RESEARCH DOSSIER
Intergroup Conflict and Dehumanization
Abstract. Intergroup conflict develops through the interaction of perceived incompatible collective interests, group identification, threat appraisal, social representation, institutional incentives, and expectations about opponents. Dehumanization is a distinctive family of judgments that deny attributes of humanness or experience; its presence and consequences require direct measurement rather than inference from ordinary dislike. This fundamental research monograph traces the historical formation of realistic conflict, minimal-group and social-identity research, moral exclusion, infrahumanization, mind perception, and contemporary models of blatant and reciprocal dehumanization. A critical narrative synthesis examines laboratory experiments, meta-analyses, community samples during asymmetric warfare, a reconciliation radio field experiment, and research on intergroup contact and collective-blame interventions. The analysis separates group affiliation from hostility, perceived threat from objective competition, dehumanizing language from discriminatory action, and reciprocal psychological processes from symmetry of political power or responsibility. It evaluates the causal limits of cross-sectional correlations, laboratory paradigms, and intervention follow-up while specifying research questions about institutional accountability, durable behavior change, and digitally mediated group relations in the Artificial Era. The synthesis identifies multiple conditional pathways linking social conflict and moral exclusion and emphasizes evidence standards for evaluating prevention and reconciliation. This is a critical narrative monograph of published literature, not a registered systematic review, original experiment, or meta-analysis.
Keywords: intergroup conflict; dehumanization; social identity; realistic conflict; outgroup threat; moral exclusion; infrahumanization; blatant dehumanization; meta-dehumanization; reciprocal hostility; group-based inequality; intergroup contact; collective blame; peacebuilding; social psychology; Artificial Era
Scientific Object, Scope, and Questions
Intergroup conflict denotes structured opposition between social groups whose members understand a dispute in terms of group membership, collective interests, identity, status, territory, resources, rights, or moral standing. Dehumanization denotes an appraisal in which another person or group is denied some attributes associated with humanness. The joint scientific object of this dossier is the interaction between group-based antagonism and representations of opponents as less fully human, together with the institutional and situational conditions that intensify or restrain this interaction. The relationship is contingent: Brewer (1999) shows why attachment to an ingroup cannot be treated as equivalent to hatred of an outgroup, while Haslam (2006) shows that dehumanization also occurs outside open conflict.
The central questions are how perceived competition and threat shape group boundaries; how these boundaries become attached to stereotypes, diminished mind attribution, or denial of moral consideration; whether dehumanization predicts hostility beyond prejudice and ideology; how reciprocal expectations produce feedback; and which interventions change actual cooperation or harm rather than only stated attitudes. The unit of analysis varies from individual judgment and interaction to organizational communication and societies experiencing sustained conflict. Group asymmetries in power, exposure to violence, and political institutions remain properties of the situation that cannot be inferred merely from individual-scale psychological measures.
This is a critical narrative synthesis of published empirical and conceptual research. Historical laboratory experiments, validated questionnaires, community surveys, conflict-zone studies, intervention trials, and meta-analyses are compared by their inferential reach. The selection is thematic and evidence-oriented; it has no preregistered search protocol, exhaustive screening flow, quantitative pooled estimator, or new data collection. Research claims are attributed to accessible original publications or identifiable abstracts and bibliographic records. References to the Artificial Era appear only where mind attribution, classification, and technologically mediated communication generate empirically tractable questions.
Defining Conflict, Hostility, and Dehumanization
Conflict is a relation of incompatible or perceived incompatible claims. Its psychological indicators include felt threat, competitive allocation, distrust, negative emotional orientation, support for exclusionary policy, and willingness to impose costs on an outgroup. These indicators do not form a single interchangeable scale: a person may prefer the ingroup when distributing benefits while rejecting coercion or violence. The minimal-group findings of Tajfel et al. (1971) identify discrimination in allocation under thin categorization conditions; they do not by themselves demonstrate a disposition toward severe conflict. The distinction between ingroup preference and outgroup aggression is essential to causal analysis.
Dehumanization refers to a particular structure of social representation rather than a synonym for dislike. In Haslam (2006), denial of uniquely human attributes is associated with animalistic forms of dehumanization, while denial of human-nature attributes is associated with mechanistic forms. Kteily et al. (2015) investigates explicit, blatant ratings that place some groups lower on a pictorial scale of perceived evolutionary advancement. Infrahumanization studies, including Vaes et al. (2003), examine the differential attribution of emotions construed as uniquely human. These measurement families capture related but nonidentical judgments, so results should be evaluated at the level of their operational definition.
Moral exclusion is a further construct: Opotow (1990) defines the removal of persons or groups from the perceived domain of fairness and moral obligation. Dehumanization may support that exclusion, and moral exclusion may be expressed through policies without overt animalistic imagery. Kelman (1973) treats authorization and routinization as additional social processes relevant to large-scale harm. A rigorous study therefore records what actors believe about victims, which norms they invoke, what roles they occupy, and what behavior follows, instead of collapsing representation, approval, participation, and consequences into one variable.
Historical Development of the Field
Modern social-psychological conflict research took shape through field experiments, studies of prejudice, group categorization, and moral responsibility. Sherif et al. (1961) documented the experimentally arranged formation of groups at a summer camp, subsequent competitive relations, and cooperative interdependence introduced through shared problems. Its original report supplies the logic of changing intergroup relations across stages, while also revealing how profoundly context and experimenter organization shape the phenomenon. The study offers a historically influential demonstration, not an unrestricted estimate of the frequency of conflict in contemporary populations.
In Allport (1954), prejudice research integrated learned stereotypes, categorization, group context, and conditions favorable to contact. Tajfel et al. (1971) later showed that biased allocation could be produced under minimal classification. The resulting intellectual transition expanded explanation from prior disputes over material resources to the psychological significance of group distinctions themselves. Brewer (1999) subsequently emphasized that ingroup identification and outgroup hostility are not the same construct. Together these developments challenge simple accounts that locate all hostility in competition or, conversely, in category membership alone.
The study of extreme intergroup harm broadened this agenda. Kelman (1973) examined dehumanization alongside authorization and routinization; Opotow (1990) introduced moral exclusion as a way to analyze the shifting boundaries of justice; Bandura (1999) systematized mechanisms that disengage moral self-regulation. The empirical dehumanization literature grew through studies of emotion attribution and mind perception, followed by Haslam (2006), Haslam & Loughnan (2014), and direct measurement of blatant dehumanization in Kteily et al. (2015). This history records overlapping research traditions whose concepts must remain distinct when applied to current conflicts.
