RESEARCH DOSSIER
Social Norms and Normative Expectations
Abstract. Social norms organize human behavior through expectations about what relevant others do, approve, and may sanction. This fundamental research monograph reconstructs the conceptual development of social-norm psychology, distinguishes empirical expectations, normative expectations, descriptive and injunctive cues, and examines how reference groups, social visibility, sanctions, and pluralistic ignorance affect behavior. It critically synthesizes experimental research on norm salience, conservation, school networks, belief correction, cultural tightness–looseness, and dynamic norms. The analysis compares behavioral experiments, surveys, field randomized trials, network methods, and cross-cultural designs, separating causal message effects from unmeasured mediating processes. Particular attention is given to norm change, distributional harms, methodological validity, digital interfaces, and emerging artificial communication. The synthesis identifies persistent challenges in reference-group elicitation, second-order belief measurement, longitudinal durability, cross-cultural invariance, and algorithmic selection of apparent consensus. This work is a critical narrative synthesis of existing research, not an original experiment, systematic review, or meta-analysis.
Keywords: social norms; normative expectations; empirical expectations; descriptive norms; injunctive norms; reference groups; pluralistic ignorance; social influence; norm perception; social sanctions; norm change; cultural tightness; norm interventions; dynamic norms; network experiments; Artificial Era
The Scientific Object of Social Norms and Normative Expectations
Social norms organize conduct through expectations attributed to other members of a relevant social group. The central psychological problem is how an individual forms beliefs about what others do, what others approve, and what consequences follow compliance or violation. A norm may operate in a household, professional community, neighborhood, online network, or political institution, and its force changes with the identities of the people whose expectations matter. Contemporary work treats reference groups, situational cues, conditional preferences, and anticipated responses as research variables rather than interchangeable descriptions of conformity. The conceptual synthesis of Legros & Cislaghi (2020) shows why distinctions among collective practices, beliefs, and social obligations are indispensable.
The scientific object is the present organization and causal operation of norms: the structure of normative and empirical expectations; their measurement; their relation to observable acts; and the conditions under which shared expectations persist, fragment, or change. Norm acquisition across development is treated as an antecedent, while the full psychology of minority influence, formal obedience, and social learning has separate research domains. A person can know a norm yet decline to follow it, follow it privately, imitate it without endorsement, or endorse it while assuming incorrectly that peers disagree. An adequate model therefore distinguishes perceived prevalence, perceived approval, personal attitudes, anticipated sanctions, and actual behavior Bicchieri (2017).
This monograph critically integrates original experiments, field interventions, cultural comparisons, and cross-disciplinary theoretical research. It reports a narrative evidence synthesis rather than a preregistered systematic review or a new quantitative meta-analysis. Research claims are calibrated to their designs: randomized messages identify effects of those messages in their tested settings, while survey correlations and cross-national comparisons support more limited inferences. The organizing question is how expectations become psychologically consequential and under what conditions accurate or corrected information can transform a social practice Miller & Prentice (2016) Tankard & Paluck (2016).
Conceptual Boundaries: Customs, Conventions, Moral Rules, and Social Norms
A recurring behavior is not automatically a social norm. Some repetitions follow material constraints, habitual efficiencies, individual preferences, or legal requirements. Others persist because actors condition choices on what they believe their reference group does or expects. The diagnostic approach in Bicchieri (2017) asks whether an actor's preference is conditional on empirical expectations, normative expectations, or both. This makes social interdependence an empirically testable proposition. Investigators can ask whether privately changing beliefs about relevant others changes behavior while holding personal resources and explicit incentives as constant as feasible.
Conventions such as which side of a path to walk on often solve coordination problems without demanding moral endorsement. An injunctive norm additionally conveys approval, disapproval, or perceived obligation; its existence does not establish that every actor endorses the underlying value. Legal rules may reinforce social expectations, but a law's presence and a community's willingness to sanction violations must be measured separately. The review by Legros & Cislaghi (2020) documents persistent differences between theories that locate norms in individual expectations and those that locate them in stable patterns of collective relations. The levels can be made complementary by specifying how shared institutions generate individual beliefs.
Moral judgment, personal attitude, and norm perception can converge or diverge. An employee may believe a practice is unfair while assuming every colleague approves it; a neighbor may prefer conservation while believing nobody else practices it. Such discrepancies are central explanatory cases. Prentice & Miller (1993) demonstrated systematic differences between students' private discomfort with campus drinking practices and their estimates of peers' attitudes. The result exemplifies why aggregated overt behavior cannot substitute for the direct measurement of private preferences and beliefs about others.
Empirical Expectations and Descriptive Norms
Empirical expectations are beliefs about how members of a relevant group actually behave in a specified situation. In a survey these can be elicited as predicted proportions, frequency estimates, or beliefs about identifiable peers; in an experiment they can be manipulated through credible information about group behavior. The precision of reference group, situation, and time horizon matters. A descriptive message that a majority acted in some way may carry information about feasibility, popularity, social identity, or an apparently sensible strategy, and different mechanisms predict different effects Cialdini et al. (1990) Goldstein et al. (2008).
In dictator-game experiments Bicchieri & Xiao (2009) independently manipulated beliefs about the behavior and approval of other dictators. When the two types of expectations conflicted, empirical expectations significantly predicted giving behavior in the study, whereas normative expectations did not add a statistically significant contribution after accounting for empirical expectations. This is evidence from a particular experimental task, not a universal hierarchy between the two kinds of belief. It demonstrates the value of separate manipulations because the same observed decision can be sustained by different perceived social facts.
Descriptive information can shift a group toward undesirable conduct when it reveals that violations are common. Schultz et al. (2007) showed in a field experiment on residential electricity use that feedback on neighborhood averages encouraged reductions among high-use households while risking a rebound among initially low-use households. The behavioral interpretation depends on baseline position relative to the reported average. Aggregate improvements can therefore coexist with deterioration in an important subgroup. A norm intervention must specify both the target action and the audience's initial distance from the reference behavior.
Normative Expectations, Injunctive Signals, and Approval
Normative expectations concern what a person believes relevant others think one ought to do. They differ from the person's own moral commitment and from simple observations of prevalence. Injunctive cues can include explicit disapproval, symbolic feedback, public statements, visible sanctions, and institutional rules interpreted as legitimate. Cialdini et al. (1990) distinguished injunctive norms from descriptive norms in naturalistic littering studies and demonstrated that their behavioral effects depended partly on which norm was salient. This framework directs researchers to measure both informational content and attention to that content.
