Chapter 7 - Social Influence
THIS CHAPTER WILL DISCUSS:
1. How group decisions often tend to be more extreme than individual decisions.
2. How groups sometimes choose a weaker one of their options because they are not all aware of information supporting their best choice.
3. How normative and informational influence both explain these tendencies.
In this chapter, we shift attention from the issues of conformity and deviance, which concern group structure, and focus on the process of social influence, which occurs when group interaction causes members to conform.
We partially discussed social influence in Chapter 6. In many experiments that we have examined so far, however, group interaction was artificial. For instance, in some of the studies by Asch and by Sherif, interaction was limited to the expression of choices. In other experiments, also by Asch and by Moscovici and Schachter, interaction was artificial because the "choices" of confederates were not subject to group influence. These contrived circumstances hide insights about the process of social influence. Here, we look at studies that examine more natural group interaction. We can learn more about social influence by studying these naturally interacting groups.
We will look at decisions that naturally interacting groups make and then turn to proposals that account for this behavior. All these proposals relate to the process of social influence.
Natural Group Situations Versus Contrived Situations
We have stated that test groups can interact artificially or naturally. This distinction influences our ability to study the concept of social influence.
When groups are contrived, as in the experiment of Asch that we discussed earlier, interaction is not natural. One reason why Asch's groups were artificial was that social influence could occur in only one direction. The majority could, and often did, influence the deviant. The deviant, however, could never influence the majority because the majority was made up of confederates who had predetermined wrong judgments of the lines. They never changed their opinions. The deviant was the only real participant in the experiment.
In such situations, confederates lead the group discussion. Such artificial groups can tell us something about how social influence works, but they cannot tell us all.
Instead, we must turn to experiments that examine more natural, as opposed to contrived, group discussion. As we shall see, natural interaction reveals a great deal about the process of social influence in groups.
Effects of Social Influence
To prepare for our discussion, let us consider the effects of social influence. How might social influence affect the decisions of a group?
Imagine that someone asks a group of four people to make a decision. The group members can choose from among a number of options, and they each know all these options. Every member also has an opinion as to which option is best. Under these circumstances, group members are likely to have differences of opinion. During the meeting, the members state their opinions. They examine the options and discuss the reasons for their preferences. At the end of the meeting the group must decide which option is best.
Scientists interested in the process of social influence have studied this type of situation. In it, all individual opinions somehow become "transformed" into one group decision. How does this "transformation" take place? What is the process? Scientists have been trying to answer these questions.
In this chapter, we will read about some of the characteristics of this “transformational process.” We will also discover some of the proposed explanations for how a group moves from a tangle of individual opinions to one group decision. All proposals look at social influence. They each differ, however, in their hypotheses concerning the role of social influence in the "transformational" process.
Simpson study. One example of a more true-to-life study was performed by Simpson way back in 1939. Each member of a four person group first viewed a set of four pictures and ranked the pictures in order according to how much she or he liked them. Then they met together and talked as a group about how to rank-order the pictures. Finally, each person made a second individual judgment. It turned out that of the 108 people in the groups, the opinions of 106 converged toward the average ranking of the group members for the pictures from their first to the second individual rankings. Looked at another way, a measure of difference in opinion decreased an average amount of 36 percent between the two rankings.
Simpson also asked 24 other people to rank the pictures twice without discussing them as a group. He then treated these people’s rankings as if they were members of four-person “groups” so that he could measure how much they differed from one another, the same way he did for the rankings produced by members of real groups. In this case 18 of the people’s rankings converged from the first to the second ranking, with an average amount of 9 percent. This is a far less convergence than for the real groups. Simpson then concluded that, although thinking twice about the problem led to some convergence in individual rankings, having the opportunity to talk about it resulted in much more movement toward agreement.
Mock jury studies. For a second example, we return to the mock jury studies discussed in Chapter Two. Recall at that time our description of “social decision schemes.” Social decision schemes are rules that groups use to combine individual members’ decisions into a group decision. For example, if the odds that a group chooses a particular option are based on whether more than half of the members support it, then the social decision scheme the group is using is a “majority model.” On other words, a particular option will be chosen if a majority of the group favors it. If instead the probability that a group chooses an option is based on the proportion of members that favor it, the group is applying a “proportionality model.” For example, if three-quarters of the members of a group like a particular option, the odds that the group will choose that option is three-quarters. Social decision scheme theorists then propose mathematical equations to represent these and other models. They then see which equation best predicts group decisions. As we discussed in Chapter 2, research using mock juries shows that the “majority model” equations are normally the best predictors of jury verdicts. This finding implies that a “majority wins” rule is likely at work in this situation.
THE GROUP POLARIZATION EFFECT
From what we have seen so far, it seems that when group members begin their discussion with opinions that differ from one another, they end up with opinions that have become closer together. If the meeting starts with a majority of the group on one side of the issue, then the minority usually shifts to become closer to that majority. If there is not minority but, instead, different group members are on opposite sides, then all of them usually meet somewhere in the middle.
This means that if we can assign a number to represent each group member’s opinion at the beginning of the discussion, then we can use a simple “averaging” social decision scheme to predict what the group decision will be. If, on a scale ranging from 0 to 10, John starts the meeting at 2, Paul at 3, George at 4, and Richard at 7, then we would expect that the group decision will be the average of these numbers, which is 4. We would also hypothesize that each person’s individual viewpoint might have changed, particularly Richard’s.
It turns out that things are a little more complex than that. When scientists looked at groups in situations such as the one just described, they found a new phenomenon that has come to be called the "group polarization effect." To understand this effect, we need to examine the research that led to its discovery. Its discovery, in turn, led to some proposals regarding how social influence works.
A "Choice Dilemma"
Imagine that someone asked a four-member group to make the following decision:
Mr. Jones is a married man with two children. He has a steady job that pays him about $60,000 a year. He can easily afford the necessities of life, but few of the luxuries. Mr. Jones' father died recently. He carried a $40,000 life insurance policy. Mr. Jones would like to invest this inheritance in stocks. He is well aware of the secure "blue chip" stocks and bonds. They would pay approximately six percent on his investment. On the other hand, Mr. Jones has heard that the stock of a relatively unknown company, Company X, might double its present value. This could happen if the buying public favorably receives a new product which is currently in production. However, if the public does not like the new product, the stock would decline in value and Mr. Jones would lose his investment.
Imagine that you are advising Mr. Jones. You must choose the lowest probability that you consider acceptable before you would advise Mr. Jones to invest in Company X. For example, do you think that Mr. Jones should not invest in Company X under any circumstance? Do you think that Mr. Jones should invest in Company X only if the odds are 9 in 10 that the stock will double in value? Or will you accept odds of 5 in 10? 1 in 10? What odds would you be willing to accept?
As you can see, the four members of the group face a number of different probabilities that Company X will succeed, and they must choose from among them. Each member has an opinion about the lowest probability of success that he or she would accept before advising Mr. Jones to invest in Company X. The members will probably disagree about the lowest acceptable probability. Some might think that even if the odds are only 3 in 10 that the stock will do well, the opportunity is too good to pass up. They would consider it a good gamble. Others will find these odds too chancy. They might want odds of 7 in 10. During the meeting, the group members state their opinions and the reasons behind them. By the end of the meeting they must come to a decision concerning the lowest probability that the group would accept before advising Mr. Jones to invest in Company X.
Scientists interested in social influence have studied this exact situation. Researchers have come to call this type of problem a "choice dilemma." Faced with choice dilemmas, groups must choose between two options. One option has an attractive outcome but only some probability of success. The other option has a less attractive outcome but will definitely succeed. For example, a group must choose between traveling a long way to a stadium in hopes of getting tickets to a popular baseball game or staying at home and watching the game on television. If they travel to the stadium, they may miss the game entirely. Staying at home, however, is not as much fun. The choice dilemma involves an attractive, risky plan and a safe plan. They need to examine the odds that they will be able to get tickets and then decide what to do.
