Psychology by BogdanovaThe Encyclopedia of Psychology for the Artificial Era

RESEARCH DOSSIER

Political Radicalization, Extremism, and Violence

Abstract. Political radicalization, extremism, and violence encompass distinct political, social, and behavioral processes, requiring separate measures of radical belief, endorsement of harm, intention, participation, and actual violent conduct. This fundamental research monograph critically synthesizes the historical development of radicalization studies, psychological and multilevel explanations of grievance, significance seeking, moral disengagement, group identity, relational influence, individual differences, and institutional conditions. Original offender studies, qualitative disengagement research, field-wide meta-analyses, systematic reviews, and intervention experiments are examined with explicit attention to causal inference, rare outcomes, measurement, selection bias, and external validity. The synthesis evaluates evidence on mental health, the limitations of violent-extremism risk assessment instruments, digital environments, and social reintegration. It also examines institutional prevention, the protection of lawful dissent, and research priorities for human–artificial political information environments. The scientific conclusion is that pathways differ and that attitudes, intent, and observed violence must remain distinct outcomes. This work is a critical narrative synthesis of published research; it reports neither original participant data nor an independently registered systematic review.

Keywords: political radicalization; violent extremism; political violence; terrorism psychology; social identity; significance quest; moral disengagement; radical attitudes; violent intentions; behavioral radicalization; risk and protective factors; lone-actor violence; deradicalization; disengagement; prevention; political psychology; forensic psychology; artificial political communication

Introduction: Explaining Political Radicalization Without a Single Pathway

Political radicalization, extremism, and violence form an interconnected but differentiated research domain. Political radicalization concerns changes in commitments, identities, narratives, and preferred methods of political action. Extremism refers to a position or practice judged markedly distant from a specified political or moral baseline; its substantive content depends on that baseline and on whether the criterion is ideological exclusion, rejection of pluralism, endorsement of coercion, or actual conduct. Political violence denotes observable attempts to inflict physical harm for political ends. These outcomes overlap in some trajectories but remain analytically distinct. The central scientific problem is explaining how a minority of people move from grievance or identification toward endorsing, preparing for, or committing violence while many people with intense grievances remain nonviolent Borum (2011a) McCauley & Moskalenko (2008) Wolfowicz et al. (2021).

This monograph offers a critical narrative synthesis of original social-psychological research, criminology, terrorism studies, forensic case comparisons, meta-analyses, and intervention experiments. It examines mechanisms at individual, relational, organizational, and institutional levels and interrogates what each research design actually measures. Its questions concern the meaning and measurement of radicalization; the historical formation of competing models; the explanatory contribution of threat, identity, significance, moral exclusion, and social networks; the transition from attitudes to behavior; and the evidence for disengagement and prevention. The synthesis is based on published evidence rather than original participant data or an independently registered systematic-review protocol Schuurman (2020) Obaidi et al. (2026).

A research vocabulary that collapses radical belief into criminal behavior produces misleading inference. Protest, dissent, doctrinal intensity, and morally uncompromising opposition to injustice can be nonviolent and constitutionally protected. Conversely, violent action sometimes occurs with thin, inconsistent, or opportunistically acquired ideological commitments. Accordingly, the explanandum is specified at each step: expressed ideological commitment, support for political violence, behavioral intention, organization-linked activity, documented threat behavior, violent offense, or cessation of participation. Explanation requires a theory connecting the measured outcome to mechanisms and context, with attention to classification error and political labeling Horgan (2008) Borum (2011b).

Conceptual Boundaries: Radicalism, Extremism, Terrorism, and Political Violence

Radicalism is historically relational: a demand for fundamental transformation can be classified as radical within a particular institutional order without implying hostility to civilians or endorsement of violent methods. The term extremism acquires empirical content only when researchers state their reference point and exclusion criterion. A multidimensional operationalization can distinguish ideological distance from mainstream opinion, categorical hostility to outgroups, unwillingness to tolerate political rivals, belief that violence is legitimate, and documented participation in harmful acts. These indicators need not covary strongly; their empirical relationship is a question for investigation Borum (2011a) Doosje et al. (2016).

Terrorism is an especially contested subset of political violence, involving questions of intended intimidation, target selection, organizational responsibility, and legal jurisdiction. Psychological data alone cannot settle political or legal definitions. Cross-study synthesis requires coding whose definition is declared and consistently applied. If one database counts threats, another counts completed offenses, and a third includes support expressed on a questionnaire, pooled estimates describe mixed outcomes. The distinction between state, nonstate, organized, and individual political violence also matters: authority, command systems, opportunity, and exposure differ across those cases McCauley & Moskalenko (2008) Schuurman (2020).

Researchers must separate the ideological content of a claim from the process by which the claimant comes to adopt it. The same broad ideological family can include parliamentary actors, peaceful movement participants, supporters of disruptive civil disobedience, and perpetrators of violence. Likewise, one violent episode may reflect tactical decisions, retaliation, coercion, or personal motives in different proportions. A useful construct taxonomy therefore records political goal, legitimacy beliefs, group identification, behavioral preparation, actual conduct, and setting as distinct variables. Such specificity is a condition for meaningful comparison across political positions and countries Wolfowicz et al. (2020) Obaidi et al. (2026).

Historical Development of Psychological Research

The modern psychology of radicalization emerged from several partially separate literatures: crowd and intergroup psychology, political socialization, social movement research, studies of war and collective conflict, criminology, and research on terrorism. Rather than identifying a single founder, the history tracks changing scientific questions. Initial explanations often sought characteristic offender profiles or treated ideological commitment as a direct cause. Subsequent work shifted toward processes, situations, networks, and trajectories, including the recognition that selection into militant groups and selection into violence are different events Horgan (2008) Borum (2011a).

McCauley and Moskalenko's 2008 framework organized mechanisms across personal, group, and mass levels, moving the field beyond a homogeneous step-by-step account. Horgan's 2008 analysis separated becoming involved, participating, and disengaging and explained why the failure of psychological profiling did not imply that psychological mechanisms were irrelevant. Borum's two 2011 reviews questioned definitions and emphasized that many proposed stage models rested on plausible descriptions rather than prospective tests. These interventions reshaped the research agenda around the diversity of routes and the need for observed behavioral outcomes McCauley & Moskalenko (2008) Horgan (2008) Borum (2011a) Borum (2011b).