Not all people in the reference network carry equal influence. A supervisor's disapproval may change career expectations; a respected peer's approval may change reputational incentives; a family elder's evaluation may influence belonging. These consequences are hypotheses about concrete mechanisms that should be tested with independent measures of actual sanctions, expected sanctions, and behavior. Bicchieri & Chavez (2010) experimentally manipulated public information about available choices in an ultimatum-game variant and found contextual variation in fairness judgments and offers, highlighting the link between normative expectations and perceptions of what counts as an appropriate action.
In the electricity field experiment Schultz et al. (2007), adding an injunctive signal to descriptive feedback eliminated the observed boomerang pattern. This finding is especially informative because it reveals an interaction: telling a low-consuming household that its consumption is below average can inadvertently normalize higher use, while social approval of its existing restraint can stabilize low use. The joint design produces a more specific inference than labeling both messages social influence. A broader conclusion that injunctive cues always counteract descriptive feedback would require independent replication across populations, behaviors, and delivery settings.
Historical Foundations and the Evolution of Norm Research
Research on norms developed across social psychology, sociology, economics, anthropology, and political science. Laboratory traditions investigated how common judgments emerge and how reference groups affect expressed responses; later field experiments measured actual behavioral consequences in natural settings. The development of focus theory provided a decisive conceptual refinement by separating the prevalence of a practice from its approval and by treating attention as a condition of influence. The five natural-setting studies in Cialdini et al. (1990) remain an important illustration of this transition from general talk about conformity to manipulated normative cues.
The 1990s added work on mistaken collective beliefs and the maintenance of behavior through misperceived consensus. In Prentice & Miller (1993), multiple studies of students' attitudes to alcohol use documented pluralistic ignorance and its relation to perceived deviance from peers. The 2000s produced more explicit conditional-preference models and behavioral games, including Bicchieri & Xiao (2009) and Bicchieri & Chavez (2010). Subsequent programs scaled to randomized household letters, network-selected school interventions, and cross-national variation, broadening the units of analysis beyond isolated individuals Allcott (2011) Paluck et al. (2016) Gelfand et al. (2011).
Historical influence is distinguishable from present evidential authority. Foundational paradigms established productive questions, while methodological developments exposed limits in single-campus samples, narrowly defined experimental games, and cross-sectional comparisons. Legros & Cislaghi (2020) mapped disagreements across reviews concerning definitions, mechanisms, development, and relevant actors. A mature history therefore follows changing operational definitions and research designs as well as dates and names. The present synthesis places individual belief formation, group structure, field causality, and cultural institutional context within one coherent analytic sequence.
Norm Salience, Attention, and Situational Interpretation
Norms exert influence through a chain from exposure to interpretation, motivation, and behavior. A person who knows that a practice is common may ignore it when attention is elsewhere or interpret it as irrelevant to a different social identity. The focus theory experiments of Kallgren et al. (2000) varied means of directing attention toward anti-littering norms and found norm-consistent behavior especially under conditions of normative focus. These findings identify attention as a moderator. They do not establish that attention alone generates an enduring shared norm, because focus may change action in the immediate situation without changing the underlying social expectation.
Attention is selective across concurrent norms. A workplace may contain simultaneous expectations for loyalty, openness, efficiency, and procedural fairness. A message can activate one of these considerations while leaving the others unchanged. The psychological mechanism may involve automatic categorization of the relevant group, an inference about likely reactions, or conscious strategic planning. Nolan et al. (2008) found that respondents underestimated the influence of descriptive norm information on their conservation choices even when norm-oriented messaging performed effectively. Self-reported motives thus need not reproduce the causal hierarchy estimated from experimental behavior.
A well-designed study identifies the object of normative focus through a manipulation check, measures behavioral rather than solely verbal compliance, and considers whether the intervention also conveys incentives, identity cues, or factual information. Researchers should distinguish changes in reported norm salience from changes in privately held normative expectations. Repeated measures can assess persistence. Outside a narrow task, an identical numerical claim about group behavior may elicit different interpretations depending on trust in the sender and perceived similarity to the reference group Goldstein et al. (2008) Miller & Prentice (2016).
Reference Groups, Social Identity, and Local Relevance
Normative expectations are indexed to an audience: family members, classmates, neighbors, co-workers, professional peers, cultural communities, or network contacts. A global claim about what people generally do can have weaker behavioral relevance than information about immediately comparable others. In two hotel field experiments Goldstein et al. (2008) found that normative messages were more effective than a conventional environmental appeal and that a situationally specific room-based reference increased the effectiveness of the message. The result suggests that perceived local relevance can matter independently of broad numerical prevalence.
The identification of a reference group creates a measurement problem. Survey respondents may report beliefs about people they admire, people who can punish them, or people visible in a digital feed; these sets need not coincide. The social-ecological framework reviewed by Legros & Cislaghi (2020) distinguishes norm targets, drivers, and beneficiaries or victims, while highlighting disagreement over which actors constitute the relevant group. Elicitation should establish whose judgments matter to the respondent before asking how frequently the norm is followed or approved. Otherwise, apparent individual inconsistencies may reflect comparisons to different groups.
Reference-group boundaries can also become contested. New entrants may adopt some local practices while rejecting others, and members can hold competing identities that activate different obligations. An observed group average describes neither the distribution of preferences within the group nor the reach of influential subgroups. For intervention design, sampling a community's visible behavior without identifying salient referents risks measuring a population whose practices have little psychological relevance to the target actors. Network-informed designs such as Paluck et al. (2016) address this question by measuring who receives attention before selecting intervention messengers.
Sanctions, Reputation, Reciprocity, and Internalized Obligation
Some norms persist through anticipated approval, shame, exclusion, loss of trust, or material penalties. Others become embedded in personal identity and are followed even without observation. The key empirical task is to distinguish the belief that a violation will be judged negatively from the belief that punishment will actually be imposed. The conditional-preference framework of Bicchieri (2017) distinguishes expected behavior and expected approval and asks under what conditions those beliefs matter to a decision. Norms with similar public behavioral compliance can depend on very different private incentives.
Reputational considerations are especially relevant when behavior is observable, audiences communicate, and future exchange depends on social standing. Experimental designs can vary privacy, audience identity, sanction opportunities, and repeated interaction to test these pathways. Yet public visibility can also change perceived accountability through institutional rules rather than informal norms. A researcher who observes larger compliance under public conditions must distinguish anticipated social judgment, direct surveillance, and explicit legal punishment. Bicchieri & Chavez (2010) illustrates how information about the available choice set can alter fairness expectations, thereby changing the meaning of an observed response.