Now let us go back to the choice dilemma facing the group that needs to advise Mr. Jones. In that dilemma, Company X is the more attractive proposition. Investing in Company X is an attractive idea, but it is a risky choice. The other option, investing in blue-chip stocks and bonds, is less attractive but safer. It is an assured option. Mr. Jones would know exactly what he was getting if he bought stocks and bonds.
The group must decide the odds for success that they would require from the more attractive but chancier option, Company X, before they recommend it.
Definition of "Risky" and "Cautious" Decisions
Before we look at how a "natural" group might deal with a choice dilemma, let us define some terms.
Imagine that the group decides to advise Mr. Jones to invest in Company X as long as the odds are 2 in 10. This means that the group is willing to allow Mr. Jones to take a chance with Company X even though the company is unlikely to succeed. The group has made a "risky" choice.
In contrast, the group advises Mr. Jones to invest in Company X only if the odds of success are 8 in 10. In this case, the group is willing to recommend Company X only if the company is highly likely to succeed. The group has made a "cautious" decision.
In general, a group is risky if it recommends the more attractive but unsure option at odds of success that are less than 5 in 10. In other words, if the choice has a less than 50-50 chance for success and the group still recommends it, the group is being risky. A group is being cautious, however, if it decides to recommend the more attractive but chancier proposition at odds of success that are more than 5 in 10.
What would a "natural" group do with a choice dilemma? We can think of several possible scenarios. Let us hypothesize about some of them.
In the first scenario, all members of the group have the same opinion. The group will almost always decide on that opinion. Thus, if all four members of our example group wanted the odds to be at least 6 in 10, the group decision would be 6 in 10.
In the second scenario, some group members tend toward risk and some tend toward caution. The group decision will probably be a compromise. We could estimate that compromise by figuring the arithmetic average of each member's individual opinion. For instance, two members of our group want odds of 7 in 10, a cautious choice, and the other two want odds of 3 in 10, a risky choice. The arithmetic average is the probability of 5 in 10. Hence, we would expect the group to compromise and agree that they could recommend Company X if the probability is 5 in 10.
In a third scenario, all group members are on one side, either risk or caution. They disagree, however, about the exact probability they would need to choose the more attractive but unsure option. Once again, we can use math to predict the group's compromise. For example, two members want the probability to be 4 in 10, and two want it to be 2 in 10. They clearly all agree on a risky option but disagree on the acceptable level of risk. By averaging their opinions, we would predict that the group would most likely agree to a 3 in 10 probability of success.
What decisions did the "natural" groups actually make with choice dilemmas? Were the three scenarios we hypothesized correct?
Researchers have experimented to see what actually occurs when groups make decisions about choice dilemmas. Many of these studies applied a particular experimental method. It was first used in a study of choice dilemmas by Wallach, Kogan, and Bem (1962).
In their study method, a number of participants first worked alone and chose acceptable odds for each of a series of 12 choice dilemmas including the one about Mr. Jones and his investments. We will label each participant's first decision the "prediscussion opinion."
Next, the researchers placed the participants in groups, and each group made a decision about acceptable odds. Finally, each participant made one last individual decision about which odds would be acceptable, again working alone. We will label the participants' last choice the "postdiscussion opinion."
What did the researchers find? One significant finding related to the third scenario above. As you will recall, in that scenario all group members are on the same side, either risk or caution. They disagree, however, about the probability they need to recommend the more attractive, but hazardous, option. We hypothesized that we could mathematically average the opinions of the group members and that this average would indicate what they would eventually agree upon.
The research findings, however, showed this idea to be false. What we expected did not happen.
For example, we are looking at the results of a group of four people in the experiment. The prediscussion opinions of two members leaned toward risk, but not too much risk. They would advise Mr. Jones to invest only if the odds were at least 4 in 10 that Company X will succeed. The other two members of the group were rather bold and would advise Mr. Jones to invest if the odds were only 2 in 10 for success.
What happened when these four people came together in a group? Earlier, we hypothesized that the group would compromise. In this case, we might expect the members to decide on the mathematic average of their prediscussion opinions. This would work out to the odds of 3 in 10.
In the Wallach, Kogan, and Bem experiment, however, this did not happen.
In 10 out of the 12 choice dilemmas that the group examined, the eventual decision was riskier than the mathematical average would predict. For example, the group that we imagined above might decide that Mr. Jones should buy the stock at lower odds than the mathematic average of 3 in 10. They might decide on the lower, more daring odds of only 2.5 in 10.
Researchers have called this phenomenon the "risky shift."
In addition, researchers found that the group meetings appeared to have a lasting influence on members. The participants' postdiscussion opinions were riskier than their prediscussion ideas. Further, these postdiscussion opinions approximated the group decision. Researchers found this true both immediately after the group meeting and several weeks later when they again asked the participants for their postdiscussion opinions. Such a result is evidence that members did not merely comply with their groups. Instead, an actual opinion change took place.
The study also examined control groups, consisting of people who voiced only two individual opinions, with no intervening group discussion. In this situation, no shift took place. Thus, the research showed that members need some experience of group discussion before risky shifts can take place.
Soon enough, scientists discovered that groups do not always experience risky shifts. Sometimes what happens is a "cautious shift." The cautious shift occurs, as its name implies, when groups make more cautious decisions than individuals. For example, two members of a four-person group individually decide to advise Mr. Jones to make the risky investment at odds of 6 in 10. The other two members decide to advise him to make the investment at odds of 8 in 10. When the group meets, its decision is more cautious than the average of their individual opinions. Perhaps the group decides on the odds of 7.5 in 10. It took a few years for the results of the cautious shift to become evident to researchers.
Remember that only 2 of the 12 choice dilemmas in the Wallach et al. study failed to result in risky shifts. The experimenters considered these outcomes flukes. Subsequent research, however, found these results genuine. Further studies used the 12 choice dilemmas from the Wallach experiment and found that the same two items consistently led to cautiousness. Why would this happen? Certain themes ran through these two dilemmas that caused participants to be careful. Scientists took this knowledge and soon learned how to write choice dilemmas that would lead to either risky or cautious behavior. For example, experimenters learned that they could create a cautious shift if they described the hardships that would befall the protagonists' dependents if the risky option should fail (Rabow et al., 1966).
Researchers also learned that risky shifts occur when group members make daring choices from the start. Cautious shifts occur when members are initially rather careful. These two results come together in the overall finding of the group polarization effect. This effect takes place when groups have members who lean toward one direction, either risk or caution, but do not agree on the same odds. Such groups tend to make decisions in the direction that the members favor. Group decisions, however, are usually more extreme than the individual members' prediscussion judgments.
Groups with members who lean toward the risky direction make decisions that are more daring than the individual members would make. People who are initially cautious make extremely careful judgments when they are in groups.
For example, four students have a group project for a class. They must decide between two projects. One is sure to succeed, and they are guaranteed a grade of C. This project is not very imaginative, however, so the highest grade they could get would be a B. The other project might fail, in which case they would receive a D. This project is very imaginative, however, and if it worked out, they would get an A.
According to the group polarization effect, if at the beginning of their discussion the group members leaned toward the guaranteed but unimaginative project, by the end of the meeting they would be strongly in favor of it. This would be a cautious shift. If instead they began the discussion slightly in favor of the chancy but imaginative project, afterward they should be enthusiastic about it. This would be a risky shift. In both cases, the students' opinions after the discussion were in the same direction, but more extreme than they were before the discussion.
The group polarization effect was an exciting and surprising discovery for the researchers who looked at choice dilemmas. Scientists subsequently have performed hundreds of studies exploring it.
