RESEARCH DOSSIER
Restorative Justice and Victim–Offender Processes
Abstract. Restorative justice comprises structured, consent-based practices in which people affected by wrongdoing can participate in understanding harm, negotiating accountability, and pursuing repair. This fundamental research monograph examines victim–offender mediation, dialogue, conferencing, and related criminal-justice processes as psychological, institutional, and evaluative phenomena. It reconstructs the historical development of restorative ideas and distinguishes normative principles and hypothesized psychological mechanisms from experimentally tested outcomes. The study synthesizes original psychological experiments, randomized conferencing evaluations, contemporary meta-analyses, qualitative research, and international safeguards. Randomized conferencing studies support improved victim experiences and modest average reductions in some reoffending measures among selected consenting participants; broader program syntheses find smaller and more heterogeneous behavioral effects. Evidence for sexual violence and intimate partner violence remains particularly limited with respect to long-term victim well-being, while coercion and secondary victimization are serious implementation concerns. This critical integrative review organizes operational definitions, causal inference, program fidelity, and research priorities; it is neither an original experiment nor a registered systematic review.
Keywords: restorative justice; victim–offender mediation; victim–offender dialogue; restorative conferencing; procedural justice; victim satisfaction; informed consent; restitution; reintegrative shaming; accountability; repeat offending; victim safety; posttraumatic symptoms; sexual violence; intimate partner violence; forensic psychology; crime victimization; program evaluation; causal inference; Artificial Era
Introduction: The Scientific Problem of Restorative Justice
Restorative justice refers to a family of practices through which people affected by wrongdoing may participate in identifying the harm, acknowledging responsibility, considering obligations, and pursuing repair. In criminal justice, victim–offender processes may include direct dialogue, indirect communication, structured mediation, conferencing with supporters, and community-based circles. These processes involve an identifiable harmful event and people with unequal roles in relation to that event. The term “restorative” describes an aim, whereas its psychological and legal achievements are questions for research. An opportunity to speak can improve a person's experience of justice without repairing material loss; a completed restitution agreement can restore resources without repairing trust; an offender can acknowledge responsibility without being reconciled with the victim. These outcomes must be investigated independently. UNODC (2020) United Nations Office on Drugs and Crime. (2020). Handbook on restorative justice programmes (2nd ed.). United Nations. (2020) sets out this range of program models.
The principal scientific questions concern participant agency, institutional legitimacy, emotional responses, social influence, accountability, agreement implementation, and subsequent harm. The evidence base spans laboratory research on reconciliation needs, criminological theories of reintegration, naturalistic program evaluation, randomized conferencing experiments, and systematic research syntheses. Their findings answer different questions. An experimental effect on recorded reconviction, for example, cannot establish the subjective quality of a victim's experience. A favorable interview with a participating victim cannot establish that the program reduces offending. The review therefore evaluates the match between construct, method, and conclusion.
The present monograph is a critical integrative research synthesis. It surveys identifiable foundational scholarship and the major controlled and comparative evaluations, analyzes limits of evidence, and develops an explicit account of what can presently be inferred. It is not an original participant experiment, a registered systematic review, or a comparative legal manual. Its central conclusion is conditional: carefully facilitated conferencing among consenting participants has randomized evidence for improved victim-reported outcomes and a modest average reduction in some measures of repeat offending; effects across the broader, heterogeneous category of restorative programs are more variable, and evidence for high-risk interpersonal violence demands separate scrutiny. Strang et al. (2013) Strang (2013) and Fulham et al. (2025) Fulham provide distinct foundations for that assessment.
Research Object, Definitions, and Boundaries
A restorative process* is a facilitated sequence of decisions, communications, and potential agreements involving those affected by an offense. A restorative program* is an institutional arrangement defining eligibility, referrals, safeguards, facilitator roles, documentation, legal interfaces, and follow-up. The program and process have different units of analysis. Two participants may report different experiences of the same conference; two conferences may be delivered under markedly different program rules. The terms victim–offender mediation* and victim–offender dialogue* often emphasize communication between directly affected persons, whereas conferencing* generally admits supporters, family, and other stakeholders. A circle* can distribute participation more widely. None of these terms by itself specifies consent quality, whether parties meet face to face, the criminal justice stage, or what legal effect an agreement receives. The Council of Europe (2018) Council of Europe (2018) recommendation expressly recognizes multiple restorative formats.
The core constructs require operational definitions. Victim participation comprises invitation, adequate information, capacity to choose, opportunity to decline, control over disclosures, and ability to withdraw. Accountability comprises acknowledgment of relevant conduct, recognition of consequences, specific reparative commitments, and follow-through; it is not identical with displayed shame or an apology. Repair may concern monetary restitution, replacement of property, practical assistance, information, recognition, boundaries for future contact, or community obligations. Reconciliation is a potential relational development, while forgiveness is an individual response that cannot be required. Victim satisfaction, procedural justice, perceived safety, psychological symptoms, restitution completion, and repeat offending are separate outcome families.
The forensic boundary is important. A restorative program can operate before charges, in diversion, following adjudication, during imprisonment, or after release. At each stage different rules govern admissions, legal advice, confidentiality, enforceability, sentencing consequences, and official discretion. Legal jurisdiction and institutional arrangements consequently form part of the intervention, not merely its background. The UN Economic and Social Council (2002) United Nations Economic and Social Council. (2002). Basic principles on the use of restorative justice programmes in criminal matters (Resolution 2002/12). United Nations. (2002) principles and EU Directive 2012/29/EU European Parliament & Council of the European Union. (2012). Directive 2012/29/EU establishing minimum standards on the rights (2012) establish internationally influential safeguards while leaving actual domestic implementation to competent institutions. Scientific inference requires recording such institutional differences.
Historical Formation and International Development
Modern restorative practices were formed through overlapping trajectories: victim-oriented responses to crime, criminal restitution initiatives, community mediation, youth justice reform, and criticism of conventional state-controlled processing of offenses. North American victim–offender reconciliation and mediation experiments were influential in the late twentieth century. Family group conferencing developed through different legal and social institutions in Australasia. The resulting transnational history involves contributions from community organizers, victim advocates, legal scholars, criminologists, and Indigenous practitioners whose traditions must be represented in their specific historical and political contexts. A single narrative of universal cultural origins misrepresents that diversity.
John Braithwaite's Crime, Shame and Reintegration* (1989) Braithwaite (1989) provided a criminological explanation for the possibility of condemning harmful conduct while sustaining pathways to responsible social membership. Heather Strang (2002) Strang (2002) placed victims' experiences at the center of empirical assessment, while Kathleen Daly (2002) Daly (2002) interrogated idealized accounts of restorative justice, including claims about cultural origins and straightforward psychological transformation. Their different contributions illustrate the relation among a theoretical model, victim-centered outcome research, and the critical historiography of practice. Historical importance does not automatically certify an empirical mechanism.