Internalized obligation adds a temporal dimension. A person may eventually treat a norm as self-endorsed, sustaining it without immediate external evaluation, while another complies only when sanctions seem likely. Cross-sectional attitudes cannot reconstruct this developmental trajectory on their own. A stronger account requires longitudinal observation of private endorsement, perceived social expectations, and behavior as audiences and incentives change. This extends the social-norm framework toward moral development and identity research without collapsing either neighboring domain into the present dossier Legros & Cislaghi (2020) Bicchieri (2017).
Pluralistic Ignorance, False Consensus, and Misperceived Approval
Pluralistic ignorance occurs when individuals privately reject or question a practice yet mistakenly believe that others endorse it. Public compliance then becomes misleading evidence, because each person's observation of apparently willing participation reinforces the mistaken estimate of group approval. Prentice & Miller (1993) documented this pattern in campus alcohol attitudes: students perceived themselves as more uncomfortable with prevailing drinking practices than their peers. The study demonstrates an epistemic mechanism of norm persistence, while its campus setting limits direct generalization to other social practices.
Not every misperception is pluralistic ignorance. A person may overestimate how often peers engage in a behavior while accurately estimating approval, or underestimate widespread approval when the behavior is rare due to external constraints. Careful measurement separates the actual distribution of private attitudes, the actual distribution of behavior, and people's second-order beliefs about those distributions. Bursztyn et al. (2020) studied misperceived support for women's work outside the home among young married men in Saudi Arabia, using private reports and information interventions to examine the behavioral consequences of correcting estimates of peers' attitudes.
Belief-correction experiments show the promise and limits of informational change. Randomized disclosure can identify whether recipients revise beliefs or decisions, but treatment effects may also depend on whether the information seems credible and whether respondents face real opportunities for action. In Bursztyn et al. (2020) the intervention increased men's willingness to support job-search assistance, while follow-up evidence connected revised beliefs to wives' subsequent job-search activity. These effects were observed within specific institutional and family arrangements. Transferring the same approach elsewhere requires measuring local control over opportunities, legal constraints, household negotiation, and reference-network structure.
Measurement Architecture and Construct Validity
Measurement begins with a specified action, context, population, time period, and reference group. Researchers should separately elicit a respondent's personal preference, own reported action, perceived prevalence, perceived approval, expected sanction, and confidence in those estimates. A statement such as 'people around here think this is normal' conflates behavior, approval, and typicality. The distinctions in Bicchieri (2017) and the synthesis of Legros & Cislaghi (2020) imply that a single generic norm item is rarely sufficient to identify the normative pathway under study. Qualitative elicitation can reveal which groups and consequences matter before standardized questions are written.
Construct validity requires a relationship between an operational indicator and the explanatory claim. A respondent's guess about group behavior can measure an empirical expectation, but it also may reflect projection of personal behavior, exposure to selective examples, or misunderstanding of the question. Normative expectations can be estimated with parallel questions asking what peers believe should happen, with separate private attitude samples establishing the comparison standard. Prentice & Miller (1993) illustrates the need to compare respondents' estimates with the actual distribution of private attitudes instead of interpreting every perceived social difference as factual.
Repeated measurement supplies evidence about stability and contextual sensitivity. The same individual may express a different norm perception when surveyed privately, in front of coworkers, after reading group feedback, or at another life stage. Researchers should document question wording, confidentiality, anchoring, reference population, and measurement invariance. Disagreement between modes may be theoretically informative rather than measurement error; public reports can themselves be a social act. Studies that label a scale 'social norm' should report whether it represents descriptive frequency, injunctive approval, sanction expectations, or a tested multidimensional construct Tankard & Paluck (2016) Legros & Cislaghi (2020).
Experimental Identification of Norm Effects
To establish a causal effect of norm information, investigators can randomize descriptive messages, injunctive messages, their combination, and an appropriate nonnormative control. Factorial designs test whether one type of information amplifies, replaces, or weakens another. Outcome measures should reflect actual target behavior when ethically and practically feasible. In Schultz et al. (2007), baseline electricity consumption revealed heterogeneous reactions to descriptive feedback, and the injunctive signal addressed the upward adjustment among low consumers. This illustrates why average treatment effects may conceal opposing subgroup mechanisms.
A randomized message establishes a causal effect of being assigned that message, subject to randomization quality and missingness; it does not necessarily isolate the latent psychological mechanism. Credibility, social identity, salience, informational usefulness, and demand effects may vary together. Manipulation checks should therefore measure changes in the intended beliefs and alternative mediators without assuming that a posttreatment correlation proves mediation. A follow-up interval can separate transient cue responses from persistent norm revision. The comparative evidence reviewed by Miller & Prentice (2016) emphasizes matching message content, target population, and behavioral opportunity.
Natural experiments and field randomization strengthen ecological relevance but introduce implementation variability. In Allcott (2011), randomized Home Energy Reports across a large set of households yielded modest average electricity reductions; the reports bundled neighbor comparisons with other features, so attribution to a single psychological component requires caution. In Paluck et al. (2016), randomized school interventions were combined with network mapping and an assessment of conflict reports. These studies illustrate different estimands, units of randomization, and outcome types. Their effect estimates should not be pooled informally as if they measured one context-independent 'norm effect'.
Field Evidence in Energy, Environmental Practice, and Consumer Settings
Environmental decisions offer clear behavioral outcomes and repeated exposure to social information. Goldstein et al. (2008) compared hotel towel-reuse appeals and found descriptive norm messages effective, especially with locally relevant comparisons. The behavior was inexpensive and observable within a short stay, which may make results sensitive to opportunity and immediate context. Hotel guests were not asked to revise all their environmental values; the experiment examined a situated consumer act. The scientific contribution lies in demonstrating that appropriately framed comparisons can move behavior without requiring a prior general environmental attitude shift.
Household energy research extends the setting to repeated consumption over time. Nolan et al. (2008) documented a gap between the influence of norm messages and respondents' retrospective attribution of their motives. Schultz et al. (2007) identified boomerang risk from a raw average comparison. Allcott (2011) examined a much larger randomized utility-letter program, reported an average reduction in consumption, and documented variation by prior usage. Together these studies establish meaningful behavioral sensitivity to peer comparison while revealing that the effective intervention package, population, and measurement horizon differ substantially across sites.