During the 2010s, significance-quest theory, social identity and devoted-actor research, moral-cognition approaches, and event-level offender studies expanded the explanatory repertoire. Later quantitative synthesis distinguished risk correlates of radical attitudes, intentions, and behaviors. The evidence infrastructure also changed. In Schuurman's audit of terrorism scholarship from 2007 through 2016, reliance on literature reviews remained substantial even as the use of primary data increased. The lesson for contemporary monographs is methodological: a persuasive process narrative cannot replace validated measurement, transparent sampling, temporal sequence, and replication Kruglanski et al. (2014) Gill et al. (2014) Schuurman (2020) Wolfowicz et al. (2021).

The Unit of Analysis: Persons, Networks, Organizations, and Political Systems

Individual-level accounts describe how a person perceives exclusion, threat, injustice, status loss, or opportunities for meaning. Social psychological accounts connect these appraisals to reference groups and relations between groups. Network accounts investigate exposure, recruitment, reinforcement, and ties between participants. Organizational accounts examine commitment rules, authority, incentives, and internal contestation. Institutional accounts place these processes within patterns of representation, state responses, discrimination, censorship, policing, war, and political opportunity. Because causal processes operate across levels, explanations that rely solely on personal characteristics or solely on macro grievance omit essential interactions Doosje et al. (2016) McCauley & Moskalenko (2008).

The appropriate research unit changes with the question. An opinion survey can estimate support for a political proposition among sampled respondents, yet it seldom observes later violence. A case series of convicted offenders describes features among selected actors but cannot identify how common those features are among nonoffenders. A study of groups can reveal relational reinforcement while providing limited evidence about private motivations. A national comparison can identify structural differences but cannot automatically assign the individual pathway. Each design offers information that must be interpreted at its own scale Gill et al. (2014) Wolfowicz et al. (2021).

An integrated research design links observations with time and context: what an individual believed at baseline; what events or encounters followed; what network and institutional opportunities were available; what observable behaviors changed; and what alternative outcomes occurred among comparable people. Such a program is difficult because political violence is rare, sensitive, and ethically challenging to study prospectively. This difficulty explains the continued importance of mixed methods, transparent case definitions, and careful assessment of selection bias, rather than justifying confident inference from retrospective accounts Schuurman (2020) Brouillette-Alarie et al. (2025).

Grievance, Relative Deprivation, and Perceived Injustice

Grievance refers to an evaluated discrepancy between conditions experienced or perceived and conditions believed to be deserved. Relative deprivation adds a comparison standard, which may be personal, group-based, historical, or aspirational. Political movements can organize diffuse dissatisfaction into shared injustice claims, and those claims can motivate peaceful as well as coercive political behavior. Therefore grievance alone has limited specificity for violent outcomes. A research design must measure how deprivation is interpreted, the available identities and action repertoires, and why violence appears legitimate or necessary to some actors McCauley & Moskalenko (2008) Doosje et al. (2016).

Perceived threats and perceived humiliation can be associated with anger, fear, resentment, or desires for status restoration. These emotions differ in action tendencies; their relationship to particular conduct is shaped by efficacy judgments and situational norms. The same experience of discrimination can lead to legal advocacy, withdrawal, solidarity networks, voting, protest, or an endorsement of harm. Accounts that identify grievance as a root cause must specify the additional mechanisms that distinguish these trajectories. Systematic syntheses suggest that background demographics and many common experiences have small average associations with radicalization outcomes Wolfowicz et al. (2020) Wolfowicz et al. (2021).

Context also determines the meaning of a grievance report. Participants may describe past events using narratives learned after entering an organization, and retrospective biographical interviews can reorganize earlier experiences into a coherent ideological story. This does not invalidate testimony, but it complicates temporal causal claims. Longitudinal studies, contemporaneous documents, independent behavioral records, and comparison groups help distinguish early vulnerability, subsequent rationalization, and the influence of shared movement discourse Horgan (2008) Bubolz & Simi (2015).

Social Identity, Group Boundaries, and Belonging

Political identity links personal self-understanding to a collectivity whose status, goals, and moral standing matter to the individual. Identification can increase perceived obligation, emotional responsiveness to group-related events, and sensitivity to norms. These outcomes do not intrinsically involve violence; the normative content of the group and the behavioral options presented to members determine the direction of commitment. Empirically, research must separate strength of identification from belief in violent means and from participation in an organization that legitimizes such means Doosje et al. (2016) Kruglanski et al. (2014).

Strong group boundaries can sharpen distinctions between in-group and outgroup, but antagonism takes different forms, from ordinary competition to political exclusion or dehumanization. The presence of boundary-making does not establish that a person will attack outsiders. The model requires observations of moral norms, hostile expectations, leader and peer cues, and permissible action. Group participation can also produce belonging and protective social bonds that favor withdrawal from violence, showing why a group's content and relational practices matter more than the general fact of social attachment Bandura (1999) Bubolz & Simi (2015).

Identity fusion and sacred-value research adds a more specific proposition: when personal identity is experienced as strongly joined to a valued collective, some participants report unusually high willingness to sacrifice for its purposes. Evidence from devoted-actor research supports relationships among fusion, sacred commitments, and stated costly intentions in particular samples. The outcome is usually reported willingness, not independent proof of future violence, and the relevant moral commitments can support defense or solidarity rather than aggression. Transfer to new movements therefore depends on measurement equivalence and political context Sheikh et al. (2014) Atran et al. (2014).

The Quest for Significance and Motivational Reorientation

Significance-quest theory proposes that people seek a sense that their lives matter and that threats to this standing can motivate efforts to regain it. In the framework developed by Kruglanski and colleagues, motivational activation interacts with an ideology that presents violence as a legitimate means and with social processes that transmit and reinforce this interpretation. The theory makes an important conceptual advance because a desire for significance has no intrinsically violent content; the legitimization of harmful action is a distinct explanatory step Kruglanski et al. (2014).

Significance loss may be personal, such as humiliation or exclusion, or connected to the perceived fate of a larger collective. The empirical challenge is establishing whether significance concerns precede involvement, are created or intensified by participation, or are reconstructed after an event. Surveys can test covariance between perceived significance and attitudes, while experiments can manipulate narrower experiences and observe immediate responses. Generalizing such findings to actual serious violence requires behavioral evidence, temporal follow-up, and plausible comparisons Kruglanski et al. (2014) Obaidi et al. (2026).

Motivational accounts gain specificity when they describe interactions rather than an isolated drive. An individual may experience frustration while having strong attachments to nonviolent social roles, alternative routes to achievement, or group norms rejecting harm. Another person may encounter a network that frames coercion as morally necessary. Differences in identity, perceived feasibility, social rewards, and available opportunities help explain divergent outcomes. This formulation invites falsifiable hypotheses about moderators instead of treating a broadly human need as a sufficient explanation for rare violence Wolfowicz et al. (2021) Doosje et al. (2016).

