Which theorist discovered the “risky shift phenomenon”?

217

MODULE

20 ❖

How Groups Intensify Decisions

Which effect—good or bad—does group interaction more often have? Police brutality and mob violence demonstrate its de -structive potential. Yet support-group leaders, management consultants, and educational theorists proclaim group interaction’s benefits, and social and religious movements urge their members to strengthen their identities by fellowship with like-minded others. Studies of people in small groups have produced a principle that helps explain both bad and good outcomes: Group discussion often strengthens members’ initial inclinations. The unfolding of this research on group polarization illustrates the process of inquiry—how an interesting discov- ery often leads researchers to hasty and erroneous conclusions, which ultimately are replaced with more accurate conclusions. This is a scientific mystery I can discuss firsthand, having been one of the detectives.

THE CASE OF THE “RISKY SHIFT” More than 300 studies began with a surprising finding by James Stoner (1961), then an MIT graduate student. For his master’s thesis in manage- ment, Stoner tested the commonly held belief that groups are more cautious than individuals. He posed decision dilemmas in which the participant’s task was to advise imagined characters how much risk to take. Put yourself in the participant’s shoes: What advice would you give the character in this situation?1

1This item, constructed for my own research, illustrates the sort of decision dilemma posed by Stoner.

mye35171_ch20_217-232.indd Page 217 19/11/10 11:50 AM user-f494/208/MHSF219/myr35171_disk1of1/0078035171/myr35171_pagefiles

218 PART THREE SOCIAL INFLUENCE

Helen is a writer who is said to have considerable creative talent but who so far has been earning a comfortable living by writing cheap westerns. Recently she has come up with an idea for a potentially significant novel. If it could be written and accepted, it might have considerable literary impact and be a big boost to her career. On the other hand, if she cannot work out her idea or if the novel is a flop, she will have expended consid- erable time and energy without remuneration. Imagine that you are advising Helen. Please check the lowest probability that you would consider acceptable for Helen to attempt to write the novel. Helen should attempt to write the novel if the chances that the novel will be a success are at least

1 in 10

2 in 10

3 in 10

4 in 10

5 in 10

6 in 10

After making your decision, guess what this book’s average reader would advise. Having marked their advice on a dozen such items, five or so indi- viduals would then discuss and reach agreement on each item. How do you think the group decisions compared with the average decision before the discussions? Would the groups be likely to take greater risks, be more cautious, or stay the same? To everyone’s amazement, the group decisions were usually riskier. Dubbed the “risky shift phenomenon,” this finding set off a wave of group risk-taking studies. These revealed that risky shift occurs not only when a group decides by consensus; after a brief discussion, individuals, too, will alter their decisions. What is more, researchers successfully repeated Stoner’s finding with people of varying ages and occupations in a dozen nations. During discussion, opinions converged. Curiously, however, the point toward which they converged was usually a lower (riskier) num- ber than their initial average. Here was a delightful puzzle. The small risky shift effect was reliable, unexpected, and without any immediately obvious explanation. What group influences produce such an effect? And how widespread is it? Do discussions in juries, business commit- tees, and military organizations also promote risk taking? Does this explain why teenage reckless driving, as measured by death rates, nearly doubles when a 16- or 17-year-old driver has two teenage passengers rather than none (Chen & others, 2000)? After several years of study, we discovered that the risky shift was not universal. We could write decision dilemmas on which people

7 in 10

8 in 10

9 in 10

10 in 10 (Place a check here if you think Helen should attempt the novel only if it is certain that the novel will be a success.)

mye35171_ch20_217-232.indd Page 218 19/11/10 11:50 AM user-f494/208/MHSF219/myr35171_disk1of1/0078035171/myr35171_pagefiles

MODULE 20 HOW GROUPS INTENSIFY DECISIONS 219

became more cautious after discussion. One of these featured “Roger,” a young married man with two school-age children and a secure but low-paying job. Roger can afford life’s necessities but few of its luxuries. He hears that the stock of a relatively unknown company may soon triple in value if its new product is favorably received or decline consid- erably if it does not sell. Roger has no savings. To invest in the company, he is considering selling his life insurance policy. Can you see a general principle that predicts both the tendency to give riskier advice after discussing Helen’s situation and more cautious advice after discussing Roger’s? If you are like most people, you would advise Helen to take a greater risk than Roger, even before talking with others. It turns out there is a strong tendency for discussion to accentu- ate these initial leanings; groups discussing the “Roger” dilemma became more risk-averse than they were before discussion.