Conservation studies also warn against careless extrapolation. A household with poor insulation, different occupancy, or constrained appliance access may have a higher energy baseline for reasons unrelated to willingness to conform. Targeted comparisons can misclassify such households if contextual covariates are ignored. Feedback may motivate some participants while generating anger or disbelief among others, and downstream savings depend on persistence. An ethically credible intervention should report the distribution of effects, preexisting differences, contact frequency, the reliability of comparison data, and whether recipients can understand or challenge the reference standard Miller & Prentice (2016) Allcott (2011).
Network Experiments, Social Referents, and Peer Visibility
Social networks alter the exposure and credibility of norm signals. The number of people who endorse a practice is one variable; their location in a communication network is another. Paluck et al. (2016) mapped peer ties in 56 schools before randomly assigning an anti-conflict intervention to selected students in intervention schools. The design allowed researchers to compare effects associated with more socially visible participants. The reported reduction in disciplinary conflict reports supports the relevance of network placement, though disciplinary records are an administrative outcome shaped by reporting and institutional practices.
Social referents can transmit beliefs about what is typical or approved without directly punishing deviations. This is a plausible mechanism behind diffusion from central or conspicuous actors to less visible peers. The school experiment found more favorable results when seed groups contained students identified as social referents Paluck et al. (2016). Network position, however, is conditional on the measurement instrument and social setting: a friendship nomination may identify a different audience than a class chat or workplace reporting line. Norm diffusion claims should identify which tie type and attention process were actually observed.
Digital networks add ranking systems that shape which examples become visible. An individual may infer widespread approval from a feed that amplifies highly engaging posts rather than from a representative sample of actual opinion. The mechanism is conceptually related to reference-group inference, but feed ranking, bots, and algorithmic recommendation require independent measurement before causal effects are attributed to a norm. The present dossier treats them as extensions of the expectation problem and reserves detailed platform-specific causal designs for adjacent research on digital social influence Tankard & Paluck (2016) Legros & Cislaghi (2020).
Evidence Appraisal, Replication, and Statistical Inference
Evidence quality depends on matching the claim to the research design. A controlled laboratory manipulation can isolate a particular informational contrast under standardized conditions. A field trial tests the delivered intervention among people facing real constraints. Cross-national comparisons describe macro-level patterns while leaving many alternative explanations available. A review of social norms must therefore distinguish experiments on perceived approval, interventions that bundle multiple cues, survey estimates of social belief, and qualitative studies of lived sanction systems. The overview by Legros & Cislaghi (2020) documents conceptual heterogeneity that makes this comparison especially necessary.
Important appraisal criteria include allocation concealment when feasible, preregistration or transparently reported hypotheses, attrition, clusters and interference, baseline balance, measurement reliability, and confidence intervals. Contagion is especially relevant: people assigned to different conditions may share information, so unit-level independence cannot be assumed. In network experiments such as Paluck et al. (2016), spillovers are often substantive effects rather than nuisances. In household trials such as Allcott (2011), treatment heterogeneity should be reported alongside the overall mean. Outcome measures should not be treated as interchangeable merely because all are described as norm compliance.
Replication is an inferential program, not a single yes-or-no status. Conceptual replications can test whether behavioral direction persists when reference groups, cultural institutions, and delivery channels change. Direct replications of a specific message require careful matching of exposure, measurement, and outcome definitions. Publication bias and selective reporting can distort the visible literature. Miller & Prentice (2016) offers a critical synthesis of intervention mechanisms, while Legros & Cislaghi (2020) shows how distinct definitions can produce apparent theoretical contradictions. The present manuscript makes no new pooled estimate and does not claim a comprehensive exhaustive search.
Cultural Variation, Tightness–Looseness, and Institutional Context
Normative expectations vary in strength, content, and sanction structure across communities. A frequent methodological error is to treat national categories as homogeneous psychological populations. In a cross-national investigation involving 33 nations, Gelfand et al. (2011) characterized cultural tightness and looseness through the strength of prevailing norms and tolerance of deviance, relating these patterns to broader ecological and institutional conditions. The comparisons illuminate meaningful macro-level variation, but the associations do not by themselves identify which historical or policy factors caused a given society's present pattern.
Within-country and within-group heterogeneity remains consequential. Urban neighborhoods, professions, generations, and subcultures can have divergent reference groups and distinct opportunities for sanction. A respondent's broad country-level score does not identify the obligations operating in a specific workplace meeting or family negotiation. Legros & Cislaghi (2020) argues for explicit consideration of relevant actors and the social meaning of practices. Comparative investigations should therefore estimate both between-context variation and the distributions within contexts, using translated instruments whose psychological meaning is assessed rather than presumed equivalent.
Cross-cultural work also faces selection and expression biases. Privacy, trust in researchers, social desirability, and the availability of safe dissent can differ substantially, affecting apparent levels of consensus. The experimental work on labor-force norms in Saudi Arabia Bursztyn et al. (2020) illustrates how private beliefs and visible practice can diverge under institutional constraints. Generalizing a belief-correction result requires knowledge of the local reference group, control over opportunities, and the way the intervention information is received. Norms are historically situated expectations, not timeless national personality traits.
Gender, Status, and Asymmetrical Costs of Norm Violation
Norms may establish expectations about occupations, household decisions, leadership, appearance, and public speech. Their influence cannot be evaluated solely from the perspective of those who are publicly visible; individuals who bear restrictions may have less ability to express dissent or to act on private preferences. Bursztyn et al. (2020) examined a specific example in which young married men underestimated peers' support for women's work outside the home. When corrected, men's actions related to supporting wives' job searches changed. This provides causal evidence about the information intervention in that setting, while the surrounding institution of household decision-making remains essential context.
A prescriptive gender norm can be measured as perceived approval of a behavior for different targets, expected reputational consequences, and observed choices when opportunities are experimentally or naturally varied. The use of identical questions for different genders is necessary for many comparisons, but attention must also be given to whether the action carries the same resources and hazards across groups. Observed lower participation can result from external opportunity constraints even when approval is widespread. The diagnosis of a norm requires separating what people want, what others expect, and what actors can actually do Bicchieri (2017) Legros & Cislaghi (2020).