Competition, Interdependence, and Realistic Conflict
Realistic conflict accounts locate an important source of antagonism in the perceived incompatibility of collective goals. Scarce resources, unequal access, rival political claims, and institutional rules can make competition salient and tie individual advantage to group victory. In the staged Robbers Cave investigation, Sherif et al. (1961) described how competitive arrangements affected social relations, then how superordinate tasks altered cooperation. The mechanism is interdependence: when reward structures define the success of one side as the loss of another, rivalry is reinforced, whereas credible shared tasks can alter incentives and interaction patterns.
Perceived competition requires separate assessment from objective scarcity. A policy may intensify material inequality while survey respondents report symbolic fears; the reverse can also occur. Riek et al. (2006) synthesizes evidence on several threat constructs and negative outgroup attitudes, showing why realistic threat, symbolic threat, group-esteem threat, intergroup anxiety, and stereotypes should be measured separately. Inferences concerning resource competition should combine institutional conditions, distributional data, perceived stakes, and behavioral decisions. A survey association between threat and hostility does not establish whether threatening conditions caused hostility or hostility increased threat perception.
Cooperation is constrained by credibility. Shared problems may remain ineffective when parties expect exploitation, leaders punish compromise, or the benefits of cooperation are distributed unequally. Intergroup trust games and repeated resource-allocation paradigms can isolate expectations and responses to incentives, but artificial experimental payoffs only approximate historical struggles in which membership and security carry exceptional costs. The theoretical lesson is that realistic conflict and social categorization interact: categories identify parties, institutions determine payoffs, and communication influences expectations about whether cooperation is safe Sherif et al. (1961) Tajfel et al. (1971) Riek et al. (2006).
Threat Appraisal, Power, and Hierarchy
Intergroup threat is a family of appraisals about anticipated loss, disruption, or danger. A realistic threat concerns material welfare or security; a symbolic threat concerns norms, identity, or worldview; group esteem and distinctiveness threats concern the position and meaning of collective membership. Riek et al. (2006) examined 95 samples and reported positive relations between five studied threat categories and negative outgroup attitudes, with some moderation by outgroup status. As a meta-analysis of correlational relations drawn from varied designs, the synthesis supports systematic covariation more securely than a common causal direction.
Hierarchy alters these appraisals. Pratto et al. (1994) introduced social dominance orientation as preference for group-based inequality and related it to attitudes about group relations and policies. Such ideological preferences may shape which disparities are considered legitimate and whose suffering merits attention. Their association with dehumanization and exclusion, however, is not a license to infer individual motives from group membership or to flatten differences among institutions. Power is relational and can affect opportunity, resources, security, and the visibility of harms irrespective of respondents' self-reported attitudes.
A stronger multilevel account tracks the joint operation of perceived threat, structural advantage, access to coercive institutions, and the narrative of deservedness. In asymmetric conflicts, measures of hostility among two populations can be superficially comparable while their material exposure and capacity to inflict harm differ substantially. The findings of Bruneau & Kteily (2017) make the importance of measuring each side visible, while the study's design cannot itself establish symmetry of the underlying conflict conditions. Description of reciprocal psychological processes must be combined with an independent assessment of power and action.
Animalistic, Mechanistic, and Blatant Dehumanization
The theoretical differentiation between animalistic and mechanistic dehumanization responds to an ambiguity in the claim that someone is regarded as less human. Haslam (2006) distinguishes attributes often understood as uniquely human, such as refinement and rationality, from human-nature attributes, including emotional depth and warmth. Denial of the former suggests an animal-like target; denial of the latter suggests a cold, inert, or mechanical target. An empirical instrument built for one denial should not be relabeled as evidence for another. The model extends to interpersonal, occupational, and institutional relations, although its application to active group conflict requires attention to enemy identity and group norms.
Blatant dehumanization typically uses more direct ranking or explicit language about the humanity of an outgroup. The scale developed and validated by Kteily et al. (2015) was designed to test whether overt dehumanization explains judgments and outcomes beyond subtle indicators and generalized prejudice. It is important to distinguish the historical imagery used as a measurement stimulus from biological evolutionary science: the pictorial ranking measures a social attitude, not an objective hierarchy among human populations. Its interpretive value lies in respondents' willingness to express a degrading characterization of a social group.
Explicit and subtle measures can show different responses to political events, audience expectations, and social desirability. A participant might avoid blatant statements while still selectively denying complex emotions; another might endorse a harsh public portrayal without producing the specific attribution pattern captured by an infrahumanization task. Haslam & Loughnan (2014) reviews this diversity. Convergence across self-report, linguistic analysis, behavioral outcomes, and external observations provides a more credible account of dehumanization than any instrument considered alone.
Infrahumanization and the Attribution of Emotions
Infrahumanization research examines the differential assignment of emotions treated by participants as distinctively human. The hypothesis is that people can regard fellow group members as capable of complex emotions to a greater extent than members of another group. Vaes et al. (2003) tested consequences through four experiments, including a lost-email task and conformity paradigms, studying reactions to expressions of uniquely human emotion. The strength of this research design lies in manipulating experimentally relevant information and recording behavior or choices within defined tasks.
The emotion categories used to operationalize uniquely human experience are culturally and theoretically situated. They should not be assumed to represent universally agreed natural kinds. Differences in language, knowledge of targets, and attribution norms can confound measures if investigators fail to control them. Infrahumanization can also arise in familiar, relatively peaceful groups, making it unsuitable as a direct diagnostic test for violent intent. The construct contributes to conflict analysis by identifying a mechanism through which an outgroup may appear less psychologically complex while remaining explicitly acknowledged as human Vaes et al. (2003) Haslam & Loughnan (2014).
A complete measurement program would pair attribution tasks with baseline emotional knowledge, perceived similarity, group identification, and behavioral outcomes. If a group is rated as less capable of remorse or grief, the researcher must establish whether this reflects denial of human emotional capacity, low familiarity, a specific stereotype, or an inference from reported behavior. Experiments can test controlled contrasts; observational evidence can determine whether the same pattern occurs during real conflict. The inference from attributed emotion to willingness to harm requires its own demonstration rather than automatic extension.