The institutional language became more standardized through the United Nations Basic Principles (2002) United Nations Economic and Social Council. (2002). Basic principles on the use of restorative justice programmes in criminal matters (Resolution 2002/12). United Nations. (2002), Directive 2012/29/EU European Parliament & Council of the European Union. (2012). Directive 2012/29/EU establishing minimum standards on the rights (2012), Council of Europe Recommendation CM/Rec(2018)8 Council of Europe (2018), and the UNODC Handbook (2020) United Nations Office on Drugs and Crime. (2020). Handbook on restorative justice programmes (2nd ed.). United Nations. (2020). These instruments provide procedural concepts and safeguards. Their adoption cannot be used as evidence that an intervention reduces violence or trauma. The international growth of the term “restorative justice” creates a special scientific obligation to distinguish institutions with meaningful participant-centered practices from nominally restorative activities that offer few actual choices.
Theoretical Foundations: Harm, Responsibility, and Repair
Restorative theory begins with the recognition that a wrongful act creates consequences for particular people, communities, and institutions. Formal legal decisions may classify the offense and impose a sanction while leaving the victim's questions and practical needs partially unanswered. A restorative response can give those questions a recognized place: what happened, why, who was affected, what responsibility follows, and what further safeguards are required. This structure is normatively attractive because it connects responsibility with identified harms. Whether it achieves its aims depends on selection, preparation, interaction, resources, and implementation. Strang (2002) Strang (2002) demonstrates why victims' actual reported interests must guide assessment.
Reintegrative shaming theory proposes that an explicit denunciation of misconduct may coexist with respect for the person's continuing membership in a moral community. Braithwaite (1989) Braithwaite (1989) interpreted this combination as potentially consequential for crime control. Its psychological plausibility rests on distinctions between disapproval of conduct, stigmatizing exclusion of the person, social support, and credible obligations for change. The same conference can, however, produce public humiliation, defensive reactions, or pressure to perform regret. Observed emotion is therefore not a validated indicator of durable accountability. Researchers require operationalized evidence of acknowledgment and subsequent behavior.
Repair is multidimensional. Material repair can be observed in payments or replaced resources. Informational repair can be measured by whether a participant receives relevant, credible explanations. Relational repair concerns trust, boundaries, or desired contact, and institutional repair concerns access to a process judged respectful and legitimate. These dimensions may move in different directions. A victim who receives restitution might remain fearful; a victim who obtains a respectful explanation may elect never to see the person again. Daly (2002) Daly (2002) identifies the dangers of treating culturally attractive restorative narratives as universal descriptions of actual practice. The scientific task is to preserve outcome specificity.
Procedural Justice, Legitimacy, and Voice
Procedural justice research examines how people evaluate the process through which authorities and institutions make decisions. Perceived opportunity to speak, neutrality, respectful treatment, and trustworthy institutional motives influence these judgments. Tyler (2006) Tyler (2006) analyzes the psychological bases of legitimacy, understood in part as a judgment that institutional authority is proper and worthy of voluntary cooperation. Restorative justice can create avenues for voice that are otherwise scarce in criminal proceedings; it can also reproduce inequalities when officials or family members determine who may speak and what is permissible.
Victim voice must involve meaningful discretion. A person may need to decline direct contact, have an advocate present, limit the subjects discussed, receive interpretation, or ask for an indirect exchange. Equal time allocations alone cannot establish fairness if intimidation, disability barriers, language differences, or social status prevent genuine influence. The person responsible for harm also requires a clear understanding of process and legal consequences. Procedural fairness can be measured through the parties' own ratings, recorded facilitatory conduct, and documented opportunities for correction and withdrawal, with care not to confuse reported satisfaction with freely exercised choice.
It is possible for participants to regard a meeting as fair even when they dislike the final outcome. Conversely, an agreement may appear practically advantageous despite having been reached through pressure. Research should distinguish satisfaction with the facilitator, the meeting, the agreement, the criminal justice institution, and the long-term outcome. Fulham et al. (2025) Fulham found favorable patterns for procedural justice and satisfaction in adult program evaluations while offering a substantially more qualified account of recidivism. The contrast is not an inconsistency: these are different psychological and behavioral endpoints.
Victim Needs, Empowerment, and Consent
People affected by wrongdoing may want information, acknowledgment, protection, financial restitution, symbolic recognition, or an opportunity to describe personal consequences. Preferences can change with time, new evidence, perceived threats, and the course of legal proceedings. Victim-centered practice requires asking about those needs rather than assuming that all participants seek forgiveness or relationship restoration. Strang (2002) Strang (2002) demonstrates the importance of investigating victims' own assessments of justice, including how their priorities differ from those reflected in ordinary legal processing.
Informed consent is a continuous condition, rather than a one-time signature. A meaningful choice requires understandable information about the alternatives, the legal position of statements, confidentiality limitations, expected emotional demands, support services, the possible presence of others, and the right to withdraw. Practical constraints such as work obligations, transportation costs, economic dependence, fear of retaliation, and institutional expectations can make nominally voluntary participation coercive. The safeguards in EU Directive 2012/29/EU European Parliament & Council of the European Union. (2012). Directive 2012/29/EU establishing minimum standards on the rights (2012) are designed to protect victims from secondary and repeat victimization during restorative services. Implementation research should establish whether such safeguards are real in practice.
Agency is best measured in a longitudinal and multidimensional manner. Preconference interviews can establish information and expectations; immediate interviews can assess perceived freedom and responsiveness; later assessments can determine whether boundaries, safety, and practical commitments endured. Recruitment denominators matter: a satisfaction survey of completers tells little about all victims offered the process, particularly those who declined because of safety concerns. Respect for refusal and preservation of access to other remedies must be counted as ethically appropriate program outcomes, even though they do not produce a completed restorative meeting.
Moral Identity, Shame, and Observable Accountability
Restorative encounters bring the consequences of wrongdoing into a personal and social setting. The person responsible may hear a detailed account of injury that had previously been abstracted into charges or penalties. The encounter may encourage guilt about an action, shame about a self-image, empathy, defensive rationalization, or practical commitment. These are psychologically different phenomena, and their measurement requires care. The theoretical claim of Braithwaite (1989) Braithwaite (1989) concerns reintegrative social disapproval rather than indiscriminate public shaming. When social disapproval instead becomes stigmatizing exclusion, the hypothesized pathway changes.
A program should not equate visible remorse with valid accountability. Tearfulness, body posture, eye contact, speech fluency, and style of apology have culturally variable meanings and can be intentionally performed. More defensible process measures include a clear acknowledgment of the established facts, accurate description of consequences, willingness to answer reasonable questions, acceptance of specific obligations, and fulfillment of those obligations. Even these indicators are not interchangeable: understanding an impact does not necessarily cause durable change, and an agreement may be fulfilled for external legal incentives rather than moral transformation.
The empirical challenge is identifying whether restorative exposure changes a psychological mediator and whether that change subsequently affects offending or interpersonal conduct. Offenders willing to participate may already differ in motivation, family resources, and anticipated legal outcomes. Repeated measurements and suitable comparison conditions are needed to separate selection from change. A scientifically useful account treats reintegrative theory as an explanatory model with testable predictions, not as automatic proof that conferencing produces rehabilitation. Latimer et al. (2005) Latimer (2005) specifically cautioned about self-selection in early program evaluations.