Moral Cognition, Justification, and Dehumanization

Moral disengagement theory describes processes through which self-sanctions against harming others can be weakened or selectively bypassed. Bandura identified moral justification, euphemistic labeling, advantageous comparison, displacement or diffusion of responsibility, minimization of consequences, and dehumanization among the mechanisms by which conduct becomes easier to rationalize. The scientific contribution is a set of processes potentially observable in discourse and judgment, rather than a diagnosis of a type of person. Their relevance to political violence should be tested through links to specified beliefs and outcomes Bandura (1999).

In conflict settings, perceived threats and narratives of victimhood can interact with moral categories, permitting actors to classify some targets as outside the ordinary sphere of moral concern. The evidentiary distinction is between utterances that describe an outgroup negatively, attitudes endorsing harms, and concrete actions. Content analysis can quantify rhetorical patterns, but rhetorical prevalence is not a direct estimate of violent behavior. A robust interpretation asks whether changes in discourse precede conduct and whether comparable rhetoric exists in nonviolent contexts McCauley & Moskalenko (2008) Bandura (1999).

Moral conviction can also motivate protective intervention, peaceful resistance, and principled refusal to participate in cruelty. Its role is therefore conditional on the substantive standard being defended and on the perceived acceptability of means. The analytic question concerns how normative commitments relate to both constraint and permission. Researchers should include measures of restraint, support for rights, and opposition to civilian harm when examining moralized politics, so that intensity of conviction is not treated automatically as an indicator of danger Atran et al. (2014) Voelkel et al. (2024).

Perceived Threat, Authoritarian Dispositions, and Hierarchy

Threat perception is a politically consequential appraisal of danger to security, status, values, or a familiar social order. It can increase support for protective institutions or restrictive policy, and its association with violence depends on further mechanisms. Authoritarianism research investigates orientations to conformity, conventional order, and coercive responses to perceived deviance. Social dominance frameworks investigate preferences for group hierarchy. These constructs should not be fused with observed extremist violence, because their measurement domains and political manifestations differ Obaidi et al. (2026) Wolfowicz et al. (2021).

Some research traditions posit that perceived societal threat activates latent tendencies toward stronger authority and norm enforcement; others emphasize ideological socialization and enduring individual differences. The tests require comparing models that allow reciprocal influence: threats may alter attitudes, while prior political commitments shape which events count as threatening. Cross-sectional correlations cannot resolve these directions. Experimental threat manipulations may establish short-term attitude effects but do not by themselves demonstrate changes in criminal conduct Doosje et al. (2016) Obaidi et al. (2026).

A defensible synthesis treats threat as a general motivational context whose expression is filtered by institutions, identities, and ideologies. Its effects can include support for law enforcement, protest, electoral change, withdrawal, solidarity, or exclusion, depending on conditions. It is essential to identify the relevant population and political circumstance. Measures developed in one national setting should undergo translation, invariance, and criterion validation before generalized claims are made about a second society Wolfowicz et al. (2020) Brouillette-Alarie et al. (2025).

Individual Differences, Personality, and Psychological Vulnerability

Individual differences research asks why people facing similar situations respond differently. Obaidi, Bergh, Benningstad, and Dovidio propose integrating broad trait-descriptive frameworks with process-oriented explanations of social perception, motivation, and cognition. Their 2025 online review, appearing in the 2026 journal volume, challenges the assumption that perpetrators must share either an ordinary uniform psychology or a common psychiatric diagnosis. Traits may influence selection into environments as well as sensitivity to those environments; both pathways require testing Obaidi et al. (2026).

Correlational studies can identify associations between measured traits and ideological attitudes, but their effects vary with the operational outcome, sampling frame, and instruments used. A personality score is not a direct prediction of a rare violent event. Statistical models that appear accurate in case–control studies can fail when applied in populations where the event's base rate is very low. The field-wide reviews by Wolfowicz and colleagues found generally small associations for many sociodemographic predictors and highlighted greater, though still context-dependent, estimates for selected criminogenic and psychological correlates Wolfowicz et al. (2020) Wolfowicz et al. (2021).

Clinical claims require special precision. Mental disorder can coexist with grievance-fueled targeted violence, and individual symptoms may affect some pathways, but a diagnosis is neither a general definition of political extremism nor a sufficient explanation of an offense. Studies comparing lone-actor and group-based offenders confront small and selected samples, disputed diagnostic reconstruction, and differential information availability. Evidence reviews favor differentiated analysis of symptoms, functioning, personal circumstances, and behavior, with mental healthcare separated from ideological surveillance Corner & Gill (2015) Corner et al. (2018) Sarma et al. (2022).

Small Groups, Networks, and Relational Reinforcement

Radicalization frequently involves social relationships, although the organizational form can vary from durable groups to loose communities and private interpersonal ties. Peer recognition can confirm identity, normalize narratives, and make commitments costly to abandon. Networks also influence whether people encounter alternative accounts and whether they perceive violent action as approved by relevant others. McCauley and Moskalenko distinguish personal grievances from group processes and mass conflict dynamics, suggesting a set of mechanisms that need not occur in a fixed order McCauley & Moskalenko (2008).

Strong ties and social environments can facilitate recruitment, but they can also anchor individuals to nonviolent responsibilities. An adequate network analysis distinguishes frequency of contact, trust, normative influence, brokerage, and actual participation. Observing that perpetrators had contacts with a movement does not establish that contact caused violence; selection and homophily may operate. Longitudinal relational data and comparison with nonviolent members would help identify whether exposure changes trajectories over and above prior beliefs Sageman (2008) Doosje et al. (2016).

Accounts that call an actor isolated should distinguish social solitude from organizational independence. Gill, Horgan, and Deckert analyzed 119 lone-actor terrorists and found no uniform profile. Their results challenge the everyday assumption that an individual offender must have been detached from broader networks. The sample concerns identified perpetrators and therefore cannot estimate prevalence or prospective risk in the general population. Its main scientific importance is the detailed attention to prior conduct and relational contexts rather than caricatured psychological portraits Gill et al. (2014).