DO GROUPS INTENSIFY OPINIONS? Realizing that this group phenomenon was not a consistent shift toward increased risk, we reconceived the phenomenon as a tendency for group discussion to enhance group members’ initial leanings. This idea led investigators to propose what French researchers Serge Moscovici and Marisa Zavalloni (1969) called group polarization: Discussion typically strengthens the average inclination of group members.

Group Polarization Experiments This new view of the changes induced by group discussion prompted experimenters to have people discuss attitude statements that most of them favored or most of them opposed. Would talking in groups enhance their shared initial inclinations as it did with the decision dilemmas? In groups, would risk takers take bigger risks, bigots become more hostile, and givers become more generous? That’s what the group polarization hypothesis predicts (Figure 20-1). Dozens of studies confirm group polarization.

• Moscovici and Zavalloni (1969) observed that discussion enhanced French students’ initially positive attitude toward their president and negative attitude toward Americans.

• Mititoshi Isozaki (1984) found that Japanese university students gave more pronounced judgments of “guilty” after discussing a traffic case. When jury members are inclined to award damages, the group award similarly tends to exceed that preferred by the median jury member (Sunstein, 2007a).

mye35171_ch20_217-232.indd Page 219 19/11/10 11:50 AM user-f494/208/MHSF219/myr35171_disk1of1/0078035171/myr35171_pagefiles

220 PART THREE SOCIAL INFLUENCE

Another research strategy has been to pick issues on which opinions are divided and then isolate people who hold the same view. Does dis- cussion with like-minded people strengthen shared views? Does it mag- nify the attitude gap that separates the two sides? George Bishop and I wondered. So we set up groups of relatively prejudiced and unprejudiced high school students and asked them to respond—before and after discussion—to issues involving racial atti- tudes, such as property rights versus open housing (Myers & Bishop, 1970). We found that the discussions among like-minded students did indeed increase the initial gap between the two groups (Figure 20-2).

Group Polarization in Everyday Life In everyday life people associate mostly with others whose attitudes are similar to their own. (Look at your own circle of friends.) Does everyday group interaction with like-minded friends intensify shared attitudes? Do nerds become nerdier and jocks jockier? It happens. The self-segregation of boys into all-male groups and of girls into all-female groups accentuates over time their initially modest gender differences, notes Eleanor Maccoby (2002). Boys with boys become gradually more competitive and action oriented in their play and fic- tional fare, and girls with girls become more relationally oriented. On U.S. federal appellate court cases, “Republican-appointed judges tend to vote like Republicans and Democratic-appointed judges tend to vote like

Before discussion

After discussion

0Neutral

Oppose

Favor Group A

Group B–

+

FIGURE 20-1 Group polarization. The group polariza- tion hypothesis predicts that discussion will strengthen an attitude shared by group members.

mye35171_ch20_217-232.indd Page 220 19/11/10 11:50 AM user-f494/208/MHSF219/myr35171_disk1of1/0078035171/myr35171_pagefiles

MODULE 20 HOW GROUPS INTENSIFY DECISIONS 221

Democrats,” David Schkade and Cass Sunstein (2003) have observed. But such tendencies are accentuated when among like-minded judges. “A Republican appointee sitting with two other Republicans votes far more conservatively than when the same judge sits with at least one Democratic appointee. A Democratic appointee, meanwhile, shows the same tendency in the opposite ideological direction.”

Group Polarization in Schools Another real-life parallel to the laboratory phenomenon is what educa- tion researchers have called the “accentuation” effect: Over time, initial differences among groups of college students become accentuated. If the first-year students at college X are initially more intellectual than the students at college Y, that gap is likely to increase by the time they graduate. Likewise, compared with fraternity and sorority members, independents tend to have more liberal political attitudes, a difference that grows with time in college (Pascarella & Terenzini, 1991). Research- ers believe this results partly from group members reinforcing shared inclinations.