Group Polarization and Social Influence
The size of this work, however, is not due strictly to the surprise effect of polarization. Another reason experimenters have devoted a great deal of time to the polarization phenomenon is that they hope it will help them understand a larger topic. Theorists feel that polarization is a "window" into the larger world of social influence and how it affects group decision making. The hope is that the study of polarization can lead to a general theory of social influence in groups.
We have centered our discussion on the choice dilemma. Polarization effects also occur in tasks other than those associated with risk. For instance, polarization occurs when groups deal with attitude extremity, when they guess light "movement" in autokinetic research, when they are involved in gambling, and when they need to decide how difficult a problem can get before they are no longer willing to tackle it. Thus, the research regarding polarization has been very extensive, and experiments have involved many tasks.
Predicting Group Polarization
One example of this research was performed by Davis, Kerr, Sussman, and Rissman (1974). They used social decision schemes to examine what happens when a group polarizes around either a risky or a cautious option.
The majority model. Specifically, they wanted to see which social decision scheme was best able to predict group polarization. Their findings were similar to what Davis discovered when he looked at juries. Again, the majority model was the best predictor. The group is likely to polarize around the option that the majority prefers.
For example, in a four person group all members lean toward caution but differ in their prediscussion opinions. Two members want odds of 8 in 10, another member wants odds of 7 in 10, the other, the least cautious, wants 6 in 10. The average of these odds is 7.25 in 10. The majority model, on the other hand, predicts that the group's decision will be for the odds of 8 in 10.
If the majority model is correct, the group's decision will be more extreme, in the direction of either risk or caution, than the mathematical average of the members' prediscussion opinions. In essence, in this circumstance, the majority model predicts a polarization effect.
This is exactly what happened in the Davis et al. study for both cautious and risky shifts.
Limits of the majority model. These findings alone, however, do not indicate that the majority model will always predict the group polarization effect. It is easy to think of circumstances in which the majority model does not do so. In our example, the majority opinion was also the most polarized. The two people who agreed on the odds of 8 in 10 made up the majority, and their opinion was also the most extreme, or polarized. In this case, the majority model predicted group polarization correctly. If the majority opinion was not also the most polarized, however, difficulties arise. In these circumstances, the majority model would not predict group polarization.
In an alternative group, two members want odds of 6 in 10. They make up the majority opinion. One other member wants odds of 7 in 10, and the last wants even more cautious odds of 8 in 10. All members are on the side of caution; however, the members in the majority are the least cautious. In this case, the arithmetic average is 6.75 in 10. The majority model, on the other hand, predicts a group decision for the odds of 6 in 10. Hence, the majority model does not predict group polarization, but a decision that reflects the opinions of the least cautious members.
Therefore, the majority model predicts that polarization will take place only if the group majority includes the most polarized members. This happened in the Davis et al. study. The group majorities usually included the most polarized members. That was why the majority model was a successful predictor of group polarization in their study.
Moreover, this is what appears to occur in general. When facing choice dilemmas, groups usually have a majority of members with a more extreme view than the minority. As a result of group discussion, the minority members’ opinions converge toward those of the majority, in so doing becoming more extreme. Therefore, what happens in group polarization is consistent with what happens in other situations in which social influence takes place; individual member opinions converge, and minorities move toward majorities.
General Social Influence Theories
In the past three decades, two theories attempting to explain how group polarization works have come into general acceptance among scientists. Further, scholars believe that these proposals go beyond choice dilemmas and relate to all types of group decisions.
Why would they think this? As we know, scientists first discovered the group polarization effect when they studied choice dilemmas. After examining group polarization, scholars came to believe that it is a window into the larger process of social influence. Hence, by looking at group polarization, they could discover more about how social influence affects all kinds of group decision-making tasks. To look at group polarization, however, scholars still used choice dilemmas as a basis for research and proposals. In essence, choice dilemmas help scholars examine and theorize about group polarization, and group polarization helps them look at social influence.
For all these reasons, proposals that relate to choice dilemmas also relate to all types of group decisions. In particular, the two theories reflect larger ideas about how social influence works in all types of group contexts. Next, we will turn to a discussion of both these larger ideas about social influence and of the two theories based on those ideas that are specifically relevant to group polarization.
The first of the larger ideas about how social influence works in discussion groups is called normative influence. Normative influence occurs when group members listen to one another express opinions about the available options. The claim is that members come into a group with prediscussion opinions. During their meeting, members share these opinions. Through this process, each person learns the other members' preferences. We would then hypothesize that members are likely to adopt the option of the majority. For example, Tom hears that others in his group like Option A but dislike Option X. If normative influence occurs, this alone would be enough to cause Tom to favor Option A. In essence, social influence occurs because people want to agree with the majority. In the end, the group chooses the majority viewpoint.
An important aspect of normative influence is that social influence occurs when people believe that the majority viewpoint is correct just because it is the majority. In the above example, Tom thinks to himself that Option A must be better than Option X because he respects all of the other members in his group, so he believes that their opinion must be correct. It is important not to confuse normative influence with compliance, as we discussed it in Chapter 6. When compliance occurs, group members voice the majority position but do not believe it. With normative influence, group members come to privately accept the majority point of view.
Research in Support of Normative Influence
We have already run across several examples of social influence in which theories consistent with normative influence.
In Chapter 5, for instance, we discussed research related to expectation states theory (Berger, Fisek, Norman, & Zelditch, 1977). Participants who believed that they were lower than others in an imaginary skill performed phony tasks. The skill might be something like "contrast sensitivity." The phony task supposedly related to the imaginary skill. Researchers showed these participants answers that supposedly came from people who were allegedly more skilled at the task than the participants were. When these answers disagreed with the participants' decisions, the participants were likely to change their minds and agree with the answers they saw. In other words, merely learning the other opinions could influence a participant's decision. Thus, the results of expectation states research are consistent with the existence of normative influence.
Much of the research we discussed in Chapter 6 also seems to reflect normative influence. For example, Asch's (1956) study of conformity found that participants agreed with the majority opinion even when that opinion was wrong. As you can recall, participants looked at the lengths of lines, and 33 participants conformed to the confederates' wrong answers in at least two-thirds of the trials. In interviews after the experiment, about half these conformists claimed that they saw the line correctly but that they decided their own perceptions must have been wrong when they heard the majority choice. They went along with the majority judgment even though it was wrong.
In Moscovici, Lage, and Naffrechoux's (1969) study of minority influence, researchers saw that a minority opinion can also affect group members. Confederates, who were in the minority, claimed that a series of blue slides were green. These statements influenced participants when they later had to judge a second series of slides, which were of a color they could easily call either blue or green.
In all three cases, participants heard only the judgments that other people made, not arguments about those judgments. The mere act of hearing the opinions influenced them, a finding consistent with the presence of normative influence.
Jury Decision Models
Some mathematical models of jury decision making differ from those we have studied so far in that they do not use social decision schemes. These models assume that jury deliberation is important. During deliberation, jurors can be persuaded to change their initial opinions. Scientists have attempted to represent this deliberation process mathematically.
In addition to showing that the majority tends to persuade the minority most of the time in a jury, research has revealed a "momentum effect": When members of the minority start to join the majority, other minority members begin to agree with the majority at an ever-faster rate. The momentum slows for the last one or two people in the minority, however. They are usually more sure of their opinions, and persuading them to change their minds is more difficult (Penrod & Hastie, 1980).
These models are consistent with the impact of normative influence because of the way they hypothesize that social influence works. The models assume that social influence occurs when jurors hear one another's opinions. The models concern the number of jurors originally on each side of an issue. They estimate the probability that jurors will change their minds and how quickly that will happen, given the number of jurors on each side. The models implicitly assume that to predict changes in jurors' minds they need look only at the opinions the jurors expressed to one another.