Ideological Socialization and the Attitude–Behavior Gap

Ideological socialization refers to processes by which political explanations, group histories, normative obligations, and moral evaluations become integrated into an individual's understanding. These processes operate through education, relationships, organizations, media, and repeated discussion. Their endpoints vary: belief adoption, more intense identification, activism, withdrawal from opponents, or violent endorsement. A causal theory must explain why identical messages have divergent effects across individuals and why some actors persist in peaceful engagement Borum (2011b) Doosje et al. (2016).

The attitude–behavior gap is the central measurement issue in this domain. A respondent's stated approval of violence, expressed intention to take part, behavioral preparations, and completed offenses are not interchangeable. Different predictors may matter at each point. The 2021 Campbell systematic review of 127 studies and 206 samples extracted 1,302 effect sizes, examining over 100 putative factors across radical attitudes, intentions, and behaviors. Many included studies were cross-sectional, limiting claims about causal ordering or forecast accuracy Wolfowicz et al. (2021).

The analytic priority is to model discontinuity as well as escalation. People may adopt severe rhetoric without engaging in violence, commit offenses during acute conflicts without a deep ideological history, or disengage behaviorally while retaining beliefs. Treating ideology as a simple causal pipeline tends to obscure those possibilities. Models should state what evidence would falsify a proposed link and which competing motives or situational conditions might explain the same outcome Horgan (2008) Borum (2011a) Bubolz & Simi (2015).

From Political Support to Observable Violent Conduct

The move from endorsement to conduct involves opportunities, capacities, constraints, and contingencies that attitude questionnaires do not observe. A person may approve of violence abstractly yet decline all participation, while another may act in a localized conflict without sustained ideological endorsement. Researchers therefore distinguish cognitive radicalization, stated intention, participation in an organization, concrete planning or threats, and perpetration. Mechanisms can differ across these outcomes, and movement between them may involve reversals or long periods of inactivity rather than uninterrupted progression Wolfowicz et al. (2020) Wolfowicz et al. (2021).

Observational case studies can reconstruct stages of involvement but frequently lack a counterfactual sample of individuals who experienced comparable events and did not offend. The interpretation must account for selection on the outcome: after a violent act, formerly ordinary biographical details may be retrospectively classified as warning signs. Case–control matching, blinded coding, pre-event records, and prospective observation would improve causal discrimination, though access and ethical constraints are significant. Offense-specific information is often incomplete and collected for investigative rather than scientific purposes Gill et al. (2014) Brouillette-Alarie et al. (2025).

Behavioral outcomes should remain central when evaluating interventions. A measured decline in animosity or support for violence is potentially meaningful, but it is not equivalent to a reduction in attacks. Conversely, direct observation of rare harmful events may be infeasible in a randomized study. The appropriate evidence strategy links intermediate outcomes to later behavior where possible, tests alternative pathways, and communicates what is estimated at each step. Voelkel and colleagues' megastudy illustrates that effects on partisan animosity and support for political violence can diverge Voelkel et al. (2024).

Lone Actors, Organizational Participants, and Distinct Pathways

The lone-actor label usually describes the operational organization of an offense rather than a person completely separated from social influence. Lone actors may participate in online communities, consume movement materials, communicate with others, or act in response to wider political environments. Gill and colleagues' behavioral study of 119 cases demonstrated substantial heterogeneity and the limits of a single sociological or psychological profile. These findings justify investigating network ties, observable preparation, and personal circumstances separately instead of assuming that organizational independence implies psychological isolation Gill et al. (2014).

Comparing lone and group offenders introduces selection concerns. Detectable group activity differs from solitary conduct, and available records vary with publicity, jurisdiction, and investigation. Studies of psychiatric symptoms, employment, social relationships, and earlier offenses must avoid implying causal roles from overrepresentation among already identified offenders. If a potential predictor is also a criterion used to construct the sample, circular inference can result. Reproducing published comparisons requires transparent case inclusion, coding rules, and handling of missing data Corner & Gill (2015) Corner et al. (2018).

Grievance-fueled targeted violence also includes incidents with weak or contested political motivation. Ebbrecht's systematic review of 78 studies across several lone-actor categories identifies overlapping psychological mechanisms, yet these categories contain important differences in goals, social contexts, and opportunities. A cross-type synthesis can illuminate shared needs for belonging, grievances, and moral disengagement, but it cannot assume that every targeted offender is politically radicalized. The political nature of the target, stated motivation, and historical setting must be established from evidence Ebbrecht (2023).

Digital Information Environments and Online Influence

Digital networks provide spaces for political identity expression, collective grievance narratives, interpersonal validation, and exposure to extremist interpretations. Platforms affect discoverability and social interaction, yet evidence of exposure alone cannot demonstrate that a digital medium caused a particular person to become violent. Researchers must measure selection into communities, preexisting attitudes, offline relationships, platform governance, and subsequent conduct. Observational associations between viewing content and holding attitudes may reflect reciprocal choice as well as persuasion Sageman (2008) Doosje et al. (2016).

Important scientific distinctions include availability of content, actual exposure, attention, interpretation, behavioral engagement, and later political action. Platform traces can measure clicks or reposts at high temporal resolution, but they rarely capture the full motive or offline social setting. Surveys can gather interpretations but rely on self-report. Designs that combine consented digital traces with panel measures and independent outcomes offer a more complete picture; however, differential access to proprietary data and changing recommendation systems limit replication and generalization Schuurman (2020) Wolfowicz et al. (2021).

Artificial systems introduce new forms of political communication through conversational agents, synthetic media, and algorithmic selection. A research program for the Artificial Era should test how system outputs influence interpretation, perceived credibility, relationship-like attachment, and intentions, while versioning the model, prompts, interface, and exposure period. These are hypotheses about human behavior in mediated environments; neither radicalization nor the psychological identity of an artificial system can be inferred from a single interaction. Ethical inquiry includes the risk of selective amplification, biased classification, and unequal treatment of political communities Obaidi et al. (2026) Brouillette-Alarie et al. (2025).

Comparative Political Contexts and Cross-Cultural Generalization

Political violence occurs across different institutional, historical, economic, and legal environments. A research instrument developed around one form of ideological conflict may fail to represent another. Terms such as extremism or radicalization can carry different official meanings, and data availability differs between open democracies, conflict zones, and repressive settings. Comparative inference therefore requires documented definitions, translation equivalence, sampling practices, and sensitivity to political power relations Doosje et al. (2016) Borum (2011a).

Most factor meta-analyses cover particular populations more densely than others. The Campbell review by Wolfowicz and colleagues drew on studies largely conducted in OECD settings and found considerable variation in measures and outcomes. Its estimates should be interpreted as pooled associations within those evidence boundaries, rather than universal coefficients for all forms of political conflict. The absence of particular regions from a synthesis does not imply absence of the underlying processes; it identifies a limitation of the evidence base Wolfowicz et al. (2021).