Recognition, Reconciliation, and Social-Psychological Needs
Reconciliation research offers another potential explanation for some participant responses. In experimental interpersonal settings, Shnabel and Nadler (2008) Shnabel (2008) investigated a needs-based model in which victimization can threaten a person's power and agency, while perpetration can threaten moral identity and social acceptance. Messages addressing those differential needs increased reconciliation-related responses in their studies. This work clarifies why recognition and agency may matter psychologically; it does not establish that a criminal restorative program has the same causal effect among people who have endured violence.
A victim may experience empowerment when another person acknowledges the actual wrongdoing, provides truthful information, and respects an expressed boundary. The responsible person may find a route to participate in constructive obligations while remaining accountable. These processes are related to moral identity, trust, and social belonging. Their configuration varies by history, offense, social position, and institutional context. Laboratory conflict scenarios are particularly limited when applied to episodes involving coercive control, ongoing violence, and legal dependence, which differ from ordinary interpersonal disagreements in seriousness and persistence.
Reconciliation requires a choice about a relationship and may be inappropriate or unwanted. A person can seek acknowledgment and restitution without seeking reunion, friendship, restored intimacy, or forgiveness. Likewise, a victim can reject a proposed agreement while appreciating being treated respectfully. Researchers should avoid designing outcome measures that interpret continuing separation as program failure. The appropriate outcome is whether participation preserved agency and met the person's relevant goals safely, within the legal and practical possibilities of the case.
Narratives, Explanation, and Emotion Regulation
One reason a victim might consider dialogue is the wish to understand how or why an offense occurred. Sudden harm can unsettle assumptions about personal safety, trust, and predictability. A structured opportunity to question the person responsible may supply information that legal proceedings cannot readily provide. It can also intensify distress when the explanation is evasive, blaming, inconsistent, or includes disturbing new details. Consequently, psychological gain from narrative exchange is a plausible mechanism rather than a guaranteed benefit of disclosure.
Preparation may make a critical difference. Participants can identify which questions they wish to ask, what information they do not want, whether they wish to hear the offender directly, and how they would respond if an answer is unwelcome. The facilitator can arrange indirect communication or separate sessions and can end an encounter if boundaries are violated. The UNODC Handbook (2020) United Nations Office on Drugs and Crime. (2020). Handbook on restorative justice programmes (2nd ed.). United Nations. (2020) discusses the preparation, facilitation, agreement, and follow-up stages as a connected process. That sequence is part of treatment specification for evaluation.
Potential mediators include gains in perceived control, reduced uncertainty, recognition of the harm, perceived safety, and access to social support. Possible adverse mediators include increased vigilance, shame, fear, humiliation, or renewed intrusive recollection. Each requires a suitable measure and timeline. A symptom improvement observed after conferencing cannot alone establish which mechanism caused it. The short-term randomized trauma-symptom findings of Angel et al. (2014) Angel (2014) invite mechanism research without proving that conferencing is a form of trauma-focused psychotherapy.
Forgiveness, Apology, and Differentiated Trust
Forgiveness is an individual's possible response to wrongdoing and may have no place in their goals for a restorative process. Treating it as an expected outcome can pressure victims into declaring a moral resolution they do not feel. An apology has a different status: it is a communicative act that may acknowledge specific wrongs, accept responsibility, convey regret, and commit to future conduct. Its credibility is judged in context and may change with later behavior. Neither emotional expression nor an apology provides independent evidence that the offender is safe or will desist.
The experimental conferencing evidence summarized by Strang et al. (2013) Strang (2013) includes outcomes related to receipt of apologies and victims' judgments of sincerity. These are consequential aspects of participants' experience, but they should not be converted into claims about permanent personality transformation. An apology may be valuable when it accurately acknowledges harm; it may be painful or manipulative when it implies victim complicity or requests undeserved forgiveness.
Trust should be decomposed into trust in the specific offender, confidence in facilitator competence, and legitimacy attributed to the criminal justice institution. A victim can trust the facilitator and appreciate a fair process while maintaining complete distrust of the offender. This pattern may reflect a successful protective boundary rather than a failure of reconciliation. Evaluation should allow independent changes in these constructs and collect open-ended accounts of why respondents make the judgments they do. Daly (2002) Daly (2002) provides a critical corrective to narratives in which every restorative process culminates in emotional harmony.
Victim–Offender Mediation and Indirect Dialogue
Victim–offender mediation typically involves preparatory meetings, voluntary decisions about participation, a facilitated communication concerning an identified offense, and possible agreement-making. The communication may be face to face, conducted by video, transmitted through letters, or relayed by facilitators. Its design changes both opportunity and risk. Direct encounters can allow immediate questions and recognition but may expose the harmed person to intimidation or unwanted emotional demand. Indirect formats can preserve boundaries while limiting some conversational possibilities. These alternatives should be defined and analyzed separately.
The word mediation* can mislead when it suggests two equally responsible parties compromising over a mutual disagreement. In a criminal matter, the factual asymmetry of wrongdoing must remain visible. A facilitator's procedural impartiality involves serving the legitimate process interests of all participants, while acknowledging established facts and protecting the rights of the person harmed. The UN Economic and Social Council's Basic Principles (2002) United Nations Economic and Social Council. (2002). Basic principles on the use of restorative justice programmes in criminal matters (Resolution 2002/12). United Nations. (2002) emphasize voluntariness and informed agreement, while Directive 2012/29/EU European Parliament & Council of the European Union. (2012). Directive 2012/29/EU establishing minimum standards on the rights (2012) adds specific protection of victims.
The juvenile victim–offender mediation literature is especially instructive. Bradshaw, Roseborough, and Umbreit (2006) Bradshaw (2006) synthesized 15 studies involving 9,172 juveniles across 21 U.S. service sites and reported favorable recidivism differences. The result shaped the field, but its aggregation of different programs and research designs makes it less causally conclusive than individual randomized estimates. A modern replication should register eligible cases, referral sources, reasons for rejection, program characteristics, and comparison intervention before estimating an effect.
Conferencing, Supporters, and Community Processes
A restorative conference often includes the victim, the offender, supporters of each, and a trained facilitator. Its social context differs markedly from one-to-one mediation. Supporters may offer emotional stability, help communicate difficult experiences, reinforce accountability, and provide practical assistance with repair. They may also create pressure to protect family reputation, accept an apology, or agree to unwanted obligations. The psychological meaning of additional participants depends on their power and relationship to those most affected. The randomized program family evaluated by Strang et al. (2013) Strang (2013) studied a defined face-to-face conferencing model rather than every form of collective meeting.
Research should record who invited supporters, who had veto power, whether supporters had a prior relationship to the offender or victim, who dominated speaking time, and whether concerns raised privately were honored during the encounter. A family group can help sustain later obligations through housing, employment, transportation, and social assistance. The same group can undermine agency when its members threaten exclusion or minimize the offense. Observational fidelity measures and participant-specific interviews provide complementary information about these mechanisms.
