Group polarization happens when a discussion among like-minded people leaves them holding a stronger version of the view they started with. A group leaning toward risk becomes riskier, and a group leaning toward caution becomes more cautious.
Key takeaways
- Definition: Groups shift toward the extremes of decision-making when compared with decisions made by individuals.
- Cautious Groups: Members who are already cautious become even more cautious when they discuss a decision in a like-minded group.
- Risky Shift: Members who are less cautious beforehand tend to make riskier decisions in a like-minded group. This type of group polarization is known as a risky shift.
- Direction: On average, groups polarize toward the attitudes held before the group decision is reached.
- Explanations: Persuasive arguments, social comparison, and self-categorization each explain polarization, and they often work together.
Definition & Background
Group polarization describes how members of a group adopt more extreme positions than the initial attitudes and actions of individual group members (Baumeister, 2007).
This occurs because the collective involvement of groups amplifies individual attitudes and opinions, shifting them further in the direction of group norms (Colman, 2015).
The effect works in either direction:
- Risk-leaning groups: become riskier after discussion.
- Caution-leaning groups: become more cautious after discussion.
Polarization does not mean two factions splitting apart. It means one group moving further along the dimension it already favored.
This contradicts a common-sense prediction.
Conformity research had suggested that groups pull members toward a moderate, conservative average. Group polarization shows that group influence can radicalize as readily as it moderates. Meta-analysis confirms the effect (Isenberg, 1986).
The underpinnings of group polarization lie in Stoner’s (1961) master’s thesis on the “risky shift,” which evolved into theories about choice shift. Replications supported the basic finding. What later work overturned was the idea that groups always become riskier.
While group polarization necessitates choice shift, the choice shift does not necessitate group polarization.
Group polarization can manifest in situations ranging from gambling behavior to social media and can have positive, deleterious, or neutral effects.
Risky Shift and Early Theories
The first provocative study in social psychology to describe group polarization was Stoner’s (1961) master’s thesis. It described the discovery of the “risky shift.” Groups generally behaved in a riskier way than individuals, taking bets with higher potential rewards but lower probabilities (Cartwright, 1971).
Although later work showed that groups are not reliably risk-seeking, the phenomenon of choice shift continued to motivate research.
Psychologists had expected the opposite.
Reading conformity research, they predicted that groups would counsel more caution than individuals. In Stoner’s task, participants advised hypothetical people on how much risk to accept. After group discussion, their advice became riskier, not safer.
Research on the risky shift relied on the Choice-Dilemmas Questionnaire. Wallach, Kogan, and Bem (1962) formalized it. Each item pits a safe, modest outcome against a risky, more rewarding one.
Respondents give the lowest acceptable chance of success.
In one dilemma, a low-ranked chess player could try a risky but deceptive maneuver against a strongly favored opponent. Alone, people advised trying it only with at least a 30 percent chance of success. After discussion, they said go for it at 10 percent.
The shift lasted. Members’ own private judgments were also riskier afterward, and the change was still detectable weeks later (Wallach, Kogan, and Bem, 1962).
Traditional perspectives on social influence tended to emphasize accommodation and conformity but not necessarily the idea that groups had an “averaging effect” on individual perspectives (Friedkin, 1999).
The risky shift did not hold across every dilemma. Some items produced the opposite movement, and groups became more cautious than their members had been alone.
Take Roger, a father of two with a secure but low-paying job and no savings. A tip says a stock will triple if the company’s new product succeeds and collapse if it fails. Alone, most people recommend caution. After discussion, they become even more conservative.
The same discussion process that pushed the chess player toward risk pushed Roger’s advisers toward caution. So the shift was not toward risk as such. It was toward extremity in whichever direction the group already leaned.
Serge Moscovici, a Romanian-born French sociologist, and Marisa Zavalloni, an Italian-born Canadian social psychologist, introduced the concept of group polarization.
Moscovici and Zavalloni (1969) tested whether discussion strengthens pre-existing attitudes beyond the special case of risk.