The Social Comparison Approach to Group Polarization
A specific theory consistent with normative influence that has been applied to the study of group polarization is social comparison theory. We discussed social comparison theory in Chapter 6 as an explanation for conformity. In that theory, Festinger distinguished between "beliefs about abilities" and "opinions." He stated that beliefs about abilities have a good/bad dimension. People want to move toward the good end of the scale and away from the bad. In contrast, Festinger believed that opinions have no analogous good/bad dimension. Instead, people must choose the group consensus as a norm to follow and shift toward that norm if they want to feel that they have correct opinions.
Baron et al. Study. Baron, Dion, Baron, and Miller (1971) performed a study that they believed to demonstrate the power of normative influence in groups. As we mentioned earlier in this chapter, scientists have learned how to write choice dilemmas that lead predictably to either risky or cautious shifts. In the Baron et al. study, single participants was placed in four-member groups. The other three members of the groups were confederates instructed to argue either for risk or for caution. In one-fourth of the discussions, the group was given a choice dilemma normally leading to cautious shifts and the confederates argued for caution. Not surprisingly, the participants’ opinions shifted toward greater caution. In another one-fourth, the group was given a choice dilemma usually tending toward risk, and the confederates argued for risk. Again as one would expect, the participants’ views moved toward greater risk.
Two other circumstances were more critical. In a third one-fourth of the discussions, the group was faced what was typically a cautious-shift-provoking choice dilemma, but the confederates argued for risk. In the final fourth, the participants read a risky-shift-oriented choice dilemma but heard cautious-arguing confederates. In these two cases, would the participants shift their opinions toward the item or toward their groupmates?
The answer was their fellow group members. Baron et al. interpreted this finding as consistent with the idea that, when facing a choice dilemma, group members look to one another for the correct option, and move their opinion to the majority point of view. This interpretation is consistent with normative influence.
Research into the Theory
The social comparison approach to group polarization assumes that all that is necessary for social influence to occur in a group is that members hear one another express their opinions. Group members need not hear arguments about the advantages or disadvantages of either risk or caution to change their opinions toward the group majority. All that they need to hear is the prediscussion opinions of the other members. Each person can then decide whether his or her opinion is consistent with the rest of the group. This implies that group polarization will still occur in groups in which the members cannot argue for a particular option.
Scientists can artificially constrain group discussion so that members can voice only their prediscussion opinions. They cannot discuss arguments for or against either risk or caution. Teger and Pruitt (1967) placed these artificial constraints on groups and found that group polarization did occur under these circumstances. The amount of polarization, however, was not as great as in "normal" discussion groups.
Teger and Pruitt's results imply that social comparison may partly account for the group polarization effect. Social comparison is not the entire reason, however, because polarization was smaller than usual in their constrained groups. In the same vein, Isenberg (1986) reviewed relevant studies and found that social comparison had a consistent but only moderate effect on group polarization in those studies.
As we know, to understand social influence completely, we need to understand why the group polarization effect occurs. We have seen how the research comes close to supporting the social comparison approach to group polarization. Therefore, normative influence appears to be a viable but incomplete explanation for how social influence occurs in groups.
Figure 7.1 diagrams the social influence process from the standpoint of social comparison theory. The process is divided into input, process, and output stages. These stages are equivalent to the stages in the input-process-output model we discussed in Chapters 1 and 2. In this and subsequent diagrams in this chapter. the input steps occur before group discussion begins, the process steps occur during discussion, and the output steps are the results of discussion.
The second of the larger ideas about social influence in groups is called informational influence. Informational influence can occur when group members learn new information about the available options. As with normative influence, the claim is that members come into a group with prediscussion opinions. In contrast, the hypothesis is that the group members talk about their opinions during their meeting. Their discussion centers on the advantages and disadvantages of each option. Through this type of discussion, each member learns more about the options. This new information often causes members to change their opinions. The group then uses the members' new opinions to help it decide on the best option.
In Contrast with Normative Influence
Comparing normative and informational influence helps us to understand the latter. As you can recall, normative influence assumes that social influence occurs when people learn the opinions of others and, as a result, change their minds to agree more with them. Proponents of normative influence do not focus on what happens when a group discusses the strengths and weaknesses of options. They do not think this is significant in the social influence process.
Supporters of informational influence, on the other hand, look carefully at what happens when group members argue about the strengths and weaknesses of their options. They believe that this discussion is responsible for social influence and do not feel that group members will change their minds merely because they learn other members' opinions. They believe that group members need to have new information before social influence can occur.
Research in Support of Informational Influence
One example of research that supports the idea of informational influence is Sherif's (1935) study of how people judged imaginary "movements" of a stationary point of light. Participants formed standards for judgment about the range of the light's "movements." When two or three participants made judgments in the presence of one another, their standards converged over time. Although they were not presenting “arguments” as such, these participants’ verbalized judgments can be looked at as information about light movement that persuaded one another to adopt the same judgmental standard. A second example also covered last chapter was minority influence. It may not happen a lot, but when it does, by definition it is influence contrary to the group “norm.” Therefore, it is almost certainly a product of informational influence.
Outside of what we have covered in this book, there are many examples of information influence in the work that scientists have done in the area of persuasion. Scientists have performed a number of studies to determine whether communicative messages that include evidence as part of a persuasive appeal are more successful in changing opinions than messages that do not contain such arguments.
For example, a communicator is trying to persuade you to brush your teeth every day. One of her claims is that, if you don't do this, your teeth will rot and fall out. In essence, the person has expressed an opinion. If you don't brush daily, your teeth will fall out. Will this opinion be more likely to persuade you if evidence backs it up? This is the question that interests scientists. They want to know if the communicator will be more likely to persuade you to brush your teeth every day if she presents evidence supporting the claim that unbrushed teeth rot and fall out than if she does not present such evidence. Reinard (1988) has reviewed many such studies, which indicate that evidence does increase the persuasive power of messages. These findings are consistent with the importance of informational influence. They show that a person wanting to change opinions effectively, needs to present arguments for an opinion. Simply expressing that opinion is not enough.
Now let us examine how this idea relates to the group polarization effect.
The Persuasive Arguments Approach to Group Polarization
Burnstein and Vinokur (1973) proposed the persuasive arguments approach, which claims that, when group members must choose from among a number of options, they all know something about each option. Every person has bits of knowledge that can be viewed as arguments either for or against each option. Members base their prediscussion opinions on these arguments.
During group discussion, members talk over the familiar arguments. They support or criticize each option, hearing arguments they had not heard before. They learn new information that gives them reasons to polarize their opinions further. At the end of the discussion, the group bases its decision on the members' further-polarized opinions. Consequently, the group decision will be more polarized than an average of its members' prediscussion opinions.
Example of Process
Let us illustrate this process. Sandra and Harvey face a choice dilemma individually. Sandra thinks of three reasons (A, B, and C) to be risky. She can think of only two reasons (M and N) to be cautious. Therefore, Sandra chooses to be slightly daring and selects the slightly risky odds of 4 in 10. Harvey also leans toward risk but sees other reasons for his choice. He finds reasons A, B, and D for daring behavior and finds reasons N and O to be cautious. He therefore also chooses the slightly risky odds of 4 in 10.
When these two meet in a group, they learn from each other new reasons to be risky. Sandra hears Harvey's reason D, and Harvey hears Sandra's reason C. Thus, they both have four reasons to be daring--A, B, C, and D. Each now has more cause to be risky. Their personal opinions polarize. Perhaps they now favor riskier odds of 3 in 10. Their group decision will then be riskier than their prediscussion opinions.