Community circles introduce further variation. Some are primarily forums for acknowledgment and relationship management; others operate in conjunction with sentencing or supervision. Claims that these models simply recover a universal premodern justice system are historically unsound. Daly (2002) Daly (2002) emphasizes the diversity and contested interpretation of restorative histories. Research involving Indigenous communities must recognize actual governance institutions, cultural authority, legal jurisdiction, and historical power relations rather than treating a generic circle format as a cultural proxy.
Process Architecture: Referral, Screening, Preparation, Agreement, Follow-Up
The intervention begins at referral. Police officers, courts, corrections staff, victim advocates, or other institutions may select cases using eligibility rules and discretionary judgments. People who never receive an offer cannot enter outcome surveys of participants. Screening should establish understanding of the incident, relevant safety concerns, ability to participate voluntarily, language and accessibility needs, and potential power imbalances. Preparation can then clarify goals, discuss legal implications, choose direct or indirect contact, and set stopping rules. The UNODC Handbook (2020) United Nations Office on Drugs and Crime. (2020). Handbook on restorative justice programmes (2nd ed.). United Nations. (2020) supplies a practice-oriented framework for these phases.
A conference may end without an agreement, with a partial agreement, or with concrete commitments subject to later monitoring. All outcomes belong in a credible study. Agreement quality includes specificity, feasibility, voluntariness, proportionality, and the relevance of terms to the actual harm. An agreement that cannot be implemented can create renewed disappointment or additional sanctions. Follow-up should determine whether obligations were fulfilled, whether participants remained safe, and whether the process had unintended consequences.
Program fidelity cannot be inferred from training certificates alone. Independent observation, coded session records, facilitator supervision, documented protocol deviations, and participant complaints can help determine whether key safeguards were delivered. A program whose success metric counts only completed conferences has an incentive to enroll unsuitable cases or pressure reluctant participants. Better metrics include the proportion offered a meaningful choice, the number declining without penalty, the quality of safety planning, adherence to confidentiality rules, agreement feasibility, and longer-term adverse events.
Research Designs, Causal Estimands, and Selection Bias
The evaluation of restorative justice requires a precise causal question. Offering eligible people a conference, assigning consenting participants to conferencing, and completing a conference are different interventions. Randomized assignment among people who have already agreed to a possible meeting estimates an intention-to-treat effect in that selected population. It does not estimate the consequences of compelling reluctant victims or offenders to attend, nor does it identify effects for people excluded by screening. The distinction is essential because programs depend on consent, and consent itself may correlate with future behavior and favorable experience.
Observational comparisons are affected by self-selection, official referral criteria, family support, case seriousness, offender motivation, and the perceived consequences of participation. A participant who voluntarily enters mediation may already have lower baseline risk, better resources for compliance, or stronger incentives to cooperate. The meta-analysis of Latimer, Dowden, and Muise (2005) Latimer (2005) emphasized this issue when interpreting encouraging results. Statistical adjustment can reduce confounding for observed variables but cannot automatically eliminate hidden differences in motivation and circumstances.
Researchers should report denominators for all referred cases, all eligible cases, all consenting individuals, all randomized assignments, actual attendance, agreement completion, and follow-up participation. They should specify the primary endpoint, follow-up window, unit of assignment, analytic strategy, handling of cluster effects, and losses to follow-up. Where an outcome is an official reconviction, the result is partly dependent on detection, charging, legal procedure, and exposure to monitoring. Where the outcome is self-reported satisfaction, social desirability and attrition may matter. This separation of mechanisms of observation is as important as sophisticated statistical analysis.
Randomized Evidence on Restorative Conferencing
The Campbell review by Strang, Sherman, Mayo-Wilson, Woods, and Ariel (2013) Strang (2013) assessed ten eligible randomized trials of face-to-face restorative justice conferencing. The trials involved approximately 1,879 offenders and a victim-outcome sample of 734 interviewed victims, across different countries and offenses. Participants were willing to take part in a potential encounter before random assignment. The review found a modest average reduction in repeat offending and favorable victim-related outcomes, including satisfaction and receipt of apologies. These are unusually valuable comparisons because randomized assignment improved the credibility of the counterfactual.
The related analysis by Sherman and colleagues (2015) Sherman synthesized the controlled evidence on repeat offending and discussed the economics of conferencing in the United Kingdom trial subset. Its importance lies in testing an identifiable program model instead of aggregating every possible practice labeled restorative. The studies do not imply that a brief discussion is sufficient to resolve entrenched criminal behavior. Their findings depend on structured preparation, participant willingness, facilitator training, institutional context, and follow-up definitions. The same trainer's involvement across the set of Campbell trials promotes procedural coherence but may narrow generalizability.
The correct conclusion is an average effect for a selected consenting population exposed to a defined form of conferencing. It does not follow that every offender will desist or every victim will benefit. Rare adverse outcomes and differences among offense categories may be invisible in mean estimates. The confidence warranted by experimental assignment therefore coexists with limits to transportability. Later program evaluation should preserve the experimental design's strengths while recruiting across diverse institutions and including independent measurement of harms and participant autonomy.
Program-Wide Meta-Analyses and the Limits of Recidivism Claims
The early meta-analysis of Latimer et al. (2005) Latimer (2005) reported favorable outcomes in victim and offender satisfaction, restitution compliance, and repeat offending across heterogeneous restorative practices. Its contribution was to bring otherwise scattered evaluations into comparative view; its principal limitation was the uncertainty introduced by voluntary selection and variability in interventions. Bradshaw et al. (2006) Bradshaw (2006) later synthesized juvenile mediation programs and likewise found favorable recidivism associations. Neither result can substitute for a carefully controlled test of one named procedure.
A more recent adult meta-analysis by Fulham, Blais, Rugge, and Schultheis (2025; online 2023) Fulham included 27 studies contributing 34 samples. It found small favorable effects for general recidivism, no statistically significant overall reduction in violent recidivism, and more promising results for satisfaction, procedural justice, and accountability. The authors' attention to program and methodological moderators clarifies why the label “restorative justice” is insufficient to predict an effect. Offense mix, program fidelity, legal stage, population, and study design can change both estimates and their interpretation.
There is no contradiction between a favorable randomized trial synthesis of face-to-face conferencing and a weaker, more heterogeneous aggregate of diverse programs. They estimate different quantities. Evidence should be assessed by the intervention, the participants, the comparison condition, the endpoint, and the study design. Claims that restorative practices either categorically prevent crime or categorically fail ignore this essential structure. The strongest conclusion is that selected conferencing models have a causal evidence base for some outcomes, while broader and high-risk applications require more rigorous independent tests.
Victim Satisfaction, Perceived Fairness, and Recognition
Victim satisfaction is both important and methodologically complex. A person may be satisfied with the facilitator, the conversation, the acknowledgment of wrongdoing, the justice-system response, or the practical agreement. These forms of satisfaction should be queried separately. The experimental studies synthesized in Strang et al. (2013) Strang (2013) reported more favorable victim evaluations among conference-assigned participants, including apology outcomes. This demonstrates a pattern of reported experience under study conditions, not universal psychological recovery.