Mind Perception, Agency, Experience, and Moral Standing
Mind perception concerns the tendency to ascribe capacities such as thought, planning, feeling, and subjective experience to a target. The review by Waytz et al. (2010) explains why such attributions vary with perceiver motives and perceived similarity, and why they are consequential for moral evaluation. In intergroup settings, diminished attribution of suffering may undermine concern for victims, whereas attributed strategic agency can intensify fears of intentional danger. Thus conflict propaganda might represent an opponent simultaneously as an emotionally insensible object and a highly calculating adversary; the dimensions of mind attribution need not move in a uniform direction.
Behavioral consequences depend on the attribute denied. Diminished emotional experience may reduce spontaneous empathy; exaggerated hostile agency may magnify anticipated threat. Haslam (2006) supports the conceptual importance of separating human-nature and uniqueness dimensions, while Waytz et al. (2010) provides a broader account of why mind perception is connected to moral status. Neither review establishes one universal causal route from reduced mind attribution to aggression. The direction and strength of a pathway must be assessed with appropriate tasks and actual conduct.
Emerging artificial systems complicate conventional language about mind attribution. People can attribute agency and experience to nonbiological systems, yet such attributions are properties of a perceiver's model until relevant system capacities are independently demonstrated. The interaction of humans and artificial agents may alter which cues produce trust or exclusion, but evidence about human intergroup dehumanization cannot be transferred wholesale to artificial cognition. A responsible research program operationalizes the target's behavior, the observer's attributions, and consequences for moral decision-making separately Waytz et al. (2010).
Moral Exclusion, Responsibility, and Institutional Authorization
Exclusion from the scope of justice is a distinct route through which group conflict can become morally asymmetric. Opotow (1990) analyzes the boundary within which fairness and concern apply. When an outgroup falls outside that boundary, restrictive policies, exploitation, or indifference toward harm may be experienced as acceptable rather than as a violation of professed principles. The key measurement issue is whether perceived rights and obligations actually differ by target group under comparable circumstances. An explicit degrading label is one indicator; an observable refusal of procedural fairness is another.
The framework of Kelman (1973) adds organizational processes that cannot be reduced to personal attitudes. Authorization channels acts through accepted authority; routinization divides and normalizes procedures; dehumanization undermines the status of those affected. This tripartite explanation suggests that violent conduct may be facilitated even when participants do not repeatedly articulate overt hatred. Institutional role expectations, paperwork, command structures, and ambiguous accountability can transform choices. Historical applications require archival and organizational evidence, because the psychological theory alone cannot establish what particular actors knew or intended.
According to Bandura (1999), moral disengagement includes moral justification, euphemistic labeling, advantageous comparison, displacement and diffusion of responsibility, minimization of harm, attribution of blame, and dehumanization. The processes can reinforce one another: an organization may frame harmful action as protective, call it by a sanitized term, and distribute responsibility through multiple offices. Yet the model predicts mechanisms rather than furnishing retrospective proof of motive. Empirical interpretation benefits from contemporaneous records, decisions, and statements compared against alternative institutional explanations.
Enemy Images, Collective Blame, and Group Essentialization
Enemy images simplify a heterogeneous population into an apparently coherent, threatening agent. The cognitive move substitutes collective character for observed individual behavior, often inferring intentions from group labels. It can make an isolated act appear diagnostic of every member and can justify retaliation against people who had no involvement. Bruneau et al. (2018) examined collective blame in relation to anti-Muslim hostility and tested an intervention drawing attention to inconsistent standards of blame. The research illustrates how correcting responsibility attribution may reduce hostility without requiring a complete revision of group identification.
Essentialized group representation can imply that membership reflects a fixed nature and makes evidence of variation seem exceptional. These ideas interact with threat appraisals and moral exclusion, but they should be measured directly instead of inferred from unfavorable opinions alone. A respondent may be critical of a government or institution without making an essentialist claim about an entire population. Valid coding separates criticism of actions, predictions about organized agents, individual stereotypes, collective blame, and denial of basic human capacity.
Escalation occurs when threat and collective attribution are reciprocally updated. A harmful act becomes evidence that the outgroup is dangerous; a degrading response becomes evidence of disrespect; both sides interpret their own retaliation as defense. Kteily et al. (2016) investigates the role of perceived dehumanization by others in predicting reciprocal dehumanization and conflict-related outcomes. Such pathways identify meaningful targets for intervention, but observational or mediated analyses need careful temporal ordering to distinguish reaction from selection, rhetoric, and the wider political situation.
Individual Differences, Ideology, Disgust, and Context
People vary in endorsement of hierarchical group relations, threat sensitivity, disgust sensitivity, prejudice, and willingness to express degrading judgments. Pratto et al. (1994) introduced social dominance orientation as a preference for group-based inequality. Hodson & Costello (2007) examined interpersonal disgust and found its relation with negative outgroup attitudes was modeled through ideological orientations and dehumanizing perceptions. These studies provide evidence about individual differences and covariation, but do not make any single disposition sufficient to predict aggression in a particular setting.
Context can amplify or suppress the expression of a predisposition. Organizational incentives, media cues, norm violations, elite rhetoric, and the perceived legitimacy of authority can make degrading descriptions common or socially costly. An individual may change overt responses rapidly when group norms change while deeper appraisals remain more stable. Bandura (1999) suggests mechanisms through which harmful practices are morally reframed; Paluck (2009) supplies field evidence that perceived norms and behavior can shift even when private beliefs change little.
Research should resist essentializing the very groups it studies. It is scientifically invalid to attribute a population's average survey response to every member or to infer an inborn tendency toward violence from a national, ethnic, or religious label. Generalization requires samples that reflect relevant variation and instruments that retain their meaning across languages and contexts. Disgust and ideological scales may measure correlated but different constructs, while social structure strongly affects which respondents can translate attitudes into material harms Hodson & Costello (2007) Riek et al. (2006).
Laboratory Paradigms and Construct Validity
Laboratory investigation offers the ability to manipulate categories, descriptions, incentives, and threat cues while holding specified alternatives constant. Minimal-group allocation designs show how classifications influence distributional decisions Tajfel et al. (1971). Emotion-attribution experiments investigate whether a sender's group membership alters responses to emotionally meaningful messages Vaes et al. (2003). Such paradigms differ in their outcome variables: a token allocation, a reply to a stranger, and an explicit judgment of humanity are not equivalent evidence for real-world violence. Interpretations depend on what was randomized and what was actually observed.