Research into Proposal
The persuasive arguments approach to the group polarization effect assumes that the single factor of members learning new information about the group's options can cause social influence. Burnstein and Vinokur did not see a sharing of opinions as a significant part of the process that changes group members' opinions. They believed that new information is sufficient to cause social influence. In addition, an advocate of persuasive arguments theory might argue that the Baron et al. findings described above are consistent with it, and not with social comparison theory. In that study, the confederates presented a number of arguments for their point of view. This means that, for example, when faced with a choice dilemma implying cautious values but confederates instructed to support risk, the real participants are hearing arguments in favor of risk, some of which might be new to them. It could be the impact of these arguments that lead the participants to become more risky, and not just, as Baron et al. supposed, just the fact that the majority of the group favors risk.
There has been a great deal of research examining the persuasive arguments theory. In general, the results of this research have been supportive. For example, Ebbesen and Bowers (1974) counted the number of statements that groups made in support of the risky and cautious alternatives when talking about choice dilemmas. They discovered that the more that the discussion arguments favored risk, the more risky shift occurred in the group members’ opinions. Analogously, the more that the discussion arguments tilted toward caution, the more cautious the participants became. In addition, St. Jean (1970) restricted group discussion to arguments and forbid group members to state their opinions. His groups polarized anyway.
Isenberg (1986) reviewed many studies like these and found a strong relationship between the effects of persuasive arguments and group polarization. Most theorists now believe that a good theory about both group polarization in specific and social influence in general must include some aspects of persuasive arguments.
Limits of the Proposal
However, persuasive arguments theory has problems accounting for certain research data. These problems come from one underlying assumption of the persuasive arguments approach. The approach presumes that the arguments that are voiced during discussion are representative of the arguments with which the members of the group are familiar. As a consequence, the proportion of arguments on either side of the issue that come up during discussion equal the proportion of arguments that members know on both sides. In our example, sixty percent of the arguments that both Sandra and Harvey are familiar with are in favor of risk and forty percent are for caution. The persuasive arguments approach implies that sixty percent of the arguments that come up during their discussion will also be for risk, and forty percent of the arguments will favor caution.
However, there is strong evidence suggesting that this assumption is wrong.
Bias in discussion. As just mentioned, persuasive arguments theory presumes that the proportion of arguments that are brought up during discussion on either side of the issue is about the same as the proportion of arguments on either side with which members of the group are familiar. If so, then discussion should also include arguments new to each member on the disfavored side of the issue. Sandra will voice reason M to be cautious some time during the discussion, and Harvey will voice reason O. As a result, both will learn new arguments in favor of caution along with new arguments in favor of risk. These new arguments on either side of the issue will approximately cancel one another out, leaving Harvey and Sandra’s opinions unchanged. Therefore, group polarization will not occur. In fact, group polarization could only occur if the assumption of equal proportions of arguments is wrong and arguments are biased in favor of the side that the members originally favor.
There are studies that suggest that discussion in biased in this manner. One of them was performed by myself (Pavitt, 1994). For three choice dilemmas, some of the participants were asked to write down the arguments they knew on both sides of the issue. Then they made a decision about each dilemma in three- or four-member groups. A content analysis was performed in order to determine the proportion of arguments on both sides of the issue in both the written lists and the discussion. The discussion was consistently more polarized than the written lists. When participants leaned toward risk, 57 percent of their written arguments but 86 percent of their voiced arguments favored risk. When participants leaned toward caution, 60 percent of their written arguments but 83 percent of their voiced arguments favored caution. This finding implies that the discussion is biased in favor of prediscussion opinions.
Why do group members seem to talk so much about ideas that favor the group opinion? Why don't they instead reveal more of their personal arguments on the disfavored side? The persuasive arguments theory cannot answer these questions. It does not account for the disproportion between personal opinion arguments and group decision arguments.
It seems that the persuasive arguments proposal can account for some outcomes but not all. It does explain the effect of arguments during group discussion. It cannot reveal, however, why the group tends to verbalize only those arguments that favor the original group option. Therefore, it cannot account for the hidden profile effect.
Figure 7.2 diagrams the social influence process from the standpoint of persuasive arguments theory.
OF INDIVIDUALą PREDISCUSSION
NORMATIVE AND INFORMATIONAL INFLUENCE COMBINED
We have seen that some scientists have hypothesized that opinion expression is the key to social influence in groups while others have theorized that new information can cause group members to change their minds. Both viewpoints have a lot of support in research, which leads us to conclude that both are probably involved in social influence when groups discuss choice dilemmas. There is good reason to conclude that social influence includes both normative and informational aspects working together.
We can propose a model for how this occurs. It begins with the proposal from persuasive arguments theory that people think of arguments in favor of both the risky and cautious choice dilemma options and decide to favor the option that has the greater number of favorable arguments. However, unlike persuasive arguments theory, we do not presume that the arguments that come up during discussion are representative of all the arguments known by group members. Instead, members only bring up arguments that are consistent with what they think is the better option. Assuming that there is a group majority supporting one of the options, which usually happens with choice dilemmas, group discussion will then be biased in favor of that option. As a result, although group discussion leads to members learning new arguments relevant to their decision, they only learn more about the option that the members originally favored. Therefore their opinions become more extreme, and group polarization occurs.
In addition to discussing arguments, group members also exchange their prediscussion preference for one of the two options. In fact, this is usually the first thing that happens during group discussion, before any relevant arguments are mentioned. If group members see that there is a group majority, normative influence is bound to happen. In fact, perhaps one of the reasons that the arguments that come out during discussion are so biased in support of the proposal that the majority favors is that everyone knows that this is group’s favored option.
The Proposal and Group Polarization
Let us see how this proposal would explain the group polarization effect.
As you can recall, in a previous example, Sandra and Harvey face a choice dilemma. Sandra thinks of three reasons (A, B, and C) to be risky. She can think of only two reasons (M and N) to be cautious. She decides to select the slightly risky odds for success of only 4 in 10. Harvey also thinks of three reasons to be risky. Two (A and B) are the same as Sandra's, but the third (D) is different. He comes up with two reasons (N and O) to be cautious. He also decides to accept the slightly risky odds of 4 in 10. According to persuasive arguments theory, when Sandra talks with Harvey, she learns another argument for risk, D. Harvey also learns another argument for risk from Sandra, C. Together they both now have four reasons to be risky--A, B, C, and D. Because they both now have more reason to be risky, their opinions polarize toward more risk.
Persuasive arguments theory alone, however, cannot explain why Sandra and Harvey do not also talk about their other arguments, which support caution. If they were to talk about these, Sandra would learn a new cautious argument (O) from Harvey, and Harvey would learn a new cautious argument (M) from Sandra. Together they would now have three reasons to be cautious: M, N, and O. These cautious arguments would counteract the persuasive effect of the new risky arguments they learned from each other. Therefore, their opinions would not polarize.
What then accounts for the polarization? First, because both Sandra and Harvey decide to favor risk, they will tend to only bring up arguments that support their decision. Second, even before bringing up those arguments, they will exchange their preference for 4 out of 10 odds. According to social comparison theory, their original decision to support a risky choice is reinforced by knowing that they are in agreement about that. Thus, they never mention the arguments for caution. In this way, our proposal accounts for the group polarization effect. By combining social comparison theory with the persuasive arguments approach, we show how group polarization can occur.
Figure 7.3 diagrams the social influence process from the standpoint of the combined social comparison/persuasive arguments proposal. Note that this diagram implies a “dual-process” account. The exchange of prediscussion opinions has a direct effect on postdiscussion opinions due to normative influence. However, the trading of prediscussion opinions is also followed by the discussion of relevant arguments, which then impacts on postdiscussion opinions through informational influence. As the proportion of relevant arguments occurring during discussion is usually consistent with the majority prediscussion viewpoint, both types of influence normally reinforce one another in their effect on postdiscussion opinions and eventual group decisions.
SIDE OF ISSUE
THE HIDDEN PROFILE EFFECT
Thus far, we have seen that both normative and informational factors appear to be involved in social influence during the discussion of choice dilemmas. As scientists see the study of choice dilemmas as providing insight into more general social influence processes, it should not be surprising to find that both forms of persuasion are also relevant in other circumstances. Let us now turn to another often-researched situation.