Cross-cultural research can examine whether mechanisms such as grievance, identity fusion, moral exclusion, or significance concerns display similar relations to outcomes after establishing measurement invariance. Contextual moderators might include conflict intensity, state violence, historical discrimination, access to lawful expression, communal networks, and the availability of institutional redress. These variables should be measured at appropriate levels and tested rather than assumed. The comparative aim is to explain heterogeneity, including peaceful responses to severe grievances, without treating any population or ideology as inherently violent Obaidi et al. (2026) McCauley & Moskalenko (2008).

Study Designs, Sampling, and Construct Validity

The field draws on representative surveys, convenience experiments, offender interviews, court records, biographical case studies, digital traces, ethnography, organizational documents, and systematic evidence syntheses. Each source answers a distinct question. Surveys estimate reported orientations within their sampling frame; experiments can isolate effects on short-term measured responses; qualitative accounts reconstruct subjective meanings; and offender data describe selected acts. No single method simultaneously supplies all the causal, temporal, and contextual information needed to explain rare violence Schuurman (2020) Wolfowicz et al. (2021).

Construct validity begins with distinguishing the phenomenon to be measured from the instrument used to observe it. Radical attitudes may be operationalized through abstract approval of force, violence in specified scenarios, willingness to sacrifice, authoritarian preferences, or membership in particular political groups. These are different measures and should be examined for convergent and discriminant validity. Scales may be influenced by item wording, social desirability, threat of disclosure, translation, and changing political interpretations. Direct measures of real conduct need clear event coding and independent source corroboration Borum (2011b) Brouillette-Alarie et al. (2025).

Good inference also demands attention to base rates. Even a marker associated with offending can generate many false positives when the target event is infrequent. Researchers should report denominators, population selection, outcome prevalence, calibration, uncertainty, and the intended purpose of a score. Selecting offenders and then identifying common features yields explanatory hypotheses, not a validated screening tool for the general population. Blind coding, preregistration, replication datasets, and careful missing-data analyses are especially valuable Gill et al. (2014) Brouillette-Alarie et al. (2025).

Meta-Analytic Evidence and the Interpretation of Risk Factors

A meta-analysis aggregates estimates from studies with defined inclusion criteria, outcomes, and statistical models; its conclusions are constrained by the original literature. Wolfowicz, Litmanovitz, Weisburd, and Hasisi's 2020 field-wide review identified 57 publications and 62 individual-level factors across attitudes, intentions, and actions. Their later Campbell synthesis included 127 studies, 206 samples, and 1,302 effect sizes on more than 100 factors. Both programs emphasize heterogeneity and the limited value of many commonly cited demographic markers Wolfowicz et al. (2020) Wolfowicz et al. (2021).

The presence of a statistical association is distinct from prospective prediction and from causal effect. Pooled estimates can combine instruments with different meanings, samples differing in risk, and adjustment strategies that treat confounders differently. Publication selection may favor striking correlates, while sparse behavior outcomes yield unstable estimates. The appropriate reading is comparative: certain factors show stronger associations than others in the measured datasets, but effect-size ranking alone does not define an intervention policy or an individual's risk Wolfowicz et al. (2021).

Critical synthesis should include negative and contradictory findings. Evidence for some popular narratives is weak when tested against multiple outcomes or sample types, and empirical support varies by ideology and geography. Scientific progress therefore depends on resolving why findings differ: measurement, time period, population, moderator, sampling, or genuine contextual variation. Meta-analytic patterns are most informative when combined with longitudinal and experimental tests that can specify where a proposed mechanism operates Wolfowicz et al. (2020) Obaidi et al. (2026).

Violence Risk Assessment, Behavioral Evidence, and False Positives

Assessment of possible extremist violence belongs to forensic and behavioral risk practice, where the unit of concern is a clearly specified potential harm. Tools such as TRAP-18, ERG22+, VERA, and related structured professional judgment instruments organize information that may be relevant to case formulation. A checklist's conceptual coverage, however, is not the same as demonstrated prospective predictive validity. Risk assessment also requires context, professional judgment, dynamic change, and consideration of protective factors, rather than automatic translation of ideological language into a score Brouillette-Alarie et al. (2025).

A 2025 Campbell systematic review by Brouillette-Alarie and colleagues examined 20 eligible studies across six instruments. It found potentially useful evidence for some aspects of reliability but substantial problems in construct validation and serious limitations arising from small and retrospective samples. Crucially, the studies did not provide true prospective predictive validation for future violent extremism. Estimates from retrospective comparisons can be distorted when outcome status was already known to coders. Accordingly, a scientific report should not represent any existing risk tool as a reliable universal detector of future political violence Brouillette-Alarie et al. (2025).

The ethical and statistical risks are connected. In populations where political violence is uncommon, even a seemingly informative marker can falsely classify many peaceful individuals. Differential surveillance can also produce unequal rates of recorded suspicious activity, introducing feedback into the data used to validate assessments. Practices consistent with scientific rigor include explicit outcomes, proportionality, regular reassessment, procedural safeguards, transparency about uncertainty, and careful distinction between protected political expression and credible behavioral evidence of harm Wolfowicz et al. (2021) Brouillette-Alarie et al. (2025).

Mental Health, Psychiatric Classification, and Explanatory Limits

Political extremism is a social and behavioral research construct rather than a psychiatric diagnosis. Studies of mental disorder among individuals involved in terrorism have produced varied results depending on sample definitions, available records, and distinctions between individual and group involvement. Corner and Gill's work investigates whether simplistic claims about illness obscure important heterogeneity. Subsequent research has emphasized the need to distinguish established diagnoses, symptoms inferred from public biographies, and potential functional pathways connecting distress to particular behaviors Corner & Gill (2015) Corner et al. (2018).

Systematic review evidence must be interpreted without importing the stigma of dangerousness into clinical categories. Sarma, Carthy, and Cox examined mental disorder and terrorist behavior with explicit attention to prevalence and temporality. Studies in this area face selection effects, reliance on retrospective records, and heterogeneous disorder classifications. Even where symptoms are associated with selected cases, a causal account requires evidence of how they interacted with immediate circumstances, opportunity, grievances, relationships, and intentional action Sarma et al. (2022) Corner et al. (2018).