Design strength requires attention to blinding, credible manipulation checks, measurement reliability, preselection, demand characteristics, and the possibility that respondents infer the experimental purpose. A dehumanizing label embedded in a vignette may simultaneously convey danger, contempt, unfamiliarity, and social status. Factorial experiments can disentangle some of these features, provided manipulation validity is demonstrated independently. When a scale contains moralized language, a correlation with punishment attitudes may partly reflect shared evaluative wording rather than a distinct psychological pathway.
External validity remains the central test. The consequences of expressing a preference in a university study are different from those of acting amid displaced populations, resource scarcity, coercive institutions, and irreversible risks. Laboratory results can establish restricted causal contrasts; field studies and historical records examine whether the contrast matters in social systems. A coherent evidence assessment needs both, with claims expressed at the level supported by each design Sherif et al. (1961) Kteily et al. (2015) Paluck (2009).
Measuring Blatant and Subtle Dehumanization
Operationalization determines what a study actually learns about dehumanization. Blatant scales ask for direct evaluations of perceived humanity; human uniqueness and human nature inventories assess attribute denial; infrahumanization designs compare secondary-emotion assignments; implicit paradigms study relative associations between groups and human or animal concepts. Kteily et al. (2015) directly compared blatant with subtler measures to evaluate incremental explanatory power. Its results justify analytic differentiation among measures, not their uncritical substitution for one another.
Reliability matters within and across groups. Internal consistency alone does not demonstrate that items have a common meaning across languages or historical settings. Measurement invariance, response-style differences, local semantic conventions, and familiarity with the targets deserve explicit tests. A scale drawing on evolutionary imagery can carry different connotations depending on education and cultural context. It should be reported as a participant rating under a particular stimulus, without insinuating that the scale expresses a factual biological classification.
Behavioral validation compares reported dehumanization with independent measures such as discriminatory allocation, exclusionary choices, punitive endorsement, avoidance, willingness to engage, and direct assistance. These outcomes are affected by legal norms and social desirability. Bruneau & Kteily (2017) linked blatant dehumanization with conflict-related hostility in two community samples during wartime, making the association relevant beyond campus populations. Even then, attitude measures remain imperfect proxies for actions, and unmeasured exposure to violence can influence both.
Conflict-Zone Evidence and Reciprocal Dehumanization
The conflict-zone comparison reported by Bruneau & Kteily (2017) assessed Israelis and Palestinians during the 2014 Gaza war using respective community samples of 521 and 354 people. Blatant dehumanization appeared in both samples and was associated with hostility-related outcomes after considering certain ideological covariates. The evidence is important because it challenges an assumption that only advantaged actors can express dehumanization. It also illustrates that sample composition, historical exposure, relative power, and the available outcome measures require separate interpretation.
Associations in a cross-sectional wartime setting are embedded in current events and changing information environments. Respondents may describe opponents as inhuman because they expect violence, because they have witnessed harm, because they endorse exclusionary policy, or because common political rhetoric connects all three. Regression adjustments for ideology cannot eliminate all these alternatives. The data indicate a reliable observed relationship in the measured samples, while the causal ordering among experience, expression, and support for aggression remains to be tested through longitudinal or experimental methods.
Reciprocity is a process-level concept, not a claim that groups carry equal responsibility or capacity to inflict damage. A study can observe reciprocal contempt while independent evidence documents unequal institutions or casualties. Scientific analysis should retain both levels. Kteily et al. (2016) addresses how perceiving that an outgroup views one's group as less human may predict corresponding dehumanization, supporting research into feedback loops. Strong tests would measure both groups repeatedly, identify external events, and record changes in action.
Meta-Dehumanization and Dynamic Escalation
Meta-dehumanization means believing that another group denies one's own group's humanity. Kteily et al. (2016) tested whether these beliefs predict antagonistic responses through reciprocal dehumanization. The construct is especially relevant because conflict participants act on their representations of the other's attitudes, which may be exaggerated or erroneous. A population might retaliate against perceived collective contempt even when the beliefs attributed to the other population are heterogeneous. Researchers therefore need direct reports from both parties rather than relying exclusively on forecasts made by one side.
A dynamic causal account distinguishes initial derogation, perceived derogation, emotional response, retaliatory expression, and behavioral support for coercion. Temporal separation can reveal whether perceptions precede retaliatory attitudes, whereas simultaneous questionnaires cannot resolve feedback direction. Cross-lagged panels, repeated experience sampling, independent media-event coding, and pre-registered interventions can narrow the possibilities. Nevertheless, events affect both communities and may change perceptions in parallel, so even longitudinal results do not automatically identify a closed interpersonal feedback mechanism.
The relational perspective suggests a more precise intervention target than generic education about similarity. Correcting false perceptions about an outgroup's judgments might reduce retaliatory motivation in specific settings; where perceptions are accurate, institutional measures to change ongoing disrespect may be necessary. The evidentiary standard should focus on whether corrective information changes credible perceptions, emotional reactions, and consequential behavior. Survey changes alone can indicate progress in appraisal but cannot substitute for a reduction in injury or coercion Kteily et al. (2016) Bruneau et al. (2018).
Empathy, Counter-Empathy, and Social Neuroscience
Empathy is multidimensional: recognizing another's emotion, sharing aspects of affect, and caring about the other's welfare are analytically distinct. Cikara et al. (2014) examines failures of empathy and counter-empathic responses under intergroup competition, including contexts in which an opponent's misfortune elicits positive affect. These phenomena challenge any assumption that knowledge of another's subjective state always produces concern. The meaning of a response depends on identification, rivalry, moral appraisal, and task demands, which should be measured separately.
Neural findings can contribute to explanation when they test clearly specified mechanisms such as group identification, salience, or valuation. Cikara & Van Bavel (2014) integrates group research and neuroscience while emphasizing the contextual nature of social identity. Activation in a region associated with social cognition cannot by itself establish an attitude of dehumanization, nor does a reduced response prove moral indifference. Convergent behavioral indicators and validated interpretations of neural measures are required before drawing conclusions about everyday hostility.
The most informative designs combine manipulated intergroup context, self-report, neural or physiological measurement, and actual decisions. A participant may cognitively represent another person's pain while responding with reduced compassion because rivalry changes the perceived value of the outcome. Such cases involve motivational regulation as well as social perception. A causal model should consequently distinguish recognition of suffering, emotional resonance, and prosocial action rather than treating a single neural signal as a direct index of humane treatment Cikara et al. (2014) Waytz et al. (2010).