Status differences also influence sanctions and norm transmission. A violation may be affordable to a high-status individual and costly to a low-status member of the same community. This asymmetric enforcement can make an apparently stable norm depend on unequal power, while survey averages obscure the distribution of risk. Social interventions should document intended beneficiaries, people who enforce the practice, and those who experience its costs. Legros & Cislaghi (2020) highlights this distinction among targets, drivers, beneficiaries, and victims. Scientific neutrality here requires specifying the social mechanism and its consequences accurately, including when collective expectations sustain coercive or discriminatory arrangements.
Pluralism, Group Conflict, and Changing Normative Majorities
Members of a society often belong to overlapping reference networks with incompatible norms. A professional duty of confidentiality may conflict with expectations for public disclosure; a local practice may conflict with institutional law; a family expectation may diverge from professional standards. The practical question is which audience becomes psychologically salient, which sanctions appear credible, and which opportunities allow an individual to act. The focus-theory evidence in Kallgren et al. (2000) suggests that situational activation can influence which norm guides a decision. It does not determine how individuals resolve every enduring moral or institutional conflict.
Perceived consensus can change before the underlying distribution of private approval changes, and the opposite sequence is also possible. A group may develop widespread private disagreement while continuing to signal public adherence. Prentice & Miller (1993) supplies a classic example of misalignment between private attitudes and perceived peer attitudes. Interventions that make existing disagreement visible can therefore differ from campaigns seeking to change personal evaluations. A careful longitudinal design records private preferences, public communication, expressed expectations, and behavioral outcomes in parallel, allowing researchers to examine their temporal order.
Normative majorities are partly products of information systems. When voices differ in visibility or reputational risk, perceived popularity may not correspond to the number of people who endorse a practice. Tankard & Paluck (2016) treats perceived norms as a vehicle for social change, showing the relevance of signals that a social consensus has shifted. Yet announcements by an institution, a spokesperson, or a platform require credibility: otherwise the message may be interpreted as propaganda or an isolated opinion. The empirical issue is the relationship between an observer's reference group and the source of the apparent consensus.
Norm Change, Critical Mass, and Coordination Dynamics
Norm change may begin with alterations in private preferences, public behavior, observable sanctions, or the distribution of information about relevant others. A small set of people can sometimes coordinate a new convention if their interaction network allows the alternative practice to become visible and mutually expected. Centola et al. (2018) experimentally studied the dynamics of social convention in a structured interaction system and demonstrated a tipping point under the tested conditions. The result provides evidence that minority actions can alter a coordination equilibrium; it does not supply a universal threshold for social change in every institutional or cultural setting.
Critical-mass claims require careful attention to mechanism and network structure. A measured threshold can depend on the density of ties, repeated encounters, initial convention strength, the observability of commitments, and the availability of alternative behaviors. A laboratory coordination game abstracts away from many material and political constraints that shape everyday practices. Centola et al. (2018) explicitly investigated a controlled social convention system. It would be a category error to infer from that experiment an invariant percentage of activists or employees required to transform any real-world norm. The correct object of generalization is a family of contingent coordination dynamics.
Information about changing behavior can be influential even when the new conduct is currently uncommon. In five experiments Sparkman & Walton (2017) found that dynamic norm messages describing how people were changing helped motivate behavior favorable to environmental sustainability; one field experiment increased meatless orders in a café. This suggests that trajectories may communicate future expectations and signal emerging social value. The effect should be distinguished from the assertion that a new behavior is already prevalent. Static prevalence, perceived direction of change, and confidence in future adoption are separate measurement targets.
Interventions Based on Norm Feedback and Belief Correction
Norm feedback interventions tell recipients how a relevant group behaves or what it privately approves. Their success depends on measuring the group accurately, choosing an interpretable comparison, and delivering the information through a trusted channel. Miller & Prentice (2016) surveys interventions in student drinking and environmental behavior and argues that effectiveness depends on the form and target of norm information. The research base supports conditional, design-sensitive applications rather than a universal message formula. A program should specify whether the intended causal mediator is perceived prevalence, normative approval, future expectations, or sanction beliefs.
Belief correction is particularly promising when a measurable gap exists between true preferences and widespread perceptions. In Bursztyn et al. (2020), randomized information about peer support for wives' labor-force participation changed related household decisions and subsequent job-search behavior. Other contexts may contain accurate pessimistic beliefs rather than pluralistic ignorance, in which case simple correction cannot remove an underlying constraint. Even when beliefs are inaccurate, messaging may fail if the audience distrusts the data or doubts its relevance. A credible program thus begins with representative baseline elicitation and independent validation of the reported information.
Feedback can also reveal norms that researchers would prefer to change. Schultz et al. (2007) makes this risk concrete: a descriptive average can pull already-protective actors toward the mean. Intervention designers should test the joint influence of descriptive and injunctive cues, disclose baseline heterogeneity, and monitor inadvertent normalization of harmful behavior. A trial can predefine both desired outcomes and adverse movements in groups initially practicing the preferred behavior. Scientific evaluation should record whether the reported comparison remains truthful and whether repeated delivery maintains its effect without escalating perceived surveillance.
Dynamic Norms and the Communication of Social Change
Dynamic norms communicate perceived movement in behavior over time rather than only the current cross-sectional majority. Such messages may make emerging practices appear feasible or indicate that public expectations will change. Sparkman & Walton (2017) tested dynamic norm information in online and field experiments, observing increased interest in reducing meat consumption and favorable behavior in some settings. These experiments distinguish a trajectory cue from a static statement that most people already perform the target action. The message's predictive force depends on whether recipients believe the stated trend and identify with the relevant group.
Trend information can also mislead when a short-term fluctuation is framed as durable social transformation. Researchers should establish the reference time window, data source, and uncertainty in the direction of change. The same percentage increase may be negligible in a small subgroup or socially salient in a visible network. A stronger study compares static, dynamic, and combined messages while independently measuring expected future behavior and perceived approval. The interpretation of Sparkman & Walton (2017) remains context dependent: evidence of effective dynamic messages in sustainability tasks is a basis for new testing elsewhere, not automatic generalization.
Communication can create feedback loops. People who believe a norm is changing may act differently, and the resulting visibility may change others' expectations. To identify such reciprocal pathways, studies need more than individual self-reports collected immediately after a message. They need temporally ordered measurements, network exposure information, and behavioral outcomes. The coordination experiments in Centola et al. (2018) and the perception-centered analysis in Tankard & Paluck (2016) give complementary frameworks for studying belief diffusion and changes in collective equilibrium. They should be integrated through explicit models rather than a single broad claim about viral social change.