Clinical assessment may be necessary when there are specific concerns about distress, functional impairment, or safety, but clinical care and security evaluation have distinct purposes and professional obligations. Neither religious affiliation, political ideology, nor unconventional belief is a diagnostic symptom in itself. The research priority is identifying legitimate therapeutic needs while maintaining valid behavioral criteria for violence risk. This distinction protects scientific validity and preserves access to care for people whose political beliefs or group memberships are socially contested Corner et al. (2018) Brouillette-Alarie et al. (2025).

Disengagement, Deradicalization, and Social Reintegration

Disengagement denotes reduction or cessation of participation in violent organizations or activities. Deradicalization refers to changes in beliefs, norms, or identities judged relevant to extremism. They can occur together but are empirically independent: someone can leave violent action while retaining strong convictions, or revise beliefs without immediately escaping organizational obligations. Horgan's pathway account identifies disengagement as a separate phase requiring its own explanatory evidence. Interventions should therefore state their desired outcome before claiming success Horgan (2008) Doosje et al. (2016).

Leaving an extremist social world can involve practical, relational, and moral difficulties. Bubolz and Simi used life-history interviews with 34 former white supremacists to examine transitions out of group involvement. Their exploratory findings describe varied routes, continuing social ties, identity change, guilt, and the complexity of exit. Because participants were selected after leaving, the design provides valuable qualitative insight but cannot establish how frequently each mechanism produces successful disengagement among all members or whether a program caused exit Bubolz & Simi (2015).

A defensible framework examines changing personal roles, education and work opportunities, relationships outside violent networks, trusted support, organizational barriers to leaving, and how participants understand responsibility for earlier conduct. These elements should be evaluated as context-dependent hypotheses rather than a universal sequence. Long-term follow-up is necessary to distinguish short-term compliance, durable behavioral change, renewed association, and genuine changes in tolerance of harm. The evidence also supports treating constructive social bonds as possible pathways toward nonviolent participation Bubolz & Simi (2015) Horgan (2008).

Intervention Research: Attitudes, Norms, and Behavioral Outcomes

Interventions for polarization and extremist violence operate on different targets. Some seek to reduce hostile beliefs; others support disengagement, strengthen community ties, offer services, or change institutional conditions. A controlled evaluation must define whether success means lower reported animosity, reduced endorsement of violence, fewer threatening behaviors, stable exit, or fewer harmful events. The validity of an outcome measure is particularly important when serious violence is too rare for routine experimental observation Voelkel et al. (2024) Wolfowicz et al. (2021).

Voelkel and colleagues conducted a preregistered megastudy of 25 interventions with 32,059 participants in the United States. Their results showed that many interventions reduced partisan animosity but that effects on antidemocratic attitudes and support for political violence were more selective. This separation is an important experimental finding: improving interparty feelings does not necessarily change the normative acceptance of coercion. The results concern survey outcomes under the study's conditions and should not be presented as proven reductions in future violent offenses Voelkel et al. (2024).

Program evaluation also requires independent controls, transparent eligibility rules, preregistered hypotheses, attention to attrition, and measurement of potential harms. Participants' reluctance to disclose attitudes, institutional pressure, or program incentives can bias apparent improvement. Where randomized allocation is inappropriate, carefully matched longitudinal comparisons and multiple independent data sources may help. Researchers should report follow-up horizons and distinguish the program's mechanism from other social changes occurring during treatment Brouillette-Alarie et al. (2025) Bubolz & Simi (2015).

Community Prevention, Institutions, and Civil Liberties

Violence prevention has psychological and institutional dimensions. Evidence-informed efforts may include trusted services, opportunities for nonviolent participation, credible conflict resolution, educational support, and responses to specific threats. Their scientific rationale depends on the identified risk process and the community's context. A generalized assumption that politically marginalized or highly religious communities represent a source of danger can intensify stigma and undermine trust. Measures should be evaluated for benefits and adverse effects among both targeted participants and the broader public Wolfowicz et al. (2021) Doosje et al. (2016).

Counterterrorism and security policies can also influence perceptions of fairness, legitimacy, and collective identity. The appropriate research question is whether a particular intervention changes relevant behavior under specified institutional conditions, not whether its intentions are desirable. Surveillance, criminal justice responses, and community partnership must be examined as potential moderators of grievance and cooperation as well as sources of practical protection. Transparent oversight and clear protection of lawful dissent help separate prevention from ideological policing Borum (2011a) Brouillette-Alarie et al. (2025).

Policy transfer demands caution because violence types, local institutions, and eligible populations differ. The meta-analytic literature does not support broad demographic profiling as an effective stand-alone prevention approach, and the predictive validation of specialized assessment instruments remains incomplete. Evaluation should incorporate baseline incidence, false-positive burdens, equitable access to services, independent complaint channels, and long-term outcomes. These are concrete requirements of sound research design as well as of public accountability Wolfowicz et al. (2020) Wolfowicz et al. (2021) Brouillette-Alarie et al. (2025).

Critical Scientific Discussion: Competing Explanations and Uncertainty

The literature offers several complementary explanations: grievance and threat appraisals, significance seeking, ideological justification, social identification, network reinforcement, dispositional differences, and institutional conditions. Their explanatory adequacy depends on what is to be predicted. Some models explain the adoption of beliefs, others declared willingness to sacrifice, and still others the organization of actual offenses. A unified account cannot be established merely by combining their vocabularies; it requires explicit multilevel hypotheses, temporally ordered evidence, and comparisons with peaceful trajectories Kruglanski et al. (2014) McCauley & Moskalenko (2008) Obaidi et al. (2026).

Three recurring problems are especially consequential. First, rare-event outcomes create severe classification and base-rate challenges. Second, case histories often contain reconstructed motivations and selective documentation, making causal inference vulnerable to hindsight. Third, political and legal definitions of extremism vary, and heterogeneous measurement can create the appearance of agreement where studies investigate different constructs. Schuurman's methodological review and later Campbell syntheses document the field's progress while preserving these limitations as central scientific problems Schuurman (2020) Wolfowicz et al. (2021) Brouillette-Alarie et al. (2025).

The strongest empirical conclusions are comparatively modest and still important. There is no single universal perpetrator profile; pathways differ; attitudinal endorsement and violent behavior require separate measurement; and common background variables often contribute little predictive information. Motivational, cognitive, and relational processes plausibly interact, but their individual roles in escalation are conditionally supported and often difficult to estimate prospectively. Effective scientific synthesis requires evaluating each claim at the outcome and evidence level established by its source Gill et al. (2014) Horgan (2008) Wolfowicz et al. (2021) Obaidi et al. (2026).