Levels of Evidence and Causal Identification
Evidence accumulates across designs rather than through a single preferred metric. Conceptual reviews establish theoretical distinctions; laboratory randomization isolates proximal effects; representative surveys estimate the distribution of reported attitudes; longitudinal panels reveal temporal association; natural and field experiments test context-sensitive interventions; historical research reconstructs institutionally embedded processes. Haslam & Loughnan (2014) organizes the conceptual field, whereas Bruneau & Kteily (2017) and Paluck (2009) offer different forms of context-rich empirical evidence.
Each design has failure modes. Convenience samples can distort prevalence claims; cross-sectional mediation can be mistaken for causal explanation; attitudes can be substituted for consequential action; and measured change may not persist after intervention. Publication bias and researcher flexibility complicate any literature summarized from selected positive findings. Meta-analytic evidence about contact and threat Pettigrew & Tropp (2006) Riek et al. (2006) should be read alongside later scrutiny of rigorous designs and delayed outcomes Paluck et al. (2019). No pooled magnitude can be transported across distinct constructs without examining sampling and measurement comparability.
The dossier therefore distinguishes four classes of conclusion. Conceptual conclusions describe the meaning of constructs; descriptive findings document what was measured; causal claims are restricted to the randomized or otherwise identified contrast; and theoretical syntheses propose mechanisms compatible with multiple pieces of evidence. Statements about institutional violence require additional historical and organizational proof. Research standards improve when studies preregister primary outcomes, report uncertainty, share usable materials, and examine whether effects appear in multiple populations.
Intergroup Contact and the Reduction of Prejudice
Intergroup contact is one of the most studied strategies for changing relations across social boundaries. The theoretical tradition associated with Allport (1954) identifies conditions in which meaningful interaction can diminish stereotyped expectations. Pettigrew & Tropp (2006) combined 713 independent samples from 515 studies and reported that intergroup contact typically correlated with reduced prejudice across multiple group targets and settings. The review also found stronger outcomes under favorable arrangements, while emphasizing that its classic set of optimal conditions was better understood as an interrelated bundle than as several individually necessary conditions.
Evidence quality differs with design and outcome. Paluck et al. (2019) reassessed the field using studies involving random assignment and delayed measures, assembling 27 eligible studies. Their analysis is consistent with a broad prejudice-reduction effect while exposing gaps in evidence for long-term change and for particular populations and policy settings. A short-term shift in willingness to interact should not be treated as proof of reduced dehumanization, durable safety, or reconciliation. Contact may be ineffective or risky under ongoing coercion, unequal status, or poorly designed environments.
Programs should measure the actual interaction: cooperation quality, voluntary participation, institutional support, sustained opportunity, and the distribution of costs. Meaningful change may require repeated nonthreatening encounters and credible systems for addressing harm. It is also crucial to monitor who withdraws from participation and why; apparent improvements among remaining participants can reflect selective retention. The contact literature supplies useful general principles, but translating them into conflict-specific programs requires attention to safety, context, and behavioral outcomes Pettigrew & Tropp (2006) Paluck et al. (2019).
Superordinate Goals, Cooperation, and Trust Repair
Cooperative interdependence can change perceptions by creating tasks in which the parties' success depends on shared action. The Sherif et al. (1961) account of Robbers Cave is influential because the researchers manipulated the nature of group relations across stages, moving from competition toward jointly managed problems. A superordinate goal can change the immediate incentive to treat an opponent as a rival, but a shared label alone is insufficient if resources, recognition, and benefits remain distributed unfairly. Studies should specify the nature of the interdependence instead of calling any joint activity cooperative.
Trust repair requires predictable behavior. An intergroup counterpart may seem more humane after fair treatment because the conduct supplies information about intentions and moral regard. Yet the same event may be interpreted as strategic deception when the context is hostile. Experimental trust games can test choices under controlled risk, whereas public policy needs evidence of repeated compliance and transparent oversight. The relevant outcome is an increasingly credible capacity to cooperate, including the possibility of refusing unsafe arrangements, rather than an expectation of unconditional reconciliation.
Institutional change can complement interpersonal experience. Transparent procedures, accountability for breaches, reliable channels for complaint, and opportunities for equal participation can create stable incentives for recognition. Opotow (1990) provides the conceptual importance of including all parties within the moral community, while Kelman (1973) underscores the organizational conditions that shape responsibility. The most persuasive account of trust repair integrates psychological appraisal with verifiable behavior and governance, because favorable impressions in an unaltered coercive structure may remain fragile.
Communication, Social Norms, and Media Interventions
Communication affects conflict partly by changing expectations about what peers approve and what behaviors are socially possible. The yearlong randomized radio field experiment in Rwanda conducted by Paluck (2009) compared a reconciliation-themed drama with a health-themed control drama. The intervention changed reported perceptions of norms and several behaviors related to trust, cooperation, and open disagreement, while producing comparatively little change in personal beliefs. This finding is especially important because public behavior and private attitudes need not respond together to an intervention.
Media effects depend on exposure, credibility, participation, and local interpretation. A message that supplies counter-stereotypic information may fail when it contradicts ongoing experience or is delivered by an untrusted source. A narrative can also shift perceived norms without prompting people to endorse the message in a direct survey. Consequently, intervention assessment should separately evaluate audience reach, understanding, social discussion, perceived normative change, and behavioral response. Evidence of changed audience engagement alone does not establish changed intergroup relations.
Digital communication introduces new channels for repeating enemy images, collapsing individual behavior into collective blame, and amplifying perceived humiliation. The historically documented mechanisms of norm learning and collective attribution provide testable hypotheses, but platform-specific claims require platform-specific data. A comparative study could randomize exposure to corrective context, track perceived collective responsibility, and measure decisions about cooperation or exclusion. The Rwanda experiment supplies a precedent for field-based evaluation, not a direct estimate of algorithmic effects Paluck (2009) Bruneau et al. (2018).
Collective Blame and Targeted Interventions
Collective blame is the attribution of the actions of a few members to a larger social category. Such attribution can generate demands for punitive measures against people unrelated to the initiating act. Bruneau et al. (2018) investigated collective blame in the context of anti-Muslim hostility and compared interventions, identifying a successful hypocrisy-focused approach among several candidates. The intervention logic makes asymmetric standards of accountability salient, thereby challenging the inference from an individual perpetrator to a whole group. Its evaluation addresses specific attitudes and measured behaviors rather than establishing a universal method for ending conflict.