Suppose that Jeff, Betsy, and Becky are trying to choose the best candidate for student body president. Jeff is aware of four reasons to favor Candidate A (W, X, Y, and Z) and two reasons to favor Candidate B (A and B). He therefore leans toward Candidate A as his prediscussion opinion.
Betsy has also come up with four reasons to support Candidate A (W, X, Y, and Z) and two reasons to favor Candidate B (C and D). Thus, she also prefers Candidate A.
Similarly, Becky knows two reasons in support of Candidate A (W, X, Y, and Z) and two arguing for Candidate B (E and F). As with her groupmates, she would choose Candidate A.
As you can see, all three of them have thought of the same four arguments favoring Candidate A. On the other hand, they thought of completely different arguments supporting Candidate B. Together they have a total of six such arguments: A, B, C, D, E, and F. Because together they have more arguments for Candidate B than Candidate A, will they all change their minds after they talk to one another? Yes, according to persuasive arguments theory. This theory hypothesizes that the three of them would talk openly with one another and learn all about the arguments for risk that they all know. In turn, this would cause them to change their preferences from Candidate A to Candidate B. No, according to social comparison theory. Because each favor Candidate A, they would begin discussion by exchanging those preferences and learn that they agree. This would reinforce their original preferences, and consequently they would limit their discussion to the four reasons for supporting Candidate A.
In order to see which theory’s prediction would be supported, Stasser and Titus (1985) examined circumstances such as this. In their study, groups were asked to decide which of three hypothetical candidates would be the best student body president. In the groups, each member was given a list of information about the candidates. Some of this information was “shared”; it was given to all of the group members. The rest of the information was “unshared”; provided to only one of the group members. The lists were manipulated so that the shared information all favored one candidate but the unshared favored a different one. Further, each member had more pieces of shared information than unshared. Therefore, the information each member was individually aware of favored the weaker candidate. For this reason, social comparison theory would predict that all the members would choose that candidate for their prediscussion opinion. However, if the group “pooled” all of the information they were given, they would then learn that a different candidate was really the best. In that case, the group members would change their minds and decide in favor of that stronger candidate. This latter result follows from persuasive arguments theory.
Stasser and Titus found that groups did not successfully pool all of their information. Instead, each member continued to support the weaker candidate, and the group made the wrong choice. Stasser and Titus’s finding makes sense if the members of these groups only discussed the arguments on the side of the weaker candidate that each member originally favored and the reasons for choosing the stronger candidate were never mentioned. Many later studies determined that this is true; groups both bring up and, after the first time, repeat the shared information favoring the weaker candidate much more often than the unshared information supporting the stronger candidate. In addition, if asked to recall information after the group decision is made, they remember the shared items more often than the unshared. Thus the arguments favoring the stronger candidate remain hidden from their consideration. Because of this, Stasser and Titus called this phenomenon the hidden profile effect.
The Information Sampling Model
The information sampling model was proposed by Stasser and Titus (1987) as an attempt to explain the “hidden profile” effect and other research findings that are inconsistent with persuasive arguments theory. The information sampling model starts with the idea that there is a given probability that an individual will think of a particular item of relevant information during group discussion. Beyond that idea, the model makes four presumptions. First, it is assumed that every item of relevant information has an equal chance of being thought of by any group member who is familiar with it. In other words, all pieces of information are equally memorable. Second, it is assumed that all group members are equally good at thinking of pieces of information that they are knowledgeable about. Third, it is assumed that whether one group member thinks of a given piece of information has no effect on whether another member of the group thinks of that piece of information. Fourth, it is assumed that if a group member thinks of an item of information, that member will bring it up in group discussion.
Given these presumptions, let us presume that the odds that each of the members of the student body president group think of an item of information relevant to their group’s decision is .4. Let us also presume that Betsy and Becky are aware of a particular piece of information, whereas Jeff is not. If the odds that Betsy, for example, thinks of a given item is .4, then the odds that Betsy will not think of that item is 1 minus .4, or .6. Now, as the odds that Becky thinks of that item of information is also .4, then the odds that neither Betsy nor Becky will think of it is .6 times .6, or .36. Thus the odds that either Betsy or Becky will think of it and contribute it to group discussion are 1 minus .36, or .64.
In general, if the probability that one member of a group will not think of a given piece of information is one minus the odds that this member will think of it, then the odds that no group member will think of and bring up the piece is
[1 - probability of member recall] n
and the odds that at least one member will think of and bring it up is
1 - [1 - probability of member recall] n
with n equaling the number of group members who are knowledgeable of that item. Thus if Jeff were also aware of that item of information, then the odds that it would be recalled by at least one of the investment group members is .78. As you may recall it from Chapter 2, this model and the reasoning behind it is analogous to Lorge and Solomon’s (1955) Model A.
Let us clarify the difference between the information sampling model and persuasive arguments theory. As described earlier, persuasive arguments theory implies that a given piece of information is equally likely to come up no matter whether all or only one member of the group knows it. This is one of the ways in which persuasive arguments theory predicts an unbiased discussion. In contrast, the information sampling model implies that the more group members who know a piece of information, the more likely it will be recalled. This leads to bias in group discussion.
An example of this type of bias is the hidden profile effect. Recall that in Stasser and Titus’s (1985) study uncovering this effect, items of information known to all group members, or “shared” among them, had more effect on their group’s decision than “unshared” items of information, those that only one member was familiar with. This is implied by the information sampling model. Recall that in our first example, all three members of the investment group were familiar with four arguments but only one member with each of the other six. The information sampling model leads to the prediction that the four shared arguments are more likely to come up during discussion and affect the group’s decision than the six unshared arguments.
There are several implications of this idea. First, the more group members that are aware of an item of information, the more likely that item will in essence dominate group discussion by “crowding” out unshared items of information. Stasser and Titus (1987) found research support for this implication. The researchers gave participants either 12 or 24 items of information relevant to a topic, and then formed groups in which either 33 or 66 percent of these items had been given to every member. Next, the groups discussed the topic. Afterward, each participant was asked to recall the information that came up during discussion. The researchers were interested in discovering the proportion of the information the participants recalled that was not originally given to them and thus learned during group discussion. When only 33 percent of the information was shared, one-third of the recalled information was learned during discussion if there were 12 items originally given and one-fifth was learned during discussion if there were 24 items originally given. In contrast, when 66 percent of the information was shared, only one-tenth of the recalled information was learned during discussion, no matter how many items were originally given. In summary, the more shared information, the less new information the participants learned during discussion. In some way, the shared information crowded the unshared out. It is likely that the unshared information was not brought up during discussion. Unfortunately, Stasser and Titus did not content analyze the discussion, so we cannot be sure if this is what occurred.
A second implication of the information sampling model is that the larger the group, the more likely that unshared information will not come out. This is because as the group gets larger, the likelihood that a shared item of information is thought of and brought up in discussion also gets larger; but no analogous increase occurs for unshared information. Thus the shared information is more likely to crowd out the unshared. However, research by Stasser, Taylor & Hanna (1989) was only partly in support of this second implication. The researchers performed content analyses of the discussion of 3- and 6-member groups. In the 3-member groups, 32 percent of the shared information and 18 percent of the unshared info came up during discussion. In the 6-member groups, 58 percent of the shared information was mentioned during discussion. This is a greater proportion than in the 3-person groups, which is consistent with the information sampling model. However, 19 percent of the unshared information was discussed. This is basically the same proportion as in the 3-member groups; the information sampling model would have predicted that the proportion in the 6-member groups would be distinctly lower.