Prejudice

Before discussion

After discussion

High-prejudice groups

Low-prejudice groups

4

2

0

3

1

–4

–3

–2

–1

FIGURE 20-2 Discussion increased polarization between homogeneous groups of high- and low- prejudice high school students. Talking over racial issues increased prejudice in a high-prejudice group and decreased it in a low-prejudice group. Source: Data from Myers & Bishop, 1970.

mye35171_ch20_217-232.indd Page 221 19/11/10 11:50 AM user-f494/208/MHSF219/myr35171_disk1of1/0078035171/myr35171_pagefiles

222 PART THREE SOCIAL INFLUENCE

Group Polarization in Communities Polarization also occurs in communities, as people self-segregate. “Crunchy places . . . attract crunchy types and become crunchier,” observes David Brooks (2005). “Conservative places . . . attract conservatives and become more so.” Neighborhoods become echo chambers, with opinions richocheting off kindred-spirited friends. One experiment assembled small groups of Coloradoans in liberal Boulder and conservative Colo- rado Springs. The discussions increased agreement within small groups about global warming, affirmative action, and same-sex unions. Never- theless, those in Boulder generally converged further left and those in Colorado Springs further right (Schkade & others, 2007). In the United States, the end result has become a more divided coun- try. The percentage of landslide counties—those voting 60 percent or more for one presidential candidate—nearly doubled between 1976 and 2000 (Bishop, 2004). The percentage of entering collegians declaring themselves as politically “middle of the road” dropped from 60 percent in 1983 to 45 percent in 2005, with corresponding increases in those declaring themselves on the right or the left (Pryor & others, 2005). On campuses, the clustering of students into mostly White sororities and fraternities and into ethnic minority student organizations tends to strengthen social identities and to increase antagonisms among the social groups (Sidanius & others, 2004). In laboratory studies the competitive relationships and mistrust that individuals frequently display when playing games with one another fre- quently worsen when the players are in groups (Winquist & Larson, 2004). During actual community conflicts, like-minded people associate increas- ingly with one another, amplifying their shared tendencies. Gang delin- quency emerges from a process of mutual reinforcement within neighbor- hood gangs, whose members share attributes and hostilities (Cartwright, 1975). If “a second out-of-control 15-year-old moves in [on your block],” surmises David Lykken (1997), “the mischief they get into as a team is likely to be more than merely double what the first would do on his own. . . . A gang is more dangerous than the sum of its individual parts.” Indeed, “unsupervised peer groups” are “the strongest predictor” of a neighbor- hood’s crime victimization rate, report Bonita Veysey and Steven Messner (1999). Moreover, experimental interventions that take delinquent adoles- cents and group them with other delinquents actually—no surprise to any group polarization researcher—increase the rate of problem behavior (Dishion & others, 1999).

Group Polarization on the Internet E-mail, blogs, and electronic chat rooms offer a potential new medium for like-minded people to find one another and for group interaction. On MySpace, there are tens of thousands of groups of kindred spirits dis- cussing religion, politics, hobbies, cars, music, and you name it. The

mye35171_ch20_217-232.indd Page 222 25/11/10 11:55 AM user-f469/208/MHSF219/myr35171_disk1of1/0078035171/myr35171_pagefiles

MODULE 20 HOW GROUPS INTENSIFY DECISIONS 223

Internet’s countless virtual groups enable peacemakers and neo-Nazis, geeks and goths, conspiracy theorists and cancer survivors to isolate themselves with like-minded others and find support for their shared concerns, interests, and suspicions (Gerstenfeld & others, 2003; McKenna & Bargh, 1998, 2000; Sunstein, 2001). Without the nonverbal nuances of face-to-face contact, will such discussions produce group polarization? Will peacemakers become more pacifistic and militia members more ter- ror prone? E-mail, Google, and chat rooms “make it much easier for small groups to rally like-minded people, crystallize diffuse hatreds and mobilize lethal force,” observes Robert Wright (2003b). As broadband spreads, Internet-spawned polarization will increase, he speculates. “Ever seen one of Osama bin Laden’s recruiting videos? They’re very effective, and they’ll reach their targeted audience much more efficiently via broadband.” According to one University of Haifa analysis, terrorist websites—which grew from a dozen in 1997 to some 4,700 at the end of 2005—have increased more than four times faster than the total number of websites (Ariza, 2006).