- Aim: To test whether group discussion strengthens attitudes in general, so that polarization is a broad principle and not just a risk effect.
- Method: Around 140 male secondary school students discussed their attitudes toward General de Gaulle and Americans. Attitudes were measured before and after discussion.
- Results: Initially positive attitudes toward de Gaulle became more positive, and initially negative attitudes toward Americans became more negative.
- Conclusion: Discussion intensifies the leaning a group already has. This generalized the risky shift from decisions about risk to attitudes of any kind.
The design showed polarization in both directions within one study, and its topics carried real social meaning. The sample was narrow, though: young French men of one era, which limits how far the exact effects generalize.
Lamm and Myers (1978) describe the theoretical underpinnings of group polarization. In early research on group polarization, several theories dominated:
- social decision rules (ways that individual decisions can form into a group decision),
- informational influence (where people learn as a result of hearing and giving cogent arguments considering the topic at hand),
- social comparison effects,
- responsibility dynamics (where having a large number of people in a group may make members feel less responsible for holding accountability for a given decision).
Explanations for Group Polarization
According to social network influence theory, these are the four main explanations of choice shift:
Persuasive Arguments Theory
Persuasive arguments theory posits that group polarization occurs because of the content of the arguments that happen during discussions.
This theory assumes a pool of arguments that can be applied to any issue (Friedkin, 1999). People draw on the part that supports their initial attitudes.
The group discussion’s purpose, according to persuasive arguments theory, is to provide individuals with complete information about the merits of alternative positions.
Suppose the arguments members draw on favor a position more extreme than the mean of the initial positions. Polarization is then likely to occur (Friedkin, 1999).
The key idea is that choice shift and group polarization depend on arguments. These arguments may be stated openly. They may also arise implicitly as members rethink the issue.
However, this approach emphasizes that the simple visibility of group member attitudes is not sufficient for choice shifts.
Because members already lean the same way, the pool of arguments in the room is skewed toward that shared position. Each person hears valid supporting arguments they had not considered, and each one nudges them further. Nobody caves to social pressure. Everyone is more genuinely convinced.
Burnstein and Vinokur (1977) tested this account directly.
- Aim: To separate the effect of exchanging arguments from the effect of simply knowing other people’s positions.
- Method: Experiments compared conditions where participants exchanged arguments with conditions where they only learned others’ positions. The novelty and one-sidedness of the available arguments were also varied.
- Results: Polarization grew when participants met novel, valid arguments favoring the dominant position. It was markedly weaker when they only learned others’ positions.
- Conclusion: Argument content drives polarization, which supports persuasive arguments theory. Social comparison probably contributes too, so both processes likely operate.
The design isolates a causal mechanism instead of a mere correlation. Showing that arguments matter does not show that social comparison is irrelevant, though.
Social Comparison Theory
Social comparison theory argues that people will enter a discussion espousing views less extreme than their true views because they fear being labeled deviant.
Discussion then reveals that other members hold similar but more extreme attitudes. People stop suppressing their true values and start exposing them. Some adopt the more extreme positions of others (Baron and Roper, 1976; Friedkin, 1999).
All in all, this means that moderate positions are eroded by discussions and extreme positions are encouraged (Friedkin, 1999).
This theory assumes that arguments on one extreme of the scale are more valued by the group than moderate or opposing views, as these arguments have higher visibility.
This is normative social influence. It is the drive to be accepted and to compare favorably with others. It predicts polarization even from mere exposure to others’ positions, without any exchange of arguments.
Imagine someone who privately holds a fairly risky attitude but voices a moderate one, so as not to look reckless. They then discover that the group actually admires risk.
Now they feel free to state their true view, and even to exceed it. Everyone does the same at once, so the group average moves toward the extreme.
Members may also move toward the extreme to set their group apart from outgroups they assume hold blander views.
Self-Categorization Theory
Self-categorization theory explains that group polarization happens on the basis of a person’s conformity to an extreme norm or position of the group (Abrams, Wetherell, Cochrane, and Hogg, 1990).