A major challenge is whose experience is recorded. Victims who are unwilling or unable to participate may be absent from post-conference samples, even though their exclusion shapes the ethical and social impact of the program. Individuals who report distress may also discontinue surveys. Facilitator-administered questionnaires can create pressure to express gratitude, particularly if the facilitator has provided substantial emotional support. Evaluation should therefore employ independent interviewers, explicitly document refusals, report response rates and uncertainty, and include qualitative accounts of mixed or negative experiences.
Fairness is an independent outcome. One participant may believe the process was respectful while concluding that the restitution offered was inadequate; another may feel pressured into a satisfactory financial settlement. Tyler (2006) Tyler (2006) provides a conceptual framework for analyzing the perceived legitimacy of the procedure, and Fulham et al. (2025) Fulham reports favorable procedural perceptions in its broader program synthesis. A scientifically sound outcome battery must preserve differences between voice, satisfaction, recognition, perceived safety, and material repair.
Offender Desistance, Accountability, and Behavioral Change
Repeated offending is influenced by developmental stage, prior offending, social relationships, employment, housing, substance use, financial pressure, monitoring, and opportunities for law-abiding participation. A restorative encounter may influence empathy, responsibility, identity, or support networks; it also commonly occurs within broader legal and social service systems. If an individual subsequently desists, attributing all change to the conference overlooks these parallel causes. Braithwaite (1989) Braithwaite (1989) supplies a theoretical link between reintegration and behavior, but a strong mechanism claim requires direct longitudinal testing.
Behavioral outcomes should distinguish binary reconviction, frequency of incidents, severity, violent offending, self-report, and administrative contact. A diversion scheme may alter recorded charges without changing behavior. Different follow-up windows may capture short-term gains or later convergence. Agreement completion, restitution, improved institutional cooperation, and recidivism should be analyzed as separate variables. The favorable aggregate findings in Sherman et al. (2015) Sherman and the more heterogeneous results in Fulham et al. (2025) Fulham are best understood through this outcome-specific framework.
The causal pathway from a meeting to desistance may involve social opportunities more than emotional intensity. A participant may make a specific commitment because supporters offer employment assistance or practical supervision; another may express intense regret yet return to an environment offering few viable alternatives. A program can meaningfully improve victim experience without producing a demonstrable crime reduction, and a crime-reduction program may function while being experienced as disrespectful. These are distinct grounds for evaluating effectiveness and justice.
Restitution, Cost, and Wider Social Outcomes
Restitution is measurable through promised amounts, actual payments, services performed, delays, and participants' assessments of adequacy. The ability to pay is unequally distributed, and imposing an unrealistic obligation can frustrate victims or create cascading legal consequences. A remedy that meets one victim's priorities may be immaterial to another victim seeking safety or an explanation. Latimer et al. (2005) Latimer (2005) included restitution-related outcomes in its synthesis, illustrating why they should be tracked independently of satisfaction and repeat offending.
Economic evaluation requires a stated perspective, time horizon, local costs, and a transparent valuation of outcomes. Implementation costs include trained facilitation, screening, interpretation, participant travel, accessibility provisions, safety planning, monitoring, and the resources supplied by support services. Sherman et al. (2015) Sherman discussed benefit-to-cost estimates from seven United Kingdom randomized experiments; those figures depend on local crime costs and program delivery. Their direction may be informative, but numerical transfer across jurisdictions without recalculation is unsound.
Distributional analysis is also essential. Institutional savings can be accompanied by increased emotional labor for victims or unpaid work for family members. A program may reach well-resourced communities while excluding individuals with complex needs. Long-run social value includes accessibility, safety, legitimacy, and fair allocation of both benefits and burdens. An economic calculation should reflect uncertainty around crime prevention and avoid valuing participant well-being only when it translates directly into criminal justice budget reductions.
Juvenile Justice and Developmental Participation
Juvenile restorative processes take place under distinctive developmental conditions. Adolescents may differ in impulse control, future orientation, understanding of legal consequences, sensitivity to peer evaluation, and dependence on adult authority. Their willingness to accept a proposal may therefore reflect the expectations of parents, teachers, police, or court personnel rather than independent comprehension. The same conference structure may have different social and psychological effects at different ages, making developmental accessibility a matter of both ethics and scientific validity.
The juvenile mediation meta-analysis by Bradshaw, Roseborough, and Umbreit (2006) Bradshaw (2006) found favorable comparative recidivism patterns but combined programs across sites and designs. Such an aggregate does not establish that all youth interventions have a common causal mechanism. Appropriate endpoints extend beyond reconviction to educational participation, relationships, restitution feasibility, access to support, and subsequent justice-system exposure. Diversion can change recorded legal outcomes even if underlying behavior is unchanged.
A developmental program requires age-appropriate explanations, independent safeguards where required, and a meaningful opportunity to refuse. An adolescent without income should not be offered obligations that depend on unrealistic financial resources; a child should not be asked to carry a family's reputational burden. Supporters can protect a young participant's comprehension while also exerting pressure. Studies should record who made key decisions and whether the child's account of consent differs from adults' descriptions. The UN Basic Principles (2002) United Nations Economic and Social Council. (2002). Basic principles on the use of restorative justice programmes in criminal matters (Resolution 2002/12). United Nations. (2002) provide general procedural expectations that must be adapted to the applicable juvenile legal framework.
Sexual Violence and the RESTORE Program
Restorative interventions for sexual violence raise questions involving survivor autonomy, credibility, sexual harm, institutional betrayal, fear of retaliation, and evidence preservation. Some survivors seek a setting in which responsibility can be acknowledged outside adversarial proceedings; others perceive participation as an unacceptable intrusion or risk. Neither response is evidence of the survivor's psychological health or moral standing. A rigorous program must protect the option of nonparticipation and must prevent pressure to forgive, reconcile, accept a minimized account of conduct, or withdraw from legal remedies.
Mary P. Koss (2014) Koss reported on the specially designed RESTORE restorative justice program for selected sexual assault cases. The evaluation described a small set of 22 cases, with structured preparation, carefully monitored conferencing and agreements, and a follow-up procedure. The study contributed evidence concerning feasibility and participant experiences in an unusually sensitive setting. Its nonrandomized, selected sample does not provide a population-level estimate of comparative effectiveness or establish safety across all forms of sexual violence.
The systematic review and meta-analysis by Kettrey and Reynolds (2026; online 2024) Kettrey identified four eligible studies from five reports, encompassing 692 participants, on restorative approaches in sexual assault and domestic violence. A favorable aggregate recidivism estimate did not remain statistically significant under restriction to randomized designs. The eligible studies did not report quantitative victim well-being outcomes. That absence is not a minor measurement omission: survivor outcomes are central to evaluating whether such practices achieve their stated goals. Future trials require ethical safeguards and independent long-term victim assessments before strong effectiveness claims can be sustained.