Group Polarization in Terrorist Organizations From their analysis of terrorist organizations around the world, Clark McCauley and Mary Segal (1987; McCauley, 2002) note that terrorism does not erupt suddenly. Rather, it arises among people whose shared grievances bring them together. As they interact in isolation from mod- erating influences, they become progressively more extreme. The social amplifier brings the signal in more strongly. The result is violent acts that the individuals, apart from the group, would never have committed. For example, the 9/11 terrorists were bred by a long process that engaged the polarizing effect of interaction among the like-minded. The process of becoming a terrorist, noted a National Research Council panel, isolates individuals from other belief systems, dehumanizes potential targets, and tolerates no dissent (Smelser & Mitchell, 2002). Over time, group members come to categorize the world as “us” and “them” (Moghaddam, 2005; Qirko, 2004). Ariel Merari (2002), an investigator of Middle Eastern and Sri Lankan suicide terrorism, believes the key to creating a terrorist suicide is the group process. “To the best of my knowledge, there has not been a single case of suicide terrorism which was done on a personal whim.” According to one analysis of terrorists who were members of the Salafi Jihad—an Islamic fundamentalist movement, of which al Qaeda is a part—70 percent joined while living as expatriates. After moving to foreign places in search of jobs or education, they became mindful of their Muslim identity and often gravitated to mosques and moved in with other expatriate Muslims, who sometimes recruited them into cell groups that provided “mutual emotional and social support” and “devel- opment of a common identity” (Sageman, 2004).

mye35171_ch20_217-232.indd Page 223 19/11/10 11:50 AM user-f494/208/MHSF219/myr35171_disk1of1/0078035171/myr35171_pagefiles

224 PART THREE SOCIAL INFLUENCE

Massacres, similarly, have been found to be group phenomena. The violence is enabled and escalated by the killers egging one another on (Zajonc, 2000). It is difficult to influence someone once “in the pressure cooker of the terrorist group,” notes Jerrold Post (2005) after interviewing many accused terrorists. “In the long run, the most effective antiterrorist policy is one that inhibits potential recruits from joining in the first place.”

EXPLAINING GROUP POLARIZATION Why do groups adopt stances that are more exaggerated than those of their average individual member? Researchers hoped that solving the mystery of group polarization might provide some insights into group influence. Solving small puzzles sometimes provides clues for solving larger ones. Among several proposed theories of group polarization, two have survived scientific scrutiny. One deals with the arguments presented during a discussion, the other with how members of a group view them- selves vis-à-vis the other members. The first idea is an example of infor- mational influence (influence that results from accepting evidence about reality). The second is an example of normative influence (influence based on a person’s desire to be accepted or admired by others).

Informational Influence According to the best-supported explanation, group discussion elicits a pooling of ideas, most of which favor the dominant viewpoint. Some discussed ideas are common knowledge to group members (Gigone & Hastie, 1993; Larson & others, 1994; Stasser, 1991). Other ideas may include persuasive arguments that some group members had not previ- ously considered. When discussing Helen the writer, someone may say, “Helen should go for it, because she has little to lose. If her novel flops, she can always go back to writing cheap westerns.” Such statements often entangle information about the person’s arguments with cues con- cerning the person’s position on the issue. But when people hear relevant arguments without learning the specific stands other people assume, they still shift their positions (Burnstein & Vinokur, 1977; Hinsz & others, 1997). Arguments, in and of themselves, matter.