This norm is not the average view but rather the prototypical position of the group. That prototype can take one of three forms:
- The group’s mean initial attitude
- One of the group’s most extreme attitudes
- A view in between these two
Which form it takes depends on how members view their relationships with the out-groups.
Groups choose among visible individual positions and gravitate toward a normative position.
Proponents such as Turner and Reynolds (2011) believe that members modify their attitudes to fit a prototypical position (Friedkin, 1999). This narrows the gap with group norms.
If everyone accepts a norm prior to discussion, social influence predicts that there is no choice shift (Friedkin, 1999).
Without such a preexisting norm, self-categorization theory hypothesizes that a view’s influence depends on interpersonal influence among group members. More powerful members have greater influence (Friedkin, 1999).
Self-categorization theory grew out of social identity theory. It describes referent informational influence: members infer the right attitude from the norms of a social category they belong to.
Polarization then unfolds in three steps:
- Categorize: a person categorizes the self as a member of a particular ingroup.
- Identify: they identify the prototypical position that best defines the ingroup and separates it from outgroups.
- Conform: they adjust their own attitude to match that prototype.
The prototype is often displaced away from the outgroup, toward the extreme that makes the ingroup distinctive. Conforming to it therefore moves members outward (Abrams et al., 1990).
Environmental activists debating tactics show the pattern. Their prototype is not the average of their opening views but a more committed stance that sets them apart from the moderate mainstream.
Social Decision Scheme Theory
Finally, social decision scheme theory proposes that group decisions depend on two things. One is the initial distribution of attitudes in the group. The other is the decision scheme (or decision rule) that members use to reach a decision (Davis, 1973).
Decision rules can encompass several processes (Laughlin, 1980; Friedkin, 1999):
- The majority rule
- The mean of initial attitudes
- The median of initial attitudes
- The most extreme initial position
In this view, choice shifts happen when these decision-making rules produce a group decision differing from the mean initial attitude of group members.
Lamm and Myers (1978) emphasized that these mechanisms are not necessarily mutually exclusive. They may amplify each other or operate during different phases of the decision-making process.
They work together. Persuasive arguments make the favored position more defensible, while social comparison makes it more rewarding to endorse. Self-categorization supplies the identity frame that decides which direction counts as the group’s leaning.
Different issues favor different routes. Factual issues favor the informational route, while value-laden issues favor comparison and identity. Meta-analytic evidence supports more than one process, so polarization is best seen as multiply determined (Isenberg, 1986).
Lamm and Myers supported their motivating factors with a number of early studies. Wolosin et al. (1975) support the responsibility dynamics hypothesis.
Wolosin and colleagues coupled risks of shock with monetary rewards for the decision-makers. The more shocks the group risked, the less responsibility each member felt for its decisions.
Group Polarization in Social Influence
Friedkin (1999) describes the interdependency between choice shift and group polarization. Choice shifts occur when, after a group’s interaction, the average attitude of group members differs from the member’s average initial attitude.
Meanwhile, group polarization happens when the choice shift is in the same direction as where the average initial attitude was inclined.
For example, suppose a group’s average initial attitude is slightly positive. After discussion, the average becomes more positive.
Group polarization always involves a choice shift. The reverse is not true.
Cartwright (1971) argued that “risky-shift” research needed an analysis of the interpersonal influences that led to choice shifts. The logic proceeds in four steps:
- Accommodation: Cartwright suggested that interpersonal accommodation alone can produce a shift, so no new process or separate group effect is needed.
- Power Structure: French’s (1956) and Harary’s (1959) formal theory of social power relates a population’s power structure to its structure of influential communications.
- Agreement Patterns: That communication structure is in turn related to the pattern and prevalence of interpersonal agreements (Friedkin, 1986).
- Unequal Influence: Choice shifts such as group polarization therefore arise from differences in susceptibility and inequalities of interpersonal influence (Friedkin, 1999).
If individuals have different levels of interpersonal influence, then a choice shift will occur.