Intimate Partner Violence, Coercive Control, and Continuing Risk
Intimate partner violence may involve continuing coercive control through threats, economic dependence, isolation, shared parenting, surveillance, and reputational pressure. The offense cannot always be treated as a completed isolated incident. A seemingly consensual meeting may create a new opportunity for intimidation or allow the abusive participant to collect private information. Participants may report willingness because refusal carries anticipated consequences outside the program. This creates a gap between recorded consent and effective freedom, and it makes ordinary conference eligibility criteria insufficient.
Hudson and Berger (2026) Hudson (2026) synthesized qualitative evidence on perceived benefits, risks, and implementation practices in intimate partner violence applications. Their findings emphasize the significance of victim agency and acknowledge dangers including manipulation, retraumatization, and inadequate facilitator preparation. These qualitative accounts identify plausible and observed concerns but cannot themselves estimate the causal effect of the intervention on long-term victim safety. The restricted empirical base reported by Kettrey and Reynolds (2026) Kettrey underscores the need for stronger comparative outcome data.
Safety screening must examine coercive dynamics, current threats, retaliation risk, information security, dependency, available protection, and the consequences of refusal. Independent victim services and legal routes must remain accessible. Facilitators need specialized training and authority to stop the process, with explicit protocols for disclosure and follow-up. The Council of Europe (2018) Council of Europe (2018) and EU victims' rights directive (2012) European Parliament & Council of the European Union. (2012). Directive 2012/29/EU establishing minimum standards on the rights (2012) establish relevant principles, including free and informed participation and protection from secondary victimization. Where coercion defeats meaningful consent, declining to convene a meeting is consistent with responsible restorative practice.
Power, Culture, Disability, and Equal Access
Restorative justice programs operate inside institutions and communities marked by unequal resources and authority. Income, language, race, disability, migration status, age, geography, and access to legal support may influence referrals, the ability to participate, and the degree of control over outcomes. A program cannot demonstrate equity merely by treating every invited person identically. Researchers need to examine the whole referral pathway and test whether exclusion, attrition, agreements, and harms follow patterned differences across groups.
Communication conditions are especially significant. A participant may use an interpreter, alternative communication, additional processing time, or independent assistance. A person who avoids eye contact or speaks slowly may be wrongly judged insincere; someone who speaks persuasively may dominate without having a stronger claim. Such judgments create bias if facilitators treat culturally familiar emotional presentation as a measure of responsibility. Procedural justice theory Tyler (2006) Tyler (2006) helps explain why respect and voice matter, but practical accessibility requires measurable accommodations.
Community and family participation can intensify inequalities where social power is concentrated. A respected family leader may protect someone from retaliation or, alternatively, insist on silence to preserve group reputation. This dual possibility also applies to claims that restorative practices represent local tradition. Indigenous institutions have distinct histories and claims to self-determination, which should not be collapsed into a standardized state-sponsored circle. Daly (2002) Daly (2002) warns against romanticized generalizations. Comparative studies should report local governance, historical exclusion, eligibility decisions, and whose interests shape the agreement.
Rights, Confidentiality, and Legal Consequences
Restorative processes may intersect with police investigations, prosecutorial decisions, pretrial diversion, adjudication, sentencing, imprisonment, supervision, and post-release support. Legal meaning changes across stages and jurisdictions. Statements about responsibility could have evidentiary consequences; agreements may be enforceable in some systems and merely voluntary in others. Confidentiality protections have exceptions, and neither facilitators nor researchers should promise unlimited secrecy without reviewing governing law. Psychological informed consent is incomplete when legal implications remain obscure.
The UN Economic and Social Council (2002) United Nations Economic and Social Council. (2002). Basic principles on the use of restorative justice programmes in criminal matters (Resolution 2002/12). United Nations. (2002) principles address voluntary participation, basic facts, safety, and process standards. EU Directive 2012/29/EU European Parliament & Council of the European Union. (2012). Directive 2012/29/EU establishing minimum standards on the rights (2012) establishes victim-centered safeguards within its scope. The Council of Europe (2018) Council of Europe (2018) recommendation develops a broader approach to accessibility, standards, and training. These sources are normative authorities relevant to institutional design, not controlled evaluations of psychological outcomes.
A rights-respecting process identifies who has decision-making authority, what legal advice is available, how admissions are treated, and what occurs if a participant withdraws or an agreement is breached. It separates the facilitator from the roles of prosecuting, defending, judging, or providing independent clinical care. Research reports should name the jurisdiction and legal stage, because their absence prevents interpretation of diversion effects, recorded offending, and participation incentives. The legal architecture is an active component of the treatment under evaluation.
Facilitator Competence and Implementation Fidelity
Facilitators may influence whether participants feel heard, whether dominance is constrained, and whether commitments can be carried out. Competence includes clear explanation of rights and process, detecting intimidation, preparing participants, managing emotional escalation, protecting information, supporting accessibility, and making referrals. The presence of initial training cannot be accepted as proof of faithful implementation. Fidelity must be evaluated through observation, supervision, protocol documentation, participant accounts, and independent review of difficult cases.
The UNODC Handbook (2020) United Nations Office on Drugs and Crime. (2020). Handbook on restorative justice programmes (2nd ed.). United Nations. (2020) treats standards, governance, screening, preparation, implementation, and monitoring as interconnected requirements. Research can operationalize this guidance through indicators such as consent comprehension, adequacy of preparation, availability of support, independent safety screening, facilitator neutrality, participant control of contact, and follow-up completion. Each indicator can be assessed separately and linked to outcomes only when an appropriate design supports inference.
Adverse events require explicit collection rather than reliance on voluntary complaints. Examples include threatening or retaliatory contact, humiliation, pressured disclosures, confidentiality breaches, worsening symptoms, unrealistic obligations, and denial of alternative remedies. Reports should identify seriousness, timing, response, and uncertainty about causation. A high average satisfaction score cannot neutralize a serious harm suffered by a minority. The qualitative analysis of Hudson and Berger (2026) Hudson (2026) illustrates why such events deserve dedicated attention in programs involving continuing interpersonal violence.
Ethical Evaluation and Limits of Generalization
An ethical program can be valuable even when no conference occurs. Appropriate refusal, careful exclusion after screening, referral to victim services, and the preservation of independent remedies are substantive expressions of autonomy and safety. A management system that judges effectiveness solely by the number of meetings held distorts practice toward recruitment and completion. Evaluation should instead track the quality of each decision and the range of participant outcomes, including the experiences of people who did not proceed.
Generalization requires more than repeating the favorable mean from a selected trial. Effects may vary by offense, prior relationship, population, facilitator, justice system, and level of community support. Randomized evaluations of one conferencing model have strong internal validity for their assignment question but cannot support universal claims for circles, indirect dialogue, or programs designed for ongoing domestic abuse. Strang et al. (2013) Strang (2013) and Fulham et al. (2025) Fulham should therefore be read together, retaining their different designs and scopes.