Normative Influence A second explanation of polarization involves comparison with others. As Leon Festinger (1954) argued in his influential theory of social comparison, we humans want to evaluate our opinions and abilities

mye35171_ch20_217-232.indd Page 224 19/11/10 11:50 AM user-f494/208/MHSF219/myr35171_disk1of1/0078035171/myr35171_pagefiles

MODULE 20 HOW GROUPS INTENSIFY DECISIONS 225

by comparing our views with others’. We are most persuaded by peo- ple in our “reference groups”—groups we identify with (Abrams & others, 1990; Hogg & others, 1990). Moreover, wanting people to like us, we may express stronger opinions after discovering that others share our views. When we ask people (as I asked you earlier) to predict how others would respond to items such as the “Helen” dilemma, they typically exhibit pluralistic ignorance: They don’t realize how strongly others sup- port the socially preferred tendency (in this case, writing the novel). A typical person will advise writing the novel even if its chance of success is only 4 in 10 but will estimate that most other people would require 5 or 6 in 10. (This finding is reminiscent of the self-serving bias: People tend to view themselves as better-than-average embodiments of socially desirable traits and attitudes.) When the discussion begins, most people discover they are not outshining the others as they had supposed. In fact, some others are ahead of them, having taken an even stronger position in favor of writing the novel. No longer restrained by a misperceived group norm, they are liberated to voice their preferences more strongly. Perhaps you can recall a time when you and someone else wanted to go out with each other but each of you feared to make the first move, presuming the other probably did not have a reciprocal interest. Such pluralistic ignorance impedes the start-up of relationships (Vorauer & Ratner, 1996). Or perhaps you can recall a time when you and others were guarded and reserved in a group, until someone broke the ice and said, “Well, to be perfectly honest, I think. . . .” Soon you were all surprised to discover strong support for your shared views. This social comparison theory prompted experiments that exposed people to others’ positions but not to their arguments. This is roughly the experience we have when reading the results of an opinion poll or of exit polling on election day. When people learn others’ positions—without prior commitment and without discussion or sharing of arguments—they often adjust their responses to maintain a socially favorable position (Myers, 1978). This comparison-based polarization is usually less than that produced by a lively discussion. Still, it’s surprising that, instead of simply conforming to the group average, people often go it one better. Merely learning others’ choices also contributes to the bandwagon effect that creates blockbuster songs, books, and movies. Sociologist Matthew Salganik and his colleagues (2006) experimented with the phe- nomenon by engaging 14,341 Internet participants in listening to and, if they wished, downloading previously unknown songs. The researchers randomly assigned some participants to a condition that disclosed previ- ous participants’ download choices. Among those given that information, popular songs became more popular and unpopular songs became less popular.

mye35171_ch20_217-232.indd Page 225 19/11/10 11:50 AM user-f494/208/MHSF219/myr35171_disk1of1/0078035171/myr35171_pagefiles

226 PART THREE SOCIAL INFLUENCE

Group polarization research illustrates the complexity of social- psychological inquiry. Much as we like our explanations of a phenomenon to be simple, one explanation seldom accounts for all the data. Because people are complex, more than one factor frequently influences an out- come. In group discussions, persuasive arguments predominate on issues that have a factual element (“Is she guilty of the crime?”). Social compari- son sways responses on value-laden judgments (“How long a sentence should she serve?”) (Kaplan, 1989). On the many issues that have both factual and value-laden aspects, the two factors work together. Discovering that others share one’s feelings (social comparison) unleashes arguments (informational influence) supporting what everyone secretly favors.

GROUPTHINK Do the social-psychological phenomena we have been considering in the previous modules occur in sophisticated groups such as corporate boards or the president’s cabinet? Is there likely to be self-justification? self- serving bias? a cohesive “we feeling” promoting conformity and stifling dissent? public commitment producing resistance to change? group polarization? Social psychologist Irving Janis (1971, 1982) wondered whether such phenomena might help explain good and bad group deci- sions made by some twentieth-century American presidents and their advisers. To find out, he analyzed the decision-making procedures that led to several major fiascos:

• Pearl Harbor. In the weeks preceding the December 1941 Pearl Harbor attack that put the United States into World War II, military commanders in Hawaii received a steady stream of information about Japan’s preparations for an attack on the United States somewhere in the Pacific. Then military intelli- gence lost radio contact with Japanese aircraft carriers, which had begun moving straight for Hawaii. Air reconnaissance could have spotted the carriers or at least provided a few minutes’ warning. But complacent commanders decided against such precautions. The result: No alert was sounded until the attack on a virtually defenseless base was under way. The loss: 18 ships, 170 planes, and 2,400 lives.