Examples
Book Club
Consider a book club, where members read a book ahead of time and then discuss it. An individual may think the book was good but far from the best they have ever read. Most other members may feel the same.
However, when the club gathers, the group amplifies the good elements of the book. Members leave with a more positive opinion than they came in with.
Group polarization has occurred here. The group’s attitude has strengthened from thinking the book was merely good to believing it was excellent.
Disaster Relief Organization
Group polarization does not necessarily entail negative effects. Take a disaster relief organization working with a government to decide how much aid to give a region affected by a hurricane.
The members of the organization may wish to allocate more aid than members of the government.
When these individuals discuss the question, the group’s normative value of giving aid becomes amplified. Both groups may then wish to allocate even more aid.
Gambling Behavior
Researchers have used various gambling and betting tasks to investigate the effects of group interaction on risk-taking. In one group of such studies, participants choose among different probabilities of winning, with lower probabilities having higher payoffs.
The betting items differ in the size of the stake.
Their expected value, the profit one would probably make by repeating the gamble many times, is the same (Lamm and Myers, 1978).
Blascovich and colleagues ran a series of blackjack studies. Over 20 trials, individuals playing alone did not change their risk-taking level. Groups tended to take progressively higher risks (Blascovich, Ginsburg, and Howe, 1975).
Real-World Applications
Any deliberating group of like-minded people can polarize, from the jury room to the internet. Four settings show the pattern:
- Juries: Deliberating juries tend to end up more extreme, in their leaning direction, than the average juror began, which affects verdicts and damages awards.
- Boards: Committees and policy teams decide more boldly or more cautiously than members privately favored, which can mean decisiveness or an aggressive strategy talked into recklessness.
- Groupthink: Polarization supplies the extremizing engine, and cohesion plus insulation let it run unchecked (Myers and Lamm, 1975).
- Radicalization: People who share a grievance and discuss it only among themselves harden toward the extreme.
Myers and Bishop (1970) showed the mechanism with racial attitudes. High school students sorted into high- and low-prejudice groups discussed racial issues with like-minded peers. The high-prejudice groups became more prejudiced, and the low-prejudice groups became less so.
Social Media
Cass Sunstein’s (2001) influential work on “online informational cocoons” brought renewed interest to group polarization. Iandoli, Primario, and Zollo (2021) conducted a systematic review of “dysfunctional” group polarization in the context of interaction enabled by social media.
In general, the researchers found two routes to increased polarization.
Social media can amplify and escalate social processes that also occur offline. It can also polarize in ways enabled by its design, which makes the platforms prone to manipulation.
Platforms also curate what users see. Algorithms preferentially surface engaging, often emotive or extreme content. They also feed each user more of what they already endorse, narrowing the range of views they meet.
Like-minded enclaves then drift toward extremes. Sunstein (2002) named this the law of group polarization.
Iandoli, Primario, and Zollo argue that social media encourages group polarization by enabling two things. The first is homophily, the tendency to engage with similar people, which creates groups where certain beliefs become dominant.
The second is discursive argumentation (Mercier and Sperber, 2011), the use of reasoning by individuals and groups to advance their agenda.
This manifests through four routes (Iandoli, Primario, and Zollo, 2021):
- Social Antecedents: diverse viewpoints and external triggers such as partisan media.
- Cognitive Arguments: social comparison and social identity theory, as well as persuasive arguments theory.
- Social Effects: opinion radicalization and misinformation.
- Design Affordances: content sharing and the feeding and prioritization of information.
Critical Evaluation of Group Polarization Research
Group polarization is one of the more robust findings in social psychology, yet the classic evidence has clear limits.
Strengths of the Evidence
Four features make the research persuasive:
- Robust Across Methods: The effect appears in risk decisions, political attitudes, racial prejudice, and large online datasets. A meta-analysis confirms it (Isenberg, 1986).
- Self-Correcting: The risky shift was refined, not abandoned, once cautious shifts appeared. The result was a general, testable law.
- Complementary Explanations: Informational, normative, and identity-based accounts each have empirical support and often act together (Burnstein and Vinokur, 1977; Abrams et al., 1990).