Second-Order Beliefs, Social Expectations, and Energy Behavior
First-order beliefs concern one's own judgment; second-order beliefs concern what others think or expect. Their measurement matters because a person may support a socially desirable practice yet assume the community does not value it. In research on household energy conservation, Jachimowicz et al. (2018) analyzed the relevance of second-order normative beliefs. The article contributes a specific theoretical and empirical focus on the perceived attitudes of others, extending the distinction between one's own values and expectations attributed to the social environment. It should not be equated automatically with sanction expectations or a direct demonstration of coercion.
Across conservation studies, several constructs co-occur: information about comparable neighbors, perceived approval, awareness of behavioral options, identity with environmental groups, and anticipated future conditions. Allcott (2011) tested large-scale household reporting; Nolan et al. (2008) contrasted persuasive effects with participants' retrospective judgments about why they acted; Schultz et al. (2007) demonstrated why low-use households can react differently from high-use households. Together these findings motivate multi-construct experimental designs and reject a simple reading of every energy saving as evidence of an identical social mechanism.
Testing higher-order beliefs requires asking who attributes which expectation to whom. A measure that asks what neighbors do is descriptive. A measure asking whether neighbors think conservation is important captures perceived evaluations. A measure asking whether neighbors expect a participant to conserve approaches a normative expectation, while another asking whether neighbors would disapprove of waste estimates anticipated judgment. These distinctions become crucial when claims are compared across experiments using different questionnaires Bicchieri (2017) Jachimowicz et al. (2018). Methodological convergence requires an explicit map of constructs and items.
Digital Visibility, Algorithmic Curation, and Artificial Social Signals
Digital platforms mediate apparent consensus through rankings, recommendations, engagement statistics, and selective presentation of messages. These interfaces can make a subset of conduct appear representative even when the observed feed is not a probability sample of the social world. The psychological phenomenon remains an inference about what relevant others do or approve; the technological system changes the evidence from which that inference is drawn. Tankard & Paluck (2016) shows why changes in perceived consensus may alter behavior, while Legros & Cislaghi (2020) provides the conceptual caution required to separate social expectations from norms in the environment.
Artificial agents can also display approving language, simulate group membership, or summarize supposed community beliefs. Whether a person treats such messages as statements about human normative expectations is an empirical question. A system-generated approval signal could be interpreted as an institutional rule, a prediction of popularity, a persuasive recommendation, or the expression of a simulated conversational persona. These interpretations should be measured separately. The Artificial Era provides new interfaces through which expectations are represented and conveyed; their psychological consequences require controlled studies of provenance, disclosure, trust, identity, and behavioral responses.
An appropriate experiment could vary the source of identical normative information among a human peer, a platform summary, and an artificial interlocutor while holding the factual content constant. Outcome measures could include belief revision, source credibility, reliance on the information, and choices with real consequences. This is a proposed research design rather than a reported completed experiment. Existing evidence on reference groups and communication networks Goldstein et al. (2008) Paluck et al. (2016) provides hypotheses, while platform selection and artificial-source attribution remain distinct sources of uncertainty. A claim of artificial normative agency would require an explicit operational definition and independent evidence.
Ethics, Manipulation, Accuracy, and Informed Agency
Norm interventions influence action by changing beliefs about other people's behavior or approval. This creates obligations of evidential accuracy and transparency. A message that exaggerates the frequency of a desired practice can manufacture a false descriptive norm, and a message implying sanctions where none exist can distort perceptions of communal approval. Miller & Prentice (2016) and Bicchieri (2017) emphasize the relevance of diagnosing the target norm and choosing a suitable intervention. Scientific integrity requires documented samples, clear comparison groups, and an account of how message claims were verified.
Ethical evaluation also concerns whose behavior is being changed and for whose benefit. Correcting false beliefs about widespread support may expand an individual's feasible choices, while amplifying a punitive norm may deepen exclusion. Effects must be measured for differently situated actors, including people unable to comply because of structural barriers. Legros & Cislaghi (2020) highlights beneficial and harmful social expectations within one framework. A program's success cannot be inferred solely from increasing conformity; the outcome must be justified by independent scientific and ethical considerations.
Transparency affects both validity and autonomy. Participants should know when normative statements are estimates, when data are uncertain, and when a synthetic source is presenting information about social behavior. In research, debriefing and privacy protection are especially important when eliciting beliefs about sensitive practices or sanctions. Misrepresentation may compromise both trust and interpretation of experimental effects. These requirements extend to artificial interfaces that deliver norm cues: a clear distinction between a measured community attitude and a generated persuasive statement preserves the meaning of the evidence being communicated.
Developmental Trajectories and the Reproduction of Expectations
Individuals acquire norm-relevant information through observation, instruction, participation, and responses from valued social partners. Developmental change can alter both which reference groups count and how an individual interprets apparent approval. A child who treats a caregiver's reaction as decisive may later orient toward classroom peers or institutional rules. The present research problem is the resulting organization of expectations at a given time and its subsequent revision; the separate developmental process of learning a norm belongs to social-learning research. Conceptual accounts in Bicchieri (2017) and Legros & Cislaghi (2020) indicate why prior learning histories and current conditional preferences must be measured as different constructs.
Adolescence often intensifies concern with peer inclusion, but adolescents occupy multiple groups with distinct standards, and an aggregate age effect cannot establish the expectation governing any specific behavior. Network intervention studies show a way to assess peer visibility and influence in concrete settings. Paluck et al. (2016) examined school networks and the influence of socially central intervention participants on conflict-related outcomes. Longitudinal studies should follow changes in friends, schools, roles, and private evaluations while measuring perceived approval and observable behavior separately. This would clarify whether apparent developmental shifts reflect changing beliefs, changing reference groups, or changing opportunities.
Across adulthood, changes in employment, migration, family roles, or participation in digital communities may reorganize normative audiences. Such transitions create opportunities to observe the same person's beliefs across institutions. A causal developmental account requires appropriate repeated measurement, comparable instruments, and plausible tests of alternative explanations. The cross-context variation documented in Gelfand et al. (2011) and the actor-centered synthesis in Legros & Cislaghi (2020) motivate multilevel work that follows both individuals and changing group settings. The existence of a stable practice over generations cannot by itself establish that the same psychological mechanism preserves it.