Future Research: Longitudinal, Comparative, and Computational Evidence

A more cumulative science would follow diverse participants over time, including people with intense grievances who remain peaceful, those who temporarily endorse violence, those who disengage, and individuals who come into contact with criminal justice systems. Repeated measures could illuminate transitions and reversals while reducing retrospective bias. The ethical design must protect privacy, consent, and political expression, and reporting should avoid reidentification of vulnerable participants. Such designs are demanding but offer greater leverage on causality than retrospective case selection alone Horgan (2008) Wolfowicz et al. (2021).

Cross-national collaboration should establish common outcome definitions while preserving contextual variables. Open codebooks and interoperable metadata can help distinguish violence against civilians, intimidation, threats, political crime, and lawful disruptive protest. Preplanned subgroup analyses can examine ideology, age, gender, social position, and institutional conditions without presuming universal effects. Independent replication and calibration across different base rates are essential before a risk measure is used for consequential decisions Schuurman (2020) Brouillette-Alarie et al. (2025).

New studies of digitally mediated political life should examine selection into online spaces, changes in perceived authority, human–artificial communication, group reinforcement, and the trajectories of beliefs and actions. Artificial agents can be research instruments and communication partners, but claims about their effects require documented system versions, standardized exposures, observed outcomes, and safeguards against experimental harm. The broader scientific objective is a process-based understanding of political change and conflict whose explanatory force survives movement between platforms, political traditions, and historical circumstances Obaidi et al. (2026) Voelkel et al. (2024).

Developmental Time, Life Transitions, and Biographical Change

Psychological trajectories unfold over years, but many studies observe only a narrow moment. School-to-work transitions, shifting family relationships, exposure to conflict, changed community roles, and institutional experiences can alter both political participation and opportunities for social belonging. Life-course interpretation requires specifying age-graded experiences without making youth, unemployment, or social dislocation into deterministic categories. Pooled estimates of common demographic factors are generally limited, while qualitative exit studies describe the importance of turning points and new social commitments Wolfowicz et al. (2020) Bubolz & Simi (2015).

The same individual's political convictions can change in content, salience, and behavioral expression. Some participants intensify their rhetoric after organizational exit while reducing willingness to act; others enter peaceful institutions or develop new relationships that make former practices less attractive. Retrospective narratives may attach decisive meaning to a single turning point, but longitudinal evidence is required to establish when changes began and what sustained them. Disengagement research therefore benefits from repeated measures of identity, social ties, behavior, and perceived legitimacy Horgan (2008) Bubolz & Simi (2015).

A developmental research program should follow multiple outcomes, including civic participation, tolerance of opponents, reported psychological well-being, and violence-related behavior. It should examine normative transitions and institution-level opportunities in different societies, while preserving the confidentiality of politically sensitive trajectories. The aim is to understand how people revise political commitments across time, an essential question that broad personality profiles and single-event retrospective studies are poorly equipped to answer Obaidi et al. (2026) Wolfowicz et al. (2021).

Consequences of Political Violence and the Psychology of Its Social Environment

Political violence affects more than those who directly participate in it. Communities experience fear, grief, disrupted trust, and changes in their expectations of institutions and neighboring groups. These consequences can influence subsequent threat appraisal, collective identity, political rhetoric, and the willingness to seek peaceful resolution. Explanatory work should distinguish the psychology of perpetrators, the experiences of victims and survivors, and the collective effects on communities; their causal questions and ethical obligations are different Doosje et al. (2016) Bandura (1999).

Media attention and public narratives can magnify or reshape a violent incident's social meaning. The relationship among attention, perceived threat, political emotion, and policy preferences is complex and potentially reciprocal. A comparison of messages, audiences, and institutional contexts is required before attributing changes in opinion or behavior to a particular event. Researchers should protect victims from being reduced to explanatory variables and should avoid speculative statements about motives that have not been independently established McCauley & Moskalenko (2008) Voelkel et al. (2024).

At the institutional level, responses to violence can strengthen trust or deepen perceptions of unequal treatment. This creates possible feedback loops between policy, group boundary formation, and subsequent political conduct. Measuring these dynamics requires before-and-after evidence, comparable populations, and a clear separation between observed effects and normative judgments about policies. Political psychology contributes through the study of appraisal, trust, and intergroup relations, while legal scholarship and conflict studies provide additional necessary context Wolfowicz et al. (2021) Borum (2011b).

Scientific Synthesis: A Multilevel Account of Escalation and Restraint

The evidence supports organizing the phenomenon around five linked analytic questions. What is the grievance or valued political objective? How is it interpreted through identity and moral judgment? What social networks or institutions supply reinforcement and opportunities? Which behaviors are legitimated, constrained, or actually performed? What occurs over time through persistence, withdrawal, or transition into other forms of participation? These questions capture the scope of the field without assuming that every person travels through a fixed sequence McCauley & Moskalenko (2008) Kruglanski et al. (2014) Horgan (2008).

The largest methodological distinction is between explanatory correlates and actionable prediction. Individual-difference research contributes testable moderators, while meta-analyses identify patterns of association. Behavioral case studies illuminate temporal complexity but depend on selective records. Intervention experiments offer comparatively strong evidence about immediate measured outcomes, though long-term effects on serious violence remain hard to establish. Risk tools offer structured professional reasoning but currently lack decisive prospective validation. A scientifically integrated account preserves these differences in evidentiary authority Obaidi et al. (2026) Gill et al. (2014) Voelkel et al. (2024) Brouillette-Alarie et al. (2025).

Future theory should specify when moral conviction supports restraint, when significance needs find peaceful routes, when group ties protect against or facilitate harmful involvement, and how institutional actions modify the available choices. The discipline benefits from treating nonviolent radical change as part of the comparison space. This approach improves construct validity, supports more discriminating empirical tests, and aligns public safety with accurate scientific classification Borum (2011a) Wolfowicz et al. (2021).

Conclusion

Political radicalization, extremism, and violence are best understood as historically situated processes involving political meaning, social identification, emotional appraisal, moral evaluation, relational influence, opportunity, and behavior. Foundational theories illuminate important mechanisms, but no single psychological profile or universal escalation sequence is supported across the evidence base. The adoption of radical ideas, endorsement of violence, intention to act, commission of an offense, and subsequent disengagement are different outcomes that demand different evidence Horgan (2008) McCauley & Moskalenko (2008) Obaidi et al. (2026).