There is also a distinction between an individual finding meaning in a process and a population-level estimate of its risks and benefits. Both deserve scientific attention, but they answer different questions. Case narratives can illuminate undermeasured needs, while comparative research tests common effects and exposes failures. An evidence-informed institution preserves both kinds of knowledge, reports uncertainty where it affects decisions, and gives priority to safety and valid consent in circumstances where the outcome evidence remains sparse.
Measurement Architecture and Validity of Outcome Constructs
A mature evidence base needs an explicit measurement architecture. At the participant level, relevant constructs include perceived voice, freedom of choice, psychological safety, recognition, satisfaction with separate aspects of the encounter, felt accountability, unwanted contact, and longer-term functioning. At the interaction level, relevant constructs include the distribution of speaking opportunities, facilitator interventions, apologies, threats, and agreement-making. At the institutional level, relevant variables include referral eligibility, exclusions, completion, legal consequences, restitution implementation, and administrative outcomes. Treating these levels as a single restorative “success score” conceals important divergence.
Measures must have construct validity, reliability, and defensible interpretation across populations. A one-item satisfaction rating can be useful for immediate service feedback but may be insensitive to coercion and insufficient for detecting trauma-related change. Official recidivism measures can have consistent administrative definitions and still diverge from unreported criminal behavior. A scale of posttraumatic symptoms can detect distress without proving a formal diagnosis. Studies should select measures for their own purposes, explain whether they were validated for the target population, and report missing data and uncertainty. Angel et al. (2014) Angel (2014) illustrates a symptom-specific randomized endpoint; Strang et al. (2013) Strang (2013) illustrates the different demands of victim-reported and offender outcomes.
An outcome matrix can clarify the structure without erasing meaning. At referral, one measures equitable offer and consent comprehension; after preparation, perceived safety and informed choice; during communication, adherence to boundaries and accountability; after agreement, perceived fairness and specific commitments; at follow-up, functioning, fulfilled repair, adverse events, and repeat conduct. The temporal sequence allows hypotheses about mediators, but statistical mediation should be interpreted causally only when its assumptions are defensible. An observational association between apology and later satisfaction does not establish that the apology caused the improvement.
Evidence Synthesis, Heterogeneity, and Reproducibility
The principal evidence streams differ in what they make observable. Randomized assignment improves causal inference for an eligible population. Nonrandomized program comparisons extend coverage but invite confounding. Qualitative interviews illuminate subjective experience and contextual risk but do not give a population-average intervention effect. Administrative linkage can extend follow-up but is affected by detection practices and data governance. Legal and professional guidance identifies permissible safeguards without testing their consequences. Rigorous synthesis keeps those forms of evidence distinct while identifying questions they jointly clarify.
The apparent conflict between experimental and broad meta-analytic recidivism estimates is largely an issue of population and intervention heterogeneity. Strang et al. (2013) Strang (2013) focused on selected randomized face-to-face conferences; Fulham et al. (2025) Fulham included a much wider variety of adult restorative programs and designs. Latimer et al. (2005) Latimer (2005) documents longstanding selection concerns. Meta-analysis does not erase differences in how recidivism, victim benefit, and intervention delivery are defined. Subgroup and moderator estimates need adequate sample size and independent replication before they become implementation rules.
Reproducibility requires transparent eligibility criteria, trial registration when appropriate, accessible protocols, precise intervention descriptions, clearly specified controls, and published null and adverse findings. A program must document the denominator of people to whom an intervention was offered, not only those who completed it. Studies should distinguish short-term outcomes from durable changes, report effect sizes and uncertainty, and avoid post hoc redefinition of endpoints. Prospective studies of sensitive violence contexts should have independent safety oversight and a procedure for reporting unintended harms. The restricted evidence in Kettrey and Reynolds (2026) Kettrey demonstrates the cost of leaving victim well-being unmeasured.
Restoration, Social Context, and the Artificial Era
Restorative processes exist within changing communication environments. Digital scheduling, remote meetings, accessible translation, confidential documentation, and outcome registries can reduce practical barriers, while introducing risks to privacy, authenticity, and the control of sensitive information. The relevant empirical questions concern whether remote access preserves informed consent, whether recordings alter disclosure, and whether digital systems improve accommodation without increasing surveillance or unwanted contact. Existing evidence on primarily human-facilitated conferences does not establish these effects for automated systems.
Artificial systems may support administration, translation, information access, and research auditing, but their intervention into justice processes can also shape who receives an invitation, how risk is represented, and what participants believe about legal consequences. These are high-stakes applications of algorithmic decision support. A system trained on historical records may reproduce unequal referral patterns; an apparently empathetic conversational agent can give factually inaccurate legal explanations; automated summarization may omit a qualifying condition from an agreement. Such risks are testable through controlled audits, error analysis, independent human review, and participant-centered outcome measures.
The Artificial Era · From Homo to Artificial provides a wider intellectual horizon for investigating how nonbiological systems enter institutions where human relationships, responsibility, and justice are organized. In this dossier the appropriate scientific contribution is to define observable system behavior and the consequences for human participants. There is no evidentiary foundation in the cited randomized literature for claiming that an artificial facilitator is equivalent to a trained human practitioner. The empirical research priority is to preserve provenance, clear accountability, participant autonomy, and independent oversight when technology participates in justice administration.
Future Research: Comparative Programs and Longitudinal Safety
The next generation of research should describe interventions with enough precision to permit comparison and replication. Publications should identify direct versus indirect contact, crime type, relationship between participants, preparatory work, facilitator experience, legal stage, available legal advice, presence of supporters, agreement structure, and follow-up. A standardized reporting vocabulary would permit meaningful synthesis without pretending that every restorative program delivers the same treatment. Individual or cluster-randomized designs are useful when ethically feasible and when genuine freedom to decline remains protected.
Victim-centered outcome research requires stronger longitudinal evidence. Investigators should measure agency, satisfaction, recognition, safety, perceived threat, symptom burden, functioning, practical repair, confidence in legal institutions, and the durability of chosen boundaries. Negative experiences should be solicited confidentially, not only through the program's own facilitators. Special attention is required where a participant depends economically or socially on the person who caused harm. Hudson and Berger (2026) Hudson (2026) and Kettrey and Reynolds (2026) Kettrey define the present priority: credible harm monitoring and well-being outcomes in sensitive interpersonal violence contexts.
Offender outcomes should be measured across behavior, implementation, and social integration. Frequency and severity of reoffending may convey more information than an undifferentiated binary event; economic and community conditions should be measured when testing reintegration hypotheses. Qualitative work can identify expectations and meanings absent from standardized instruments, while randomized tests can evaluate specified causal comparisons. Interdisciplinary integration with Community Psychology and Social Psychology should be guided by precisely shared constructs—participation, power, trust, norm change, group pressure—not by the loose assumption that a restorative label identifies a general psychological treatment.
Critical Scientific Discussion
Restorative justice contains three analytically separate layers. The normative layer asks what people harmed by crime are entitled to expect and what legitimate repair might involve. The explanatory layer proposes mechanisms such as recognition, agency, procedural justice, accountability, and reintegration. The empirical layer investigates actual outcomes through comparison, observation, and measurement. A statement about what a program ought to do cannot substitute for proof of what it does. Likewise, a laboratory result about reconciliation needs does not establish clinical or criminological effectiveness in a criminal justice setting.