Competing Explanations and Diagnostic Failures
Similar behavior can arise from conforming to an injunctive norm, inferring useful information from a majority, acting on a personal preference, responding to material incentives, or avoiding punishment by an authority. These explanations differ in their causal implications. Bicchieri & Xiao (2009) demonstrates the value of separating empirical from normative expectations experimentally, while Cialdini et al. (1990) shows that activating the prevalence of conduct differs from activating what people approve. Inference is weakened when investigators present a prevalence statistic and then infer an unmeasured desire for social approval from a subsequent behavioral change.
Responses to an experimental message may also arise from demand characteristics, perceived experimenter expectations, altered beliefs about practical feasibility, or a generalized reminder to conserve resources. Bundled interventions multiply these possibilities. In Allcott (2011), household reports provided comparative information in a practical delivery program; identifying the contribution of each element requires a dedicated component comparison. In Goldstein et al. (2008), the contrast between specific normative appeals and a conventional environmental message was clearer, but the underlying inference still depends on the plausibility of alternative cues. Mechanism tests should be preregistered and evaluated against rival models.
A further diagnostic error is treating consensus as uniform. Norm perceptions are distributions, often marked by disagreement, uncertainty, and unequal visibility. Prentice & Miller (1993) established the importance of contrasting private views with estimates of group approval, and Bursztyn et al. (2020) identified settings in which a perception gap carried behavioral consequences. Researchers should ask whether respondents know the reference group's actual attitude distribution, whether uncertainty is concentrated among isolated participants, and whether visible behavior is constrained. Accurate measures of disagreement are themselves fundamental scientific observations.
Scientific Integration Across Levels of Explanation
At the individual level, social norms involve cognitive representations of group behavior, second-order beliefs, situational attention, and preferences conditional on expectations. At the relational level, they involve communication, reputation, reference networks, and asymmetric sanctions. At the institutional level, they interact with laws, opportunities, public narratives, and distributed resources. Legros & Cislaghi (2020) identifies the relationship between individual and collective conceptualizations as a central debate. An integrated model links these levels through observable processes, including exposure to a signal, a measured belief revision, changed action, and subsequent changes in what others can observe.
Three empirical traditions provide complementary tests. Laboratory experiments distinguish causal message components and belief types Bicchieri & Xiao (2009) Cialdini et al. (1990). Naturalistic interventions measure behavior under practical constraints Schultz et al. (2007) Paluck et al. (2016). Comparative and network research locates individuals inside social and institutional systems Gelfand et al. (2011) Centola et al. (2018). The traditions should be combined by specifying the dependent variable and level of randomization, rather than interpreting their varied outcomes as replications of one universal effect. A model that predicts perceived approval is not automatically a model that predicts energy use, workplace behavior, or collective mobilization.
A strong scientific synthesis preserves the distinction between a shared practice and the psychological processes that reproduce it. The same aggregate regularity may be sustained by different combinations of descriptive expectations, perceived obligations, sanctions, and individual commitments. Change may occur when any one of these components shifts or when their relationships are reorganized. The explicit separation of mechanisms enables cumulative research: different laboratories can compare comparable expectation measures even when the substantive practice differs. Miller & Prentice (2016) and Tankard & Paluck (2016) show how interventions and perceptions of consensus can be studied within such a framework.
Unresolved Problems and an Empirical Research Agenda
Several central questions remain open. Researchers need more reliable methods for discovering the relevant reference group without presupposing that a convenient census category is psychologically decisive. They need better instruments for measuring uncertainty, anticipated sanction, and disagreement within networks, and stronger cross-cultural evidence that measures mean the same thing across languages and institutional settings. Legros & Cislaghi (2020) catalogues theoretical disagreements that still shape these problems. The next methodological advance is likely to come from combining elicitation, repeated individual observation, experimentally varied information, and social-network mapping with transparent validity tests.
Another open problem concerns persistence. Many experiments measure choices immediately after a normative signal, whereas sustained change requires altered expectations across repeated encounters. The time scale of observation should be justified by the proposed mechanism. Sparkman & Walton (2017) studied messages about changing behavior, and Centola et al. (2018) studied coordination tipping in a structured system; future work could test when expectations about the future convert transient action into stable collective practice. Such designs must distinguish delayed effects, selective attrition, network diffusion, and the actual replacement of a social rule.
Artificially mediated communication introduces a further agenda. Platform ranking and synthetic responses can supply apparently normative information that is abundant, selectively visible, or generated. Researchers should measure whether recipients perceive the message as evidence about human peers, as an institutional instruction, or as advice from an artificial interlocutor. A randomized study could manipulate content provenance while measuring belief accuracy, action, and trust, with explicit disclosure of synthetic materials. This is a proposed research program. Its grounding lies in the demonstrated importance of salience, credible reference groups, and expectation correction Kallgren et al. (2000) Goldstein et al. (2008) Bursztyn et al. (2020).
Conclusion
Social norms are patterns of socially interdependent expectations through which people interpret what relevant others do, endorse, and may sanction. The research reviewed here establishes that empirical expectations, normative expectations, personal preferences, and observable practices are distinct variables with consequential relationships. Laboratory studies demonstrate the influence of differently framed normative cues; field experiments show changes in energy behavior, everyday conservation, school conflict, and household decision-making; comparative research reveals variation in institutional and cultural constraint Cialdini et al. (1990) Schultz et al. (2007) Paluck et al. (2016) Bursztyn et al. (2020) Gelfand et al. (2011).
The main scientific synthesis is that the causal force of a norm depends on the audience, the informational content and credibility of the signal, the actor's situation, the distribution of sanctions, and opportunities for action. Misperceived approval can preserve publicly visible practices that many privately question, while accurate feedback can change behavior under identifiable conditions. Dynamic messages and coordination experiments provide evidence for pathways of norm change, with thresholds and effects dependent on specific designs Prentice & Miller (1993) Sparkman & Walton (2017) Centola et al. (2018). These findings support a multidimensional account rather than an assumption that all group influence operates through one psychological motive.
The next stage of research requires more precise measurement of reference groups and second-order beliefs, long-term causal designs, comparisons across social positions and cultures, and explicit testing of digital and artificial channels through which expectations are represented. These directions preserve the central insight of contemporary norm science: to explain a collective regularity one must identify what participants believe about one another and how those beliefs become consequential for action. The analytical framework of Bicchieri (2017) and the interdisciplinary review in Legros & Cislaghi (2020) provide a coherent foundation for this continuing program.
