examine One way to get those data is to examine our her hiring practices to see if she has indeed decisions in a systematic way. been disproportionately favoring those groups. Remember the MBA study groups whose But because so many people harbor such bi­ participants overestimated their individual ases, they can also be generally acknowledged, contributions to the group effort so thatthe to­ and that knowledge can be used as the basis tals averaged 139%? When the researchers for changing the way decisions are made. It is asked group members to estimate what each importantto guard against using pervasiveness of the other members' contributions were be­ to justify complacency andinaction: Pervasive­ fore claiming their own, the total fell to 121%. ness ofbias is not a mark ofits appropriateness The tendency to claim too much credit still any more than poor eyesight is considered so persisted, but this strategy of "unpacking" the ordinary a condition that it does not require work reduced the magnitude ofthe bias. In en­ corrective lenses. vironments characterized by "I deserve more Shape your environment. Research shows than you're giving me" claims, merely asking that implicit attitudes can be shaped by exter­ team members to unpack the contributions of nal cues in the environment. For example, others before claiming their own share of the Curtis Hardin and colleagues at UCLA used pot usually aligns claims more closely with the lAT to study whether subjects'implicit Areyourcompany's high what's actually deserved. As this example dem­ race bias would be affected ifthe test was ad­ onstrates, such systematic audits of both indi­ ministered by a black investigator. One group achievers all castfrom vidual and group decision-making processes ofstudentstookthe testunderthe guidance of the same mold? can occur even as the decisions are being a white experimenter; another group took the made. test with a blackexperimenter. The mere pres­ Unpacking is a simple strategy that manag­ ence of a black experimenter, Hardin found, ers should routinely use to evaluate the fair­ reduced the level of subjects' implicit anti­ ness of their own claims within the organiza­ black bias on the lAT. Numerous similar stud­ tion. But they can also apply it in any situation ies have shown similar effects with other so­ where team members or subordinates may be cialgroups. What accounts for such shifts? We overclaiming. For example, in explaining a can speculate that experimenters in class­ raise that an employee feels is inadequate, a rooms are assumed to be competent, in manager should ask the subordinate not what charge, and authoritative. Subjects guided by he thinks he alone deserves but what he con­ a black experimenter attribute these positive siders an appropriate raise after taking into ac­ characteristics to that person, and then per­ count each coworker's contribution and the haps to the group as a whole. These findings pool available for pay increases. Similarly, suggest that one remedy for implicit bias is to when anindividual feels she's doingmore than expose oneself to images and social environ­ her fair share of a team's work, asking her to ments that challenge stereotypes. consider other people's efforts before estimat­ We know of a judge whose court is located ingherown canhelp align her perception with in a predominantly African-American neigh­ reality, restore her commitment, and reduce a borhood. Because of the crime and arrest pat­ skewed sense ofentitlement. ternsinthe community, mostpeople the judge Taking the lAT is another valuable strategy sentences are black. The judge confronted a for collecting data,. We recommend that you paradox. On the one hand, she took a judicial and others in your organization use the test to oath to be objective and egalitarian, and in­ expose your own implicit biases. Butone word deed she consciously believed that her deci­ of warning: Because the test is an educational sions were unbiased. On the other hand, every andresearch too~ not a selection or evaluation day she was exposed to an environment that tool, it is critical thatyou consider your results reinforced the association between black men and others' to be private information. Simply and crime. Although she consciously rejected

PAGE 8 HARVARD BUSINESS REVIEW· DECEMBER 2003 How (Un)ethical Are You?

racial stereotypes, she suspected that she har­ portunities for younger ones. Now pretend bored unconscious prejudices merely from that, as you make your decisions, you don't working in a segregated world. Immersed in know which group you belong to. That is, you this environment each day, she wondered ifit don't know whether you are senior or junior, was possible to give the defendants a fair hear­ married or single, gay or straight, a parent or ing. childless, male or female, healthy or un­ Rather than allow her environment to rein­ healthy. You will eventually find out, but not force a bias, the judge created an alternative until after the decision has been made. In this environment. She spent a vacation week sit­ hypothetical scenario, what decision would ting in a fellow judge's court in a neighbor­ you make? Would you be willing to risk being hood where the criminals being tried were pre­ in the group disadvantaged by your own deci­ dominantly white. Case after case challenged sion? How would your decisions differ if you the stereotype ofblacks as criminal and whites could make them wearing various identities as law abiding and so challenged any bias not your own? against blacks that she might have harbored. This thought experiment is a version of Think about the possibly biased associations philosopher John Rawls's concept of the your workplace fosters. Is there, perhaps, a "veil of ignorance," which posits that only a "wall offame" with pictures ofhigh achievers person ignorant of his own identity is capa­ all cast from the same mold? Are certain types ble of a truly ethical decision. Few of us can of managers invariably promoted? Do people assume the veil completely, which is pre­ overuse certain analogies drawn from stere

HARVARD BUSINESS REVIEW· DECEMBER 2003 PAGE 9 How (Un)ethical Are You?

Just considering a trust a complex presentation to a female col­ more than good intention, is a defining char­ league or to receive a promotion from an acteristic ofan ethical manager. They must ac­ counterstereotypical African-American boss-can prompt less­ tively collect data, shape their environments, choice at the conscious biased and more ethical decision making. and broaden their decision making. What's Similarly, consciously considering counterin­ more, an obvious redress is available. Manag­ level can reduce implicit tuitive options in the face of conflicts of in­ ers should seek every opportunity to imple­ terest, or when there's an opportunity to ment affinnative action policies-notbecause bias. overclaim, can promote more objective and of past wrongs done to one group or another ethical decisions. but because of the everyday wrongs that we can now document are inherent in the ordi­ The Vigilant Manager nary, everyday behavior of good, well-inten­ Ifyou answered "true" to the question at the tioned people. Ironically, only those who un­ start of this article, you felt with some confi­ derstand their own potential for unethical dence that you are an ethical decision maker. behavior can become the ethical decision How would you answer it now? It's clear that makers that they aspire to be. \) neither simple conviction nor sincere inten­ tion is enough to ensure that you are the ethi­ Reprint R0312D; Harvard Business Review cal practitioner you imagine yourself to be. OnPoint 5526 To order, see the next page Managers who aspire to be ethical must chal­ or call 800-988-0886 or 617-783-7500 lenge the assumption that they're always un­ or go to www.hbr.org biased and aclmowledge that vigilance, even

PAGE 10 HARVARD BUSINESS REVIEW· DECEMBER 2003