Interpretation requires examining what an intervention directly modifies. A hypocrisy prompt may weaken an automatic group-generalization judgment while leaving resource anxiety or institutional threat unchanged. The effects may also depend on local norms about fairness and the credibility of the comparison. Repeating the same intervention across different communities, using independent research teams and delayed behavioral endpoints, would test whether the relevant mechanism generalizes. An intervention that succeeds on one attitude scale may fail on more consequential outcomes.
Targeted communication should avoid treating the communities under study as passive recipients of psychological correction. Participants are situated agents whose assessments can reflect actual experiences of harm. Programs should therefore combine information about attribution errors with accurate accounts of context and meaningful procedural protections. From an evidence standpoint, the scientific question is whether a designed change in responsibility judgment leads to measurable reductions in unjust exclusion, threats, or retaliatory behavior Bruneau et al. (2018) Opotow (1990).
Preventing Escalation and Protecting Moral Inclusion
Prevention requires identifying points in the conflict process where enemy representation becomes linked to permission for harm. Bandura (1999) describes moral disengagement through justification, sanitized language, diffused responsibility, denial of consequences, and victim derogation. Organizational safeguards can be designed around these routes: transparent reporting of consequences, refusal to anonymize responsibility, meaningful review of coercive acts, and explicit rules applying equally to affected populations. These proposals follow the mechanisms of the theory; their effectiveness must be established in implementation research.
Moral inclusion can be operationalized as willingness to apply comparable standards of dignity, procedural fairness, and concern regardless of group membership. Opotow (1990) indicates why the scope of justice is an important target. Researchers can test whether respondents grant equal evidentiary standards, due process, assistance, and rights to members of different categories under matched situations. Such indicators may be more informative about humane institutional treatment than changes in generalized warmth alone.
An important asymmetry concerns the burden of prevention. Where institutions control detention, force, migration, or access to essential services, their decisions can produce large harms even when general survey attitudes remain unchanged. Individual-level programs cannot replace rights protection and accountable administration. Psychology contributes by explaining mechanisms of compliance, belief, threat, and moral evaluation; an adequate prevention strategy also requires legal, historical, and organizational knowledge Kelman (1973) Bandura (1999).
Evaluating Conflict Reduction and Long-Term Change
The evaluation of a peacebuilding or prejudice-reduction program begins by specifying the outcome. Dehumanizing speech, perceived threat, reciprocal hostility, intergroup contact, discriminatory choices, and actual violence are related but distinct. A strong study establishes a baseline, comparable conditions, validated instruments, ethical procedures, and a follow-up long enough to distinguish temporary demand effects from durable change. The evidence reviewed by Paluck et al. (2019) shows why randomization and delayed measures are valuable when deciding whether contact interventions justify policy claims.
Process evaluation asks whether participants received the intended program, whether implementation differed across settings, and whether mechanisms responded as hypothesized. If an intervention changes norms Paluck (2009), the investigator should measure both norm perceptions and target behavior. If it is designed to reduce collective blame Bruneau et al. (2018), a critical test is whether altered attribution mediates a downstream change under appropriate causal identification. Attrition and contamination among connected participants can obscure comparisons, particularly in conflict-sensitive communities.
The standard of success must include the possibility of adverse effects. Forced contact can increase vulnerability; public displays of reconciliation can stigmatize dissenters; and measurement of dehumanization can itself expose participants to harmful language. Ethical research incorporates voluntary participation, privacy, safety planning, locally informed interpretation, and community feedback. Longer-term field designs should report unintended changes, heterogeneity, and failures alongside favorable averages. The objective is a defensible account of when and for whom an intervention changes intergroup life.
Developmental, Cultural, and Historical Variation
Intergroup categories acquire meaning within developmental histories of family socialization, schooling, migration, political institutions, and encounters with groups designated as outsiders. The mechanisms discussed in this dossier are not limited to a particular national setting, but their expression depends on locally meaningful identities, historical grievances, authority relationships, and the acceptable language of disagreement. Allport (1954) and Brewer (1999) provide foundational analyses of how group boundaries shape evaluations. These accounts offer conceptual starting points; direct developmental inference requires studies following participants across age or comparing well-specified cohorts.
Cultural variation matters to the meaning of emotion vocabulary, moral obligations, group identification, and claims about humanity. An infrahumanization measure designed for one language may not have equivalent validity in another. Likewise, explicit statements of hierarchy may be constrained by local reputational costs, while indirect exclusion can remain institutionally commonplace. Vaes et al. (2003) and Kteily et al. (2015) show that different instruments capture different operations, making cross-cultural calibration essential. Measurement invariance and local qualitative interpretation help distinguish construct change from linguistic response differences.
Historical comparison must combine psychological claims with independent documentation of material and political structures. The presence of degrading propaganda does not tell us by itself how decisions were made, who possessed coercive authority, or how violence unfolded. Kelman (1973) identifies mechanisms relevant to organized harm but cannot replace historians' archival reconstruction. Research across societies should therefore treat context as a potential cause and moderator, not as decorative background to a universal individual-level story.
Asymmetry, Justice, and Ethical Interpretation
A critical account of group conflict distinguishes reciprocal psychological processes from the distribution of rights, power, losses, and responsibility. Bruneau & Kteily (2017) observed dehumanizing responses in populations on different sides of an asymmetric war. The finding establishes that degrading judgments are possible in multiple positions; it does not imply an equivalence in material circumstances, legal status, exposure, or conduct. Scientific descriptions of attitudes must be accompanied by independently sourced contextual evidence when a specific conflict is under analysis.
The language of symmetry can obscure important variation if it is applied to entire communities. Surveys estimate responses among sampled participants under particular historical conditions. A conclusion about one sample cannot characterize all people associated with that group, and the measured association between hostility and policy preference cannot be used to infer the intent of leaders or armed institutions. Similarly, valid criticism of state policy must be distinguished from generalized hostility toward people on the basis of origin, religion, or ethnicity.
Ethical interpretation includes methodological responsibility to participants. Discussing explicit dehumanizing statements should preserve attribution and avoid presenting them as the narrator's judgments. Investigators must protect confidentiality in settings where expressed views can expose respondents to retaliation. Results should describe consequences with precision while recognizing that institutional protection of rights is not reducible to the psychological task of improving attitudes Opotow (1990) Kelman (1973).