It is important to note that several of the assumptions of the information sampling model are unrealistic. Some pieces of information are more memorable than others, some people are better at thinking of pieces of information than others, and people do not always contribute every item of information they think of. Nonetheless, it is clear that the information sampling model is a better representation of group information utilization than persuasive arguments theory. In particular, only the information sampling model predicts that shared information is more likely to come up in group discussion, and as a result has a greater impact on group decisions, than unshared information. The model presumes this occurs because shared information is more likely to be thought of by group members. At least three additional reasons have also been proposed for this bias. First, group members are likely to consider shared information more reliable and trustworthy just because more members are familiar with it (Larson, Foster-Fishman & Keys, 1994). Second, shared information has greater odds of being remembered over the entire discussion than unshared information. This is because group members are in a sense being exposed to shared information more often than unshared. Not only is it part of their original base of knowledge, they are more likely to hear it repeated during discussion (Larson et al., 1994). Third, a desire for conformity may lead group members to not discuss unshared information (Larson, Christiansen, Abbott & Franz, 1996).
“Opportunities” for Information
The next theoretical advance in the study of information utilization is found in Larson, Foster-Fishman & Keys (1994). Larson et al.’s thinking was actually an extension of the information sampling model. It is based on the idea that the odds that a given item of information will come up in group discussion is based on the number of “opportunities” that item has of being thought of by a group member. The number of opportunities for a given piece of information is equal to the size of the group. So, to return to our first example, if only Jeff is aware of an argument for Candidate B, then that argument has one opportunity to come up, but as all members are aware of an argument for Candidate A, that argument has three opportunities.
In a group with a given size, the number of opportunities for shared items of information to occur during discussion equals the number of shared items multiplied by the size of the group. Therefore, as all the members of the group are familiar with the 4 pieces of information supporting Candidate A, there are 12 opportunities for shared pieces to come up. In contrast, the number of opportunities for unshared items of information to occur during discussion equals that number multiplied by 1, the member who knows it. As there are 6 items of unshared information among the members of the investment group, then there are only 6 opportunities for them to come up. In a mathematical sense, this accounts for why:
1 - shared information has an advantage during discussion. The number of shared items multiplied by group size will normally be greater than the number of unshared items multiplied by group size.
2 - larger groups accentuate this effect. In a larger group, the number of shared items multiplied by group size is proportionally greater than in a smaller group.
3 - a larger proportion of shared information accentuates this effect. With more shared information, the number of shared items multiplied by group size is proportionally greater.
This third issue is particularly important, because it also implies that shared information will not have an advantage during discussion if the proportion of unshared information is large enough to overcome the group size effect. Let us imagine that the group members are all familiar with 2 pieces of information favoring Candidate A, while there are 10 pieces of information favoring Candidate B unshared among them. In this case, the number of opportunities for unshared items to come up (10) is greater than the number of opportunities for shared items (2 items multiplied by 3 members equals 6 opportunities). Thus the bias for shared information will not occur if there is a large enough proportion of unshared information.
A further implication concerns changes over time in the number of opportunities for shared and unshared information to come up during discussion. Each time that an item of information comes up for the first time during discussion, the total number of opportunities to bring forth as-yet-undiscussed items of information goes down by the number of members who know that item. If Jeff were to bring up one of the shared arguments in favor of the investment, given that all three group members are familiar with it, then the total number of opportunities for shared arguments to come up later in the discussion would decrease by three. In contrast, if Lisa were to bring up an unshared argument favoring Candidate B, then the total number of opportunities for unshared arguments would only go down by one. As you can see, the number of opportunities to bring up new pieces of shared information will go down faster than the number for unshared information. Thus the bias for shared information will decrease as discussion continues. The rate of this decrease can be calculated. For example, if the group has among them 10 shared and 10 unshared pieces of information, the odds that the first item of information is shared will be .75 but that the sixteenth item will be shared will be .30.
As with any scientific model, Larson et al.’s refinement of the information sampling model includes some unrealistic assumptions. For example, it presumes that a piece of information will only come up once in discussion, ignoring the fact that previously discussed information is often repeated. Nonetheless, Larson and associates have found evidence for these implications. For example, Larson, Christiansen, Abbott, and Franz (1996) gave three-member medical teams both shared and unshared information about some hypothetical medical cases and asked them to diagnose the cases. The probability that an item of information was shared decreased from .72 for the first brought up during discussion to .25 for the eighteenth.
Is it Normative or Informational Influence?
When Stasser and Titus (1985) originally designed the hidden profile study, they believed that their results would distinguish between the information influence explanation for social influence provided by persuasive arguments theory and the normative influence account given by social comparison theory. If group members maintained their original preference for the weaker candidate, then normative influence processes would appear to be operating. If, instead, group members switched their allegiance to the stronger candidate, then information influence would seem to be at work. They found that group members continued to support the weaker candidate, and this finding has been replicated numerous times by many different researchers. This makes it appear as if normative influence is the fuel behind what occurs when groups face hidden profiles.
But is this always true? Could informational influence also have a role in these situations? Next, we will study different points of view on this question.
The common knowledge effect. The common knowledge effect is the idea that the more people who are aware of a piece of information, the greater impact that piece of information has on group members’ opinions. Gigone and Hastie (1993) demonstrated the common knowledge effect by applying a twist to the usual hidden profile research paradigm. Members of three-person groups were provided with six different types of numerical information about 32 students, such as their high school grade point average and their self-reported enjoyment for a particular course, and asked to predict the student’s grade in that course. Of these six items, two were shared by all three group members, two were shared by two members, and two were unshared. Gigone and Hastie showed that, for four of the types of information, the extent to which a particular information item was shared was related to how strongly the group members weighed that item in their predictions. In addition, some types of information, such as high school g.p.a, were always more influential than others.
This finding is inconsistent with the assumption behind the information sampling model that all items of information are equally important. This does not, however, mean that Gigone and Hastie believe that normative influence is not important in hidden profile tasks; quite the opposite. They were also able to show that the number of items of information discussed by group members was generally not related to the manner in which the information that the group members were provided impacted on their predicted grade. In contrast, prediscussion preferences were usually related to the impact of provided information on predicted grades. This is evidence in favor of normative influence and against informational influence as responsible for their findings.
The dual process model. It did not take long for researchers to respond to Gigone and Hastie’s work. Winquist and Larson (1998) did some reasoning that led to an alternative hypothesis. A normative influence account for social influence during the discussion of hidden profile tasks implies that group members’ preferences after their discussion and their group decision will be totally consistent with their opinions before the discussion. Keep in mind that shared information will be consistent with those prediscussion opinions and unshared information will be inconsistent with them. If so, then when shared information comes up during discussion, because unshared information is consistent with the group members’ opinions, it should have no effect on them. However, when unshared information is mentioned, it will be inconsistent with those opinions. It follows that if unshared information has any impact at all on group member preferences and their eventual decisions, it must be because they provided informational influence contrary to their original opinions.
The researchers set up a situation in which each person in three-member groups received six positive and two negative pieces of information about two job candidates, implying for each member individually that neither was better than the other. However, this information was distributed among the members such that, if successfully pooled across members, one candidate would clearly stand out over the other. Results showed that decision accuracy was directly related with the amount of unshared information discussed by the group but not with the amount of shared information. This is unquestionable support for the presence of informational influence.
In response to these findings, Winquist and Larson (1998) proposed a dual-process model for social influence within groups. In summary, items of information impact in postdiscussion individual preferences and group decisions through two avenues. Shared information determines group members’ prediscussion preferences, which in turn, affect their postdiscussion opinions and group decisions. Unshared information, in contrast, affects discussion content; this turn impacts on member postdiscussion preferences and group decisions. These two influence routes actually occur sequentially, as shared information has impact before and unshared information has impact during discussion. They will, however, have different strengths in different situations. Hidden agenda situations are set up so that group members will agree with one another; when that occurs, shared information will overwhelm unshared information and normative influence will be stronger than informational influence. In circumstances in which members are less likely to disagree with one another, normative influence will be lessened, so the effect of unshared information and informational influence will be stronger.