- Practical Reach: The same process helps explain polarized juries, boards, and online communities.
Because independent methods converge on the same pattern, researchers treat polarization as a genuine principle of group influence rather than a laboratory quirk. The mechanism has stayed stable even as the scale and stakes have grown. That convergence is why the finding counts among the more secure in social psychology.
Limitations and Criticisms
Four criticisms qualify those strengths:
- Overgeneralized Risky Shift: The first decade treated discussion as reliably risk-seeking, an artifact of the dilemmas that happened to produce risky shifts.
- Artificial Methods: Classic studies used hypothetical dilemmas about strangers and ad hoc laboratory groups. Real stakes, cohesion, and history may change the result.
- Entangled Mechanisms: Persuasive arguments, social comparison, and self-categorization often operate together, so support for one rarely rules out the others.
- Variable Effect Size: The effect’s magnitude varies with the issue, the group, and how positions are elicited (Isenberg, 1986).
Polarization is therefore a reliable tendency, not an inevitability. It also needs a shared initial leaning, so genuinely mixed groups need not polarize. The classic evidence is stronger on whether the effect exists than on how much each cause contributes.
Contemporary Research
Modern work has moved from small student groups in the laboratory to computational analysis of millions of real online interactions. The main question is no longer whether discussion polarizes attitudes. It is how platform design shapes the process.
The anchor study is a cross-platform analysis by Cinelli and colleagues (2021).
- Aim: To measure whether social media produce echo chambers, meaning clusters of like-minded users insulated from opposing views, and how this varies with platform design.
- Method: The researchers analyzed more than 100 million pieces of content on divisive topics such as abortion, gun control, and vaccination. They compared Facebook, Twitter, Reddit, and Gab, mapping how information spread within and between ideological groups.
- Results: Echo-chamber structure was strong on Facebook and Twitter, which emphasize following and algorithmic feed-ranking. It was markedly weaker on Reddit, which is organized around topics.
- Conclusion: Platform design can create the homogeneous exposure that polarization theory identifies as its engine, but the effect is not an inevitable feature of the internet.
The huge, real-behavior dataset gives this study power and external validity that no laboratory experiment can match. It measures exposure and structure, though, not the attitude change that classic experiments captured. Self-selection into platforms is also hard to separate from platform effects.
When Exposure to the Other Side Backfires
Bail and colleagues (2018) paid a national sample of American Twitter users to follow a bot that retweeted messages from the opposing political side for a month. Republicans became substantially more conservative, and Democrats became slightly more liberal.
Exposure to opposing views is therefore not a reliable cure. It can deepen division when identity is engaged.
Misinformation in Echo Chambers
Del Vicario and colleagues (2016) found that unverified and conspiratorial content on Facebook spread within tightly knit, like-minded communities. The longer members participated, the more polarized their engagement grew.
Weighing the Evidence
These lines of evidence update the classic mechanisms rather than overturn them. The processes are the same. Their scale, speed, and structural drivers are new.
References
Abrams, D., Wetherell, M., Cochrane, S., Hogg, M. A., & Turner, J. C. (1990). Knowing what to think by knowing who you are: Self-categorization and the nature of norm formation, conformity and group polarization. British Journal of Social Psychology, 29(2), 97–119. https://doi.org/10.1111/j.2044-8309.1990.tb00892.x
Bail, C. A., Argyle, L. P., Brown, T. W., Bumpus, J. P., Chen, H., Hunzaker, M. B. F., Lee, J., Mann, M., Merhout, F., & Volfovsky, A. (2018). Exposure to opposing views on social media can increase political polarization. Proceedings of the National Academy of Sciences, 115(37), 9216–9221. https://doi.org/10.1073/pnas.1804840115
Baron, R. S., & Roper, G. (1976). Reaffirmation of social comparison views of choice shifts: Averaging and extremity effects in an autokinetic situation. Journal of Personality and Social Psychology, 33(5), 521.