The strongest empirical contributions include behavioral case comparisons, systematic reviews differentiating attitudes from conduct, and experiments showing that shifts in intergroup animosity do not automatically translate into changes in support for political violence. The field's main limitations remain selective sampling, heterogeneous definitions, retrospective inference, limited longitudinal evidence, and weak prospective validation of risk instruments. These limitations are integral to the interpretation of findings rather than marginal qualifications Gill et al. (2014) Wolfowicz et al. (2021) Voelkel et al. (2024) Brouillette-Alarie et al. (2025).

A cumulative psychology of political conflict should integrate individual and relational explanations with political institutions, historical trajectories, and the changing digital environments of the Artificial Era. Its standards are clear construct definitions, directly observed outcomes where possible, causal modesty proportionate to design, protection of lawful political expression, and transparent independent testing. The research aim is to explain why political commitments take such divergent forms and how contexts of coercion and peaceful participation develop over time Doosje et al. (2016) Schuurman (2020).

References

  • Atran, S., Sheikh, H., & Gómez, Á. (2014). Devoted actors sacrifice for close comrades and sacred cause. Proceedings of the National Academy of Sciences, 111(50), 17702–17703. https://doi.org/10.1073/pnas.1420474111
  • 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
  • Borum, R. (2011). Radicalization into violent extremism I: A review of social science theories. Journal of Strategic Security, 4(4), 7–36. https://doi.org/10.5038/1944-0472.4.4.1
  • Borum, R. (2011). Radicalization into violent extremism II: A review of conceptual models and empirical research. Journal of Strategic Security, 4(4), 37–62. https://doi.org/10.5038/1944-0472.4.4.2
  • Brouillette-Alarie, S., Hassan, G., Varela, W., Danis, E., Ousman, S., Madriaza, P., Pauls, I. L., Kilinc, D., Pickup, D., Pelzer, R., & Borokhovski, E. (2025). Reliability and validity of risk assessment tools for violent extremism: A systematic review. Campbell Systematic Reviews, 21(4), e70080. https://doi.org/10.1002/cl2.70080
  • Bubolz, B. F., & Simi, P. (2015). Leaving the world of hate: Life-course transitions and self-change. American Behavioral Scientist, 59(12), 1588–1608. https://doi.org/10.1177/0002764215588814
  • Corner, E., & Gill, P. (2015). A false dichotomy? Mental illness and lone-actor terrorism. Law and Human Behavior, 39(1), 23–34. https://doi.org/10.1037/lhb0000102
  • Corner, E., Gill, P., Schouten, R., & Farnham, F. (2018). Mental disorders, personality traits, and grievance-fueled targeted violence: The evidence base and implications for research and practice. Journal of Personality Assessment, 100(5), 459–470. https://doi.org/10.1080/00223891.2018.1475392
  • Doosje, B., Moghaddam, F. M., Kruglanski, A. W., de Wolf, A., Mann, L., & Feddes, A. R. (2016). Terrorism, radicalization and de-radicalization. Current Opinion in Psychology, 11, 79–84. https://doi.org/10.1016/j.copsyc.2016.06.008
  • Ebbrecht, C. K. (2023). Systematic review: Risk factors and mechanisms of radicalization in lone-actor grievance-fueled violence. Nordic Psychology, 75(2), 150–184. https://doi.org/10.1080/19012276.2022.2074524
  • Gill, P., Horgan, J., & Deckert, P. (2014). Bombing alone: Tracing the motivations and antecedent behaviors of lone-actor terrorists. Journal of Forensic Sciences, 59(2), 425–435. https://doi.org/10.1111/1556-4029.12312
  • Horgan, J. (2008). From profiles to pathways and roots to routes: Perspectives from psychology on radicalization into terrorism. The Annals of the American Academy of Political and Social Science, 618(1), 80–94. https://doi.org/10.1177/0002716208317539
  • Kruglanski, A. W., Gelfand, M. J., Bélanger, J. J., Sheveland, A., Hetiarachchi, M., & Gunaratna, R. (2014). The psychology of radicalization and deradicalization: How significance quest impacts violent extremism. Political Psychology, 35(S1), 69–93. https://doi.org/10.1111/pops.12163
  • McCauley, C., & Moskalenko, S. (2008). Mechanisms of political radicalization: Pathways toward terrorism. Terrorism and Political Violence, 20(3), 415–433. https://doi.org/10.1080/09546550802073367
  • Obaidi, M., Bergh, R., Benningstad, N. C. G., & Dovidio, J. F. (2026). Individual differences in violent extremism. Nature Reviews Psychology, 5, 29–46. First published online November 17, 2025. https://doi.org/10.1038/s44159-025-00509-y
  • Sageman, M. (2008). Leaderless jihad: Terror networks in the twenty-first century. University of Pennsylvania Press. https://www.pennpress.org/9780812206784/leaderless-jihad/
  • Sarma, K. M., Carthy, S. L., & Cox, K. M. (2022). Mental disorder, psychological problems and terrorist behaviour: A systematic review and meta-analysis. Campbell Systematic Reviews, 18, e1268. https://doi.org/10.1002/cl2.1268
  • Schuurman, B. (2020). Research on terrorism, 2007–2016: A review of data, methods, and authorship. Terrorism and Political Violence, 32(5), 1011–1026. https://doi.org/10.1080/09546553.2018.1439023
  • Sheikh, H., Atran, S., Ginges, J., Wilson, L., Obeid, N., & Davis, R. (2014). The devoted actor as parochial altruist: Sectarian morality, identity fusion, and support for costly sacrifices. Cliodynamics, 5(1), 23–40. https://doi.org/10.21237/C7clio5124901
  • Voelkel, J. G., Stagnaro, M. N., Chu, J. Y., Pink, S. L., Mernyk, J. S., et al. (2024). Megastudy testing 25 treatments to reduce antidemocratic attitudes and partisan animosity. Science, 386(6719), eadh4764. https://doi.org/10.1126/science.adh4764
  • Wolfowicz, M., Litmanovitz, Y., Weisburd, D., & Hasisi, B. (2020). A field-wide systematic review and meta-analysis of putative risk and protective factors for radicalization outcomes. Journal of Quantitative Criminology, 36(3), 407–447. https://doi.org/10.1007/s10940-019-09439-4
  • Wolfowicz, M., Litmanovitz, Y., Weisburd, D., & Hasisi, B. (2021). Cognitive and behavioral radicalization: A systematic review of the putative risk and protective factors. Campbell Systematic Reviews, 17(3), e1174. https://doi.org/10.1002/cl2.1174