The Winquist and Larson (1998) dual-process model seems to ignore the impact that shared information may have in strengthening members’ opinions when the group is in agreement. For this reason, it differs from ours, which allows for normative influence occurring not only before but at the same time as informational influence. Having said this, other research exists that provides more evidence that when groups disagree with one another about the best choice in a hidden profile situation, informational influence is important. For example, the Winquist and Larson study is just one of several that demonstrated that the extent to which a group discussed the better option was related to how often the group chose that option. Keep in mind that the better option is usually dispreferred by group members before discussion, so this discussion almost certainly led to informational influence. One of these demonstrations was by Hollingshead (1996), who performed a detailed analysis of the relationship between prediscussion preferences and group decisions in a hidden profile task. She uncovered successful minority influence in 10 of 39 three-member groups in which only one of the members preferred the best option. As we mentioned earlier in the chapter, minority influence has largely been viewed as a function of informational influence; after all, the very concept of normative influence seems inconsistent with a minority’s preferences.
After looking at all the evidence, it is impossible not to conclude that both informational and normative influence work together in groups. Although we have presented them as in some sense opposed to one another, they can also be seen as working in tandem. If a group member makes a particular argument that other group members have not heard before, informational influence might occur. At the same time, group members may see which side of the issue that member is on, and if it is consistent with the majority, some normative influence might happen simultaneously. On the other hand, as we have seen, arguments themselves are seen as more convincing the more people that voice them.
Another source of evidence pointing to a dual process is the perceptions of research participants themselves. In a group polarization study, Butler and Crano (1992) asked participants who had shifted their preferences the reasons that they believed to lead to their shift. Of 130 supplied reasons, the authors categorized 99 as consistent with informational influence (e.g., “The group gave me new insights that I had not considered,” “They brought up new ideas that I hadn’t thought of”) and 25 with normative influence (“I realized that my decisions were far from the norm,” “It helped to make decisions that would suit everyone”); the remaining six were irrelevant to both mechanisms. Recently, in an explicit evaluation of the issue, Price, Nir, and Cappella (2006) determined that both arguments (information influence) and statements of support for given options (normative influence) had discernible impacts on opinions after political discussions by groups discussing the issues on-line.
FROM THEORY INTO PRACTICE
Scientists have seen the tendency for groups both to polarize and to be fooled by hidden profiles as indications of poor decision making, and have looked for ways to overcome them. Both can be viewed as the product of a “tyranny of the majority.” This is because they are the result of either individual opinions or discussion arguments on one side of the issue crowding out those supporting the other side. Fortunately, those studying hidden profiles in particular have come up with a number of methods that have led groups to overcome the domination of shared arguments and attend as much to the unshared. I suspect these techniques would also be helpful in lessening group polarization and in paying more attention to minority points of view in all situations.
First, members with higher status in groups bring up more unshared information than those with lower status. For example, Larson, Christensen, Franz, and Abbott (1998) provided three-member medical teams with information about a medical case and asked them to make a decision. The lead physicians in the medical teams were more likely to add their unshared information to the mix than the other team members. This is probably because they have more confidence than the others and feel that they can deviate from group norms without losing their group’s respect. This implies that one way to make group members comfortable with bringing up all of the information that they have is to give them status. This can be done in various ways. For example, assigning each group member to the job of group expert for a specific part of the information the group needs to make its decision should help. Fraidin (2004) did just that, and it motivated the group members to bring up more of their information than those without such instructions. Along the same lines, Stasser, Stewart and Wittenbaum (1995) discovered that even just informing everyone in the group about what unshared information that each candidate each led to more unshared information coming up.
The experience of working on a problem may result in group members being more willing to divulge the information they have the next time they work on the same type of problem. Wittenbaum (1998) provided some participants with individual problem-solving experience. She then assigned them to a group that also included members without that experience and gave that group a similar problem. As usual, the inexperienced members mostly brought up shared information. The experienced members, in contrast, brought up as much unshared information as shared. As all group members knew which among them had the previous experience, Wittenbaum believed that they were accorded greater status in their groups, leading to greater comfort at providing unshared information. Another possible explanation for the findings is that the very experience of doing the task twice led those members have increased self-confidence.
One final method for improving group problem-solving discussion is to change the usual instructions that the group members are given. Groups are usually asked to decide on the best candidate. Stasser and Stewart (1992) gave that description to half of their groups; as is normal, only about one-third of them made the correct choice. In contrast, Stasser and Stewart instructed the other half of the groups to decide which of the candidates is most likely to be the best candidate. This slight change in wording led to almost 2/3 of the groups with those instructions overcoming their hidden profile and came to the right decision. Other instructions that have proved beneficial are instructions to rank order all of the candidates (Hollingshead, 1996) and to critically evaluate the provided information (Postmes, Spears, & Cihangir, 2001).
Chapter 6 described how scientists have generally studied conformity by using experiments that artificially limit social interaction. This chapter looked at studies that allowed interaction to proceed more naturally. These studies have led to further insights into social influence processes. We have learned that group member opinions tend to converge toward one another, with minority viewpoints in particular moving closer to those of the majority.
An insight that we have gained from more natural experiments involves what happens when group members work on a choice dilemma which pits a more attractive but risky option against a sure but less attractive alternative. Scientists have found a group polarization effect. This takes place when group members are initially in general but not exact agreement on one side of an issue involving potential risk. In this situation, the group’s final decision tends to be on the same side as their initial judgments, but it is more extreme.
Some scientists have felt that group polarization occurs because of normative influence. This consists of group members coming to change their minds to be consistent with the group majority just because they believe that the majority must have the correct opinions. Social comparison theory is consistent with a normative influence explanation for group polarization. We discussed this theory back in Chapter 6 as an account for conformity. Here, group members are seen as deciding that the group majority’s preference must be correct and then changing their own views to be more consistent with it.
Other scientists have proposed that group polarization is a result of informational influence. In this case, group members are seen as changing their minds because of the information that has come up during group discussion. Persuasive arguments theory is consistent with the idea of informational influence. This theory claims that when group members hear arguments that are consistent with their own prediscussion views but that they had not thought of earlier, they further polarize their opinions.
Research suggests that social comparison theory and persuasive arguments theory may be partial and complementary explanations for group polarization. Accordingly, I have proposed a model that attempts to combine these two theories.
The hidden profile paradigm began as an attempt to distinguish normative and informational influence explanations for social influence during group discussion. A hidden profile exists if the information each group member is individually aware of supports one decision option, but all the information the group has as a whole implies a different option. If a group is susceptible to the hidden profile and decides on the option that each member believed correct before group discussion, then any influence that occurred would be consistent with these prediscussion beliefs. This would count as normative influence. In contrast, if group members overcome the hidden profile, change their minds about the best option, and make the correct decision, then influence must have come from the information about the decision that came up during discussion.
Research has shown that most groups are fooled by the hidden profile. It has also revealed that group discussion mostly contains mention of the items of information that all group members share. These findings are consistent with a normative influence account. The information sampling model provides a framework for understanding why shared information dominates, along with other phenomena that occur when groups are faced with hidden profile tasks. One of these is the tendency for shared information to come out at the beginning of group discussions, but for unshared to gain more prominence as discussion continues.
The common knowledge effect is the tendency for items of information to have more impact on group decisions as the number of group members that are aware of them increases. Further, some research shows that prediscussion preferences are related to how much the information that group members are provided affects group decisions, but the discussion of this information is not. These results are consistent with a normative influence explanation. However, other studies reveal that when group members overcome hidden profiles, it is due to unshared items of information arising in discussion. This finding implies informational influence. As with group polarization, it is clear that when groups grapple with hidden profile problems, both types of influence are involved.