References
- Allcott, H. (2011). Social norms and energy conservation. Journal of Public Economics, 95(9–10), 1082–1095. https://doi.org/10.1016/j.jpubeco.2011.03.003
- Bicchieri, C. (2017). Norms in the wild: How to diagnose, measure, and change social norms. Oxford University Press. https://doi.org/10.1093/acprof:oso/9780190622046.001.0001
- Bicchieri, C., & Chavez, A. (2010). Behaving as expected: Public information and fairness norms. Journal of Behavioral Decision Making, 23(2), 161–178. https://doi.org/10.1002/bdm.648
- Bicchieri, C., & Xiao, E. (2009). Do the right thing: But only if others do so. Journal of Behavioral Decision Making, 22(2), 191–208. https://doi.org/10.1002/bdm.621
- Bursztyn, L., González, A. L., & Yanagizawa-Drott, D. (2020). Misperceived social norms: Women working outside the home in Saudi Arabia. American Economic Review, 110(10), 2997–3029. https://doi.org/10.1257/aer.20180975
- Centola, D., Becker, J., Brackbill, D., & Baronchelli, A. (2018). Experimental evidence for tipping points in social convention. Science, 360(6393), 1116–1119. https://doi.org/10.1126/science.aas8827
- Cialdini, R. B., Reno, R. R., & Kallgren, C. A. (1990). A focus theory of normative conduct: Recycling the concept of norms to reduce littering in public places. Journal of Personality and Social Psychology, 58(6), 1015–1026. https://doi.org/10.1037/0022-3514.58.6.1015
- Gelfand, M. J., Raver, J. L., Nishii, L., Leslie, L. M., Lun, J., Lim, B. C., Duan, L., Almaliach, A., Ang, S., Arnadottir, J., Aycan, Z., Boehnke, K., Boski, P., Cabecinhas, R., Chan, D., Chhokar, J., D'Amato, A., Ferrer, M. S., Fischlmayr, I. C., ... Yamaguchi, S. (2011). Differences between tight and loose cultures: A 33-nation study. Science, 332(6033), 1100–1104. https://doi.org/10.1126/science.1197754
- Goldstein, N. J., Cialdini, R. B., & Griskevicius, V. (2008). A room with a viewpoint: Using social norms to motivate environmental conservation in hotels. Journal of Consumer Research, 35(3), 472–482. https://doi.org/10.1086/586910
- Jachimowicz, J. M., Hauser, O. P., O’Brien, J. D., Sherman, E., & Galinsky, A. D. (2018). The critical role of second-order normative beliefs in predicting energy conservation. Nature Human Behaviour, 2, 757–764. https://doi.org/10.1038/s41562-018-0434-0
- Kallgren, C. A., Reno, R. R., & Cialdini, R. B. (2000). A focus theory of normative conduct: When norms do and do not affect behavior. Personality and Social Psychology Bulletin, 26(8), 1002–1012. https://doi.org/10.1177/01461672002610009
- Legros, S., & Cislaghi, B. (2020). Mapping the social-norms literature: An overview of reviews. Perspectives on Psychological Science, 15(1), 62–80. https://doi.org/10.1177/1745691619866455
- Miller, D. T., & Prentice, D. A. (2016). Changing norms to change behavior. Annual Review of Psychology, 67, 339–361. https://doi.org/10.1146/annurev-psych-010814-015013
- Nolan, J. M., Schultz, P. W., Cialdini, R. B., Goldstein, N. J., & Griskevicius, V. (2008). Normative social influence is underdetected. Personality and Social Psychology Bulletin, 34(7), 913–923. https://doi.org/10.1177/0146167208316691
- Paluck, E. L., & Green, D. P. (2009). Deference, dissent, and dispute resolution: An experimental intervention using mass media to change norms and behavior in Rwanda. American Political Science Review, 103(4), 622–644. https://doi.org/10.1017/S0003055409990128
- Paluck, E. L., Shepherd, H., & Aronow, P. M. (2016). Changing climates of conflict: A social network experiment in 56 schools. Proceedings of the National Academy of Sciences, 113(3), 566–571. https://doi.org/10.1073/pnas.1514483113
- Prentice, D. A., & Miller, D. T. (1993). Pluralistic ignorance and alcohol use on campus: Some consequences of misperceiving the social norm. Journal of Personality and Social Psychology, 64(2), 243–256. https://doi.org/10.1037/0022-3514.64.2.243
- Schultz, P. W., Nolan, J. M., Cialdini, R. B., Goldstein, N. J., & Griskevicius, V. (2007). The constructive, destructive, and reconstructive power of social norms. Psychological Science, 18(5), 429–434. https://doi.org/10.1111/j.1467-9280.2007.01917.x
- Sparkman, G., & Walton, G. M. (2017). Dynamic norms promote sustainable behavior, even if it is counternormative. Psychological Science, 28(11), 1663–1674. https://doi.org/10.1177/0956797617719950
- Tankard, M. E., & Paluck, E. L. (2016). Norm perception as a vehicle for social change. Social Issues and Policy Review, 10(1), 181–211. https://doi.org/10.1111/sipr.12022
Social Norms in Collective Health, Safety, and Community Action
Health-related practices often involve public signals and private risk. People may infer what is safe or acceptable from peer behavior even when those peers have incomplete information. A social-norm intervention that accurately communicates protective behavior can support collective action, but a descriptive message emphasizing widespread risky conduct may normalize the practice. The conditional logic follows Cialdini et al. (1990) and the intervention analysis of Miller & Prentice (2016). For any health or safety setting, the validity of the factual claim, audience vulnerability, and independent medical or public-health evidence must accompany the psychology of influence.
Community interventions can combine normative communication with material changes and social participation. Paluck & Green (2009) experimentally studied a radio program intended to change deference and dispute resolution in Rwanda; reported shifts in willingness to express dissent and communal problem-solving behavior were more evident than changes in many private attitudes. The findings point to a distinction between the perceived social permissibility of speaking or acting and a complete revision of individually held beliefs. They also show why intervention assessment needs outcomes beyond respondents' agreement with a message.
Measurement of success must address equity and unintended effects. Increased public compliance may reflect temporary surveillance, fear of sanction, or genuine changes in perceived legitimacy. Where sanctions are socially asymmetric, a program could increase the burden on marginalized people while improving an aggregate compliance rate. Legros & Cislaghi (2020) argues that harmful and beneficial norms need to be studied within the same general framework. The scientific purpose is to explain whose expectations govern action, who can deviate safely, and whether intervention benefits persist when monitoring or a campaign ends.