Computational Communication and the Artificial Era
Networked and artificial systems now participate in the production, distribution, ranking, and transformation of intergroup messages. Psychological research can examine how recommendation environments, synthetic messages, and human–artificial dialogue alter exposure to enemy images, perceived public opinion, and confidence about collective intentions. The underlying constructs remain recognizable: threat appraisal, identity salience, collective blame, mind attribution, and norm perception Riek et al. (2006) Waytz et al. (2010). Their causal operation within a particular platform, however, is a fresh empirical question that requires data about the platform and its users.
Artificial agents can be studied as communication partners, classifiers, or sources of social information. The fact that people ascribe intentions or emotional understanding to an agent is an observation about mind perception; it does not resolve the ontology or phenomenal experience of that system. Research on human dehumanization must not be mapped mechanically onto artificial entities, whose capacities and social positions require their own operational definitions. In the Artificial Era, the productive issue is how humans interpret intelligence, agency, and moral standing across biological and nonbiological interactions Waytz et al. (2010).
An intervention-relevant design could compare transparent source attribution, contextual information about a depicted outgroup, and explanations correcting collective blame. Researchers would then measure human participants' threat estimates, moral judgments, perceived norms, and actual sharing or exclusionary choices. Such work should assess algorithmic affordances, differential exposure, misinformation, and demographic heterogeneity without assuming that technological mediation necessarily escalates or reduces conflict. This extension is a proposed research program grounded in established psychological constructs, not an empirical result established by the classical literature.
Integrated Explanatory Synthesis
The assembled literature supports a multilevel account with linked components. Collective arrangements establish perceived interests and opportunities; categorization determines which people and actions are interpreted as belonging together; threat appraisal and identity concern organize expected danger; stereotyped representations and dehumanization shape the presumed capacities and value of opponents; moral exclusion and disengagement influence which acts appear permissible; and behavioral responses update group expectations. Evidence for particular links comes from research on interdependence Sherif et al. (1961), categorization Tajfel et al. (1971), threat Riek et al. (2006), humanity judgments Haslam (2006) Kteily et al. (2015), and moral regulation Opotow (1990) Bandura (1999).
This synthesis proposes a sequence of testable links while recognizing that causal order can vary. Hostility may precede dehumanization; dehumanization may normalize later hostility; both may be responses to observed harm. Kteily et al. (2016) highlights reciprocal processes, while Bruneau & Kteily (2017) demonstrates how strong associations appear within active conflict. Experimental evidence supports specific proximal manipulations, whereas sustained societal causal sequences require longitudinal and institutional study. The presence of several processes at one moment is insufficient to prove that every step of the proposed sequence occurred.
The model integrates incentives, representations, norms, and institutions rather than selecting a single master cause. Intergroup conflict can persist amid extensive recognition of the opponent's humanity when stakes remain irreconcilable; dehumanization can emerge in relatively ordinary settings without organized conflict Haslam (2006). A scientific account must therefore make room for conflict without contempt, contempt without violence, and humane behavior enacted despite continued disagreement. Such distinctions clarify the conditions under which the mechanisms become aligned and harmful.
Unresolved Problems and Future Research
One unresolved question is the incremental causal contribution of dehumanization beyond threat, prejudice, ideology, prior victimization, and perceived intentions. Studies often establish predictive associations but lack designs that manipulate the specific humanity appraisal independently of other evaluative content. Future research should compare multiple validated measures of dehumanization in the same populations, document their distinct outcomes, and preregister the hypotheses about mediation. Trials that shift one measure while leaving another unchanged would help specify whether the constructs differ mechanistically or only in their verbal expression Kteily et al. (2015) Haslam & Loughnan (2014).
Another problem concerns temporal dynamics. Cross-sectional reciprocal hostility measures do not identify whether violence drives dehumanization, the reverse, or both through common events. Repeated observations before and after informative events, independent coding of harmful incidents, and multi-group longitudinal panels could improve identification. Sampling groups across varied power positions would illuminate when similar psychological responses coexist with unequal capacities for action. Comparative research must publish both common trends and site-specific exceptions Bruneau & Kteily (2017) Kteily et al. (2016).
Intervention science also needs evidence about durability, safety, and consequential behavior. Contact meta-analyses establish broad patterns, yet the rigorous delayed-outcome literature is smaller Pettigrew & Tropp (2006) Paluck et al. (2019). Field experiments on norms Paluck (2009) and collective blame Bruneau et al. (2018) suggest distinct intervention pathways deserving independent replication. Future work should measure institutional outcomes, adverse effects, heterogeneity, and how networked communication interacts with local histories. The central scientific task is to distinguish changing narratives from changing relationships and changing social conditions.
Conclusion
Intergroup conflict and dehumanization are linked by identifiable psychological mechanisms operating within material, cultural, and institutional environments. Competition and identity create conditions for collective antagonism; threat appraisals and enemy representations shape what opponents are believed to intend; selective denial of humanness can narrow empathy and moral inclusion; and group norms, authority, and organizational routines may transform judgments into patterned action. Theoretical integration is strongest when these mechanisms are operationalized separately and assessed across levels of explanation Sherif et al. (1961) Haslam (2006) Kelman (1973).
The empirical record establishes that categorization can influence allocation decisions, that multiple forms of dehumanizing judgment can be measured, that such judgments correlate with hostility in conflict settings, and that designed interventions can change some attitudes, norms, and behaviors. The most difficult questions concern causal direction, long-term behavioral consequences, unequal institutional power, and the portability of interventions across societies Tajfel et al. (1971) Bruneau & Kteily (2017) Paluck (2009) Paluck et al. (2019). Scientific conclusions should therefore preserve the distinction between a demonstrated contrast, a predictive association, and an explanatory proposal.
The research program culminates in a practical criterion for better knowledge: each account of conflict should identify the groups and institutions involved, specify perceived and material interests, measure humanity judgments and moral boundaries, test behavioral outcomes, and examine consequences over time. This approach makes interdisciplinary cooperation possible among social psychology, conflict studies, cultural research, law, and research on technologically mediated social relations. It also gives a principled basis for evaluating efforts to reduce unjust exclusion while protecting people whose lives are affected by the conflict.