The best-controlled conferencing evidence supports an average benefit for specific outcomes among willing participants, with special strengths in victim-reported experience and more qualified but favorable findings on measured reoffending (Strang et al., 2013 Strang (2013); Sherman et al., 2015 Sherman). Broader adult evaluations suggest smaller and more variable effects on general repeat offending and weaker evidence for violent reoffending (Fulham et al., 2025 Fulham). This pattern is explained in substantial part by differences in study design, program type, participant selection, outcome definitions, and implementation. It rules out universal claims while supporting a serious, conditional research program.
The sharpest ethical and scientific problems occur where current danger, coercion, and trauma shape consent. Selected participants may report meaningful benefits even though no robust population-level estimate exists for survivor safety or long-term well-being. Koss (2014) Koss offers exploratory program evidence; Kettrey and Reynolds (2026) Kettrey documents the scarcity of rigorous quantitative studies; Hudson and Berger (2026) Hudson (2026) identifies qualitative safety concerns. These findings together require specialized screening, competent facilitation, independent survivor support, and transparent adverse-event evaluation. The freedom to decline a meeting is a positive requirement of legitimate restoration.
Conclusion
Restorative justice and victim–offender processes are a distinctive field of psychological and forensic research because they connect the experience of harm with the social organization of responsibility and repair. Their central concepts are observable and separable: agency, procedural fairness, recognition, accountability, satisfaction, practical restitution, psychological distress, safety, and future conduct. Effective research measures these dimensions independently and locates each process within the legal and institutional conditions in which it occurs.
Randomized evidence shows that selected face-to-face conferencing programs can improve important victim-reported outcomes and modestly reduce certain measures of repeat offending. Wider program comparisons show heterogeneous and often smaller behavioral effects. The evidence for restorative intervention in sexual violence and intimate partner violence is especially limited, while the consequences of coercion, retaliation, and renewed trauma are potentially grave. A defensible scientific synthesis therefore makes consent, accessibility, independent safety oversight, and the measurement of adverse events essential features of both practice and evaluation.
The enduring research agenda is to clarify which processes benefit which people under what conditions, through which mechanisms, over what timescales, and at what cost. Restoration becomes an empirically meaningful concept when programs demonstrate specific forms of repair while protecting participants' boundaries and legal rights. The scientific value of the field lies in its capacity to examine the complex relation between justice as an institution, harm as a human experience, and accountable conduct over time.
References
- Angel, C. M., Sherman, L. W., Strang, H., Ariel, B., Bennett, S., Inkpen, N., Keane, A., & Richmond, T. S. (2014). Short-term effects of restorative justice conferences on post-traumatic stress symptoms among robbery and burglary victims: A randomized controlled trial. Journal of Experimental Criminology, 10, 291–307. https://doi.org/10.1007/s11292-014-9200-0
- Bradshaw, W., Roseborough, D., & Umbreit, M. S. (2006). The effect of victim offender mediation on juvenile offender recidivism: A meta-analysis. Conflict Resolution Quarterly, 24(1), 87–98. https://doi.org/10.1002/crq.159
- Braithwaite, J. (1989). Crime, shame and reintegration. Cambridge University Press. https://doi.org/10.1017/CBO9780511804618
- Council of Europe, Committee of Ministers. (2018). Recommendation CM/Rec(2018)8 concerning restorative justice in criminal matters (adopted October 3, 2018). Council of Europe. https://search.coe.int/cm?i=09000016808e35f3
- Daly, K. (2002). Restorative justice: The real story. Punishment & Society, 4(1), 55–79. https://doi.org/10.1177/14624740222228464
- European Parliament & Council of the European Union. (2012). Directive 2012/29/EU establishing minimum standards on the rights, support and protection of victims of crime. Official Journal of the European Union, L 315, 57–73. https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:32012L0029
- Fulham, L., Blais, J., Rugge, T., & Schultheis, E. A. (2025; first published online 2023). The effectiveness of restorative justice programs: A meta-analysis of recidivism and other relevant outcomes. Criminology & Criminal Justice, 25(5), 1486–1512. https://doi.org/10.1177/17488958231215228
- Hudson, A., & Berger, E. (2026). Restorative justice processes and intimate partner violence: A qualitative systematic review of perceived benefits, risks, and best practices for victims. Trauma, Violence, & Abuse. Advance online publication. https://doi.org/10.1177/15248380261433020
- Kettrey, H. H., & Reynolds, N. S. (2026; first published online 2024). Is restorative justice appropriate for sexual assault and domestic violence? A systematic review and meta-analysis of the 'empirical vacuum'. Journal of Experimental Criminology, 22, 229–247. https://doi.org/10.1007/s11292-024-09651-8
- Koss, M. P. (2014; first published online 2013). The RESTORE program of restorative justice for sex crimes: Vision, process, and outcomes. Journal of Interpersonal Violence, 29(9), 1623–1660. https://doi.org/10.1177/0886260513511537
- Latimer, J., Dowden, C., & Muise, D. (2005). The effectiveness of restorative justice practices: A meta-analysis. The Prison Journal, 85(2), 127–144. https://doi.org/10.1177/0032885505276969
- Sherman, L. W., Strang, H., Mayo-Wilson, E., Woods, D. J., & Ariel, B. (2015; first published online 2014). Are restorative justice conferences effective in reducing repeat offending? Findings from a Campbell systematic review. Journal of Quantitative Criminology, 31, 1–24. https://doi.org/10.1007/s10940-014-9222-9
- Shnabel, N., & Nadler, A. (2008). A needs-based model of reconciliation: Satisfying the differential emotional needs of victim and perpetrator as a key to promoting reconciliation. Journal of Personality and Social Psychology, 94(1), 116–132. https://doi.org/10.1037/0022-3514.94.1.116
- Strang, H. (2002). Repair or revenge: Victims and restorative justice. Oxford University Press. https://doi.org/10.1093/oso/9780199251643.001.0001
- Strang, H., Sherman, L. W., Mayo-Wilson, E., Woods, D. J., & Ariel, B. (2013). Restorative justice conferencing (RJC) using face-to-face meetings of offenders and victims: Effects on offender recidivism and victim satisfaction. A systematic review. Campbell Systematic Reviews, 9(1), 1–59. https://doi.org/10.4073/csr.2013.12
- Tyler, T. R. (2006). Psychological perspectives on legitimacy and legitimation. Annual Review of Psychology, 57, 375–400. https://doi.org/10.1146/annurev.psych.57.102904.190038
- United Nations Economic and Social Council. (2002). Basic principles on the use of restorative justice programmes in criminal matters (Resolution 2002/12). United Nations. https://digitallibrary.un.org/record/475864
- United Nations Office on Drugs and Crime. (2020). Handbook on restorative justice programmes (2nd ed.). United Nations. https://www.unodc.org/documents/justice-and-prison-reform/20-01146_Handbook_on_Restorative_Justice_Programmes.pdf