Baumeister, R. F., & Vohs, K. D. (Eds.). (2007). Encyclopedia of social psychology. SAGE Publications. https://doi.org/10.4135/9781412956253
Blascovich, J., Ginsburg, G. P., & Howe, R. C. (1975). Blackjack and the risky shift, II: Monetary stakes. Journal of Experimental Social Psychology, 11(3), 224–232. https://doi.org/10.1016/S0022-1031(75)80024-2
Burnstein, E., & Vinokur, A. (1977). Persuasive argumentation and social comparison as determinants of attitude polarization. Journal of Experimental Social Psychology, 13(4), 315–332. https://doi.org/10.1016/0022-1031(77)90002-6
Cartwright, D. (1971). Risk taking by individuals and groups: An assessment of research employing choice dilemmas. Journal of Personality and Social Psychology, 20(3), 361.
Cinelli, M., De Francisci Morales, G., Galeazzi, A., Quattrociocchi, W., & Starnini, M. (2021). The echo chamber effect on social media. Proceedings of the National Academy of Sciences, 118(9), e2023301118. https://doi.org/10.1073/pnas.2023301118
Colman, A. M. (2015). A dictionary of psychology: Oxford quick reference.
Davis, J. H. (1973). Group decision and social interaction: A theory of social decision schemes.
Del Vicario, M., Bessi, A., Zollo, F., Petroni, F., Scala, A., Caldarelli, G., Stanley, H. E., & Quattrociocchi, W. (2016). The spreading of misinformation online. Proceedings of the National Academy of Sciences, 113(3), 554–559. https://doi.org/10.1073/pnas.1517441113
French Jr, J. R. (1956). A formal theory of social power. Psychological review, 63(3), 181.
Friedkin, N. E. (1986). A formal theory of social power. The Journal of Mathematical Sociology, 12(2), 103–126. https://doi.org/10.1080/0022250X.1986.9990008
Friedkin, N. E. (1999). Choice shift and group polarization. American Sociological Review, 64(6), 856–875. https://doi.org/10.1177/000312249906400606
Harary, F. (1959). Status and contrastatus. Sociometry, 22(1), 23-43.
Iandoli, L., Primario, S., & Zollo, G. (2021). The impact of group polarization on the quality of online debate in social media: A systematic literature review. Technological Forecasting and Social Change, 170, 120924. https://doi.org/10.1016/j.techfore.2021.120924
Isenberg, D. J. (1986). Group polarization: A critical review and meta-analysis. Journal of Personality and Social Psychology, 50(6), 1141–1151. https://doi.org/10.1037/0022-3514.50.6.1141
Lamm, H., & Myers, D. G. (1978). Group-induced polarization of attitudes and behavior. Advances in Experimental Social Psychology, 11, 145–195. https://doi.org/10.1016/S0065-2601(08)60007-6
Mercier, H., & Sperber, D. (2011). Why do humans reason? Arguments for an argumentative theory. Behavioral and Brain Sciences, 34(2), 57–74. https://doi.org/10.1017/S0140525X10000968
Moscovici, S., & Zavalloni, M. (1969). The group as a polarizer of attitudes. Journal of Personality and Social Psychology, 12(2), 125–135. https://doi.org/10.1037/h0027568
Myers, D. G., & Bishop, G. D. (1970). Discussion effects on racial attitudes. Science, 169(3947), 778–779. https://doi.org/10.1126/science.169.3947.778
Myers, D. G., & Lamm, H. (1975). The polarizing effect of group discussion. American Scientist, 63(3), 297–303.
Stoner, J. A. F. (1961). A comparison of individual and group decisions involving risk [Unpublished master’s thesis]. Massachusetts Institute of Technology.
Sunstein, C. R. (2002). The law of group polarization. Journal of Political Philosophy, 10(2), 175–195. https://doi.org/10.1111/1467-9760.00148
Wallach, M. A., Kogan, N., & Bem, D. J. (1962). Group influence on individual risk taking. The Journal of Abnormal and Social Psychology, 65(2), 75–86. https://doi.org/10.1037/h0044376