References
- Allport, G. W. (1954). The nature of prejudice. Addison-Wesley. https://wellcomecollection.org/works/rk5gw3qa
- Bandura, A. (1999). Moral disengagement in the perpetration of inhumanities. Personality and Social Psychology Review, 3(3), 193–209. https://doi.org/10.1207/s15327957pspr0303_3
- Brewer, M. B. (1999). The psychology of prejudice: Ingroup love and outgroup hate? Journal of Social Issues, 55(3), 429–444. https://doi.org/10.1111/0022-4537.00126
- Bruneau, E., Kteily, N., & Falk, E. (2018). Interventions highlighting hypocrisy reduce collective blame of Muslims for individual acts of violence and assuage anti-Muslim hostility. Personality and Social Psychology Bulletin, 44(3), 430–448. https://doi.org/10.1177/0146167217744197
- Bruneau, E., & Kteily, N. (2017). The enemy as animal: Symmetric dehumanization during asymmetric warfare. PLOS ONE, 12(7), e0181422. https://doi.org/10.1371/journal.pone.0181422
- Cikara, M., Bruneau, E., Van Bavel, J. J., & Saxe, R. (2014). Their pain gives us pleasure: How intergroup dynamics shape empathic failures and counter-empathic responses. Journal of Experimental Social Psychology, 55, 110–125. https://doi.org/10.1016/j.jesp.2014.06.007
- Cikara, M., & Van Bavel, J. J. (2014). The neuroscience of intergroup relations: An integrative review. Perspectives on Psychological Science, 9(3), 245–274. https://doi.org/10.1177/1745691614527464
- Haslam, N. (2006). Dehumanization: An integrative review. Personality and Social Psychology Review, 10(3), 252–264. https://doi.org/10.1207/s15327957pspr1003_4
- Haslam, N., & Loughnan, S. (2014). Dehumanization and infrahumanization. Annual Review of Psychology, 65, 399–423. https://doi.org/10.1146/annurev-psych-010213-115045
- Hodson, G., & Costello, K. (2007). Interpersonal disgust, ideological orientations, and dehumanization as predictors of intergroup attitudes. Psychological Science, 18(8), 691–698. https://doi.org/10.1111/j.1467-9280.2007.01962.x
- Kelman, H. C. (1973). Violence without moral restraint: Reflections on the dehumanization of victims and victimizers. Journal of Social Issues, 29(4), 25–61. https://doi.org/10.1111/j.1540-4560.1973.tb00102.x
- Kteily, N., Bruneau, E., Waytz, A., & Cotterill, S. (2015). The ascent of man: Theoretical and empirical evidence for blatant dehumanization. Journal of Personality and Social Psychology, 109(5), 901–931. https://doi.org/10.1037/pspp0000048
- Kteily, N., Hodson, G., & Bruneau, E. (2016). They see us as less than human: Metadehumanization predicts intergroup conflict via reciprocal dehumanization. Journal of Personality and Social Psychology, 110(3), 343–370. https://doi.org/10.1037/pspa0000044
- Opotow, S. (1990). Moral exclusion and injustice: An introduction. Journal of Social Issues, 46(1), 1–20. https://doi.org/10.1111/j.1540-4560.1990.tb00268.x
- Paluck, E. L. (2009). Reducing intergroup prejudice and conflict using the media: A field experiment in Rwanda. Journal of Personality and Social Psychology, 96(3), 574–587. https://doi.org/10.1037/a0011989
- Paluck, E. L., Green, S. A., & Green, D. P. (2019). The contact hypothesis re-evaluated. Behavioural Public Policy, 3(2), 129–158. https://doi.org/10.1017/bpp.2018.25
- Pettigrew, T. F., & Tropp, L. R. (2006). A meta-analytic test of intergroup contact theory. Journal of Personality and Social Psychology, 90(5), 751–783. https://doi.org/10.1037/0022-3514.90.5.751
- Pratto, F., Sidanius, J., Stallworth, L. M., & Malle, B. F. (1994). Social dominance orientation: A personality variable predicting social and political attitudes. Journal of Personality and Social Psychology, 67(4), 741–763. https://doi.org/10.1037/0022-3514.67.4.741
- Riek, B. M., Mania, E. W., & Gaertner, S. L. (2006). Intergroup threat and outgroup attitudes: A meta-analytic review. Personality and Social Psychology Review, 10(4), 336–353. https://doi.org/10.1207/s15327957pspr1004_4
- Sherif, M., Harvey, O. J., White, B. J., Hood, W. R., & Sherif, C. W. (1961). Intergroup conflict and cooperation: The Robbers Cave experiment. University Book Exchange. https://psychclassics.yorku.ca/Sherif/
- Tajfel, H., Billig, M. G., Bundy, R. P., & Flament, C. (1971). Social categorization and intergroup behaviour. European Journal of Social Psychology, 1(2), 149–178. https://doi.org/10.1002/ejsp.2420010202
- Vaes, J., Paladino, M. P., Castelli, L., Leyens, J.-P., & Giovanazzi, A. (2003). On the behavioral consequences of infrahumanization: The implicit role of uniquely human emotions in intergroup relations. Journal of Personality and Social Psychology, 85(6), 1016–1034. https://doi.org/10.1037/0022-3514.85.6.1016
- Waytz, A., Gray, K., Epley, N., & Wegner, D. M. (2010). Causes and consequences of mind perception. Trends in Cognitive Sciences, 14(8), 383–388. https://doi.org/10.1016/j.tics.2010.05.006
Social Identity and the Psychology of Group Boundaries
Social identification means incorporating a group membership into the working representation of self and socially relevant others. Categorization affects attention, comparison standards, interpretation of behavior, and responses to threats involving collective standing. In Tajfel et al. (1971), allocating resources in favor of the ingroup occurred without established intergroup hostility or personal gain, demonstrating the explanatory value of categorization. Yet the study does not show that every category boundary produces hostility, because the response was a particular laboratory choice under restricted conditions.
Group identities are multiple and context sensitive. Cikara & Van Bavel (2014) reviews neuroscience alongside classic group-process research and emphasizes functional relations and changing representations of us and them. Perceived norms, common tasks, leadership messages, and overlapping memberships can change which identity matters in a given encounter. Researchers should therefore assess salience, centrality, identity threat, and category accessibility rather than treat ethnic, religious, national, organizational, and digital categories as psychologically equivalent or invariably fixed.
Identity protection may encourage defensive evaluation or perceived collective victimization without requiring a judgment that the other group is subhuman. Brewer (1999) offers a caution against treating ingroup affection as logically sufficient for outgroup hostility. The transition from identification to antagonism becomes an empirical question: which threats are salient, what are the expected actions of the outgroup, which emotions organize response, and when do stereotypes become moralized? Only after these intermediate mechanisms are measured can a model explain how ordinary social boundaries become aligned with severe exclusion.