• 1. Introduction

  Karen,¹x We use pseudonyms throughout this article, to maintain the anonymity of participants in the study. a young woman in her twenties, was employed as a salesperson at a family-owned fashion retail shop. After nine months of employment, the shop owner, Jacob, found out that she had been stealing cash and garments for an estimated value of USD 5,200. Jacob’s wife, Sara, tried to resolve the conflict but failed. Karen’s apology and attempt to return the stolen goods were declined because of disagreement about the overall value of the theft. The conflict escalated, and the storeowner filed a criminal complaint.
  Karen’s case is one of many undertaken by the justice system every day. Like many others offenders, Karen is a first-timer; she is remorseful and cooperative with the police; she does not follow a criminal lifestyle. In thousands of programmes worldwide, restorative justice (RJ) processes provide opportunities for offenders to reflect on their actions, apologise and repair the harm caused to their victims. Victims participating in restorative justice processes are given the opportunity to tell their stories, receive answers to their questions and determine the appropriate reparation plan for their cases.
  This article reports on the findings of a study that analysed the characteristics of restorative justice conferences conducted by a community-based restorative justice programme in Jerusalem, which we used as a case study. The overall goal of the study was to assess the level of ‘restorativeness’ of community restorative justice (CRJ) conferences. To this end, we used a set of seventeen key ordinal features (characteristics), organised in four groups (Dancig-Rosenberg & Gal, 2014). The characteristics are relevant to a range of criminal justice mechanisms and relate to the process, substance, stakeholders and outcome of the mechanism under study. The methodology enables us to provide a comprehensive account of the nature of the mechanisms and to conduct comparative analyses between them. In the present study, we focused on one such mechanism, restorative justice, and examined one of its variants: CRJ conferences. Through an assessment of the procedural and substantive traits of CRJ conferences at the Jerusalem-based Mosaica restorative justice programme, we demonstrate a way to measure the restorativeness of this programme. Accordingly, the research questions are: (a) What are the traits that characterise the CRJ conferences conducted by the Mosaica programme? (b) In what ways do the Mosaica CRJ conferences reflect the restorativeness ideal, and how do they diverge from it? The findings of the study can enrich the knowledge about the restorativeness ideal and assess its realisation through specific, scale-based, measurable characteristics, and the challenges in implementing this ideal in real-world restorative justice programmes.

• 2. Theoretical background

  2.1 Restorative justice

  Restorative justice covers a wide range of practices that share an understanding of crime as an interpersonal act affecting people and relationships (Walgrave, 2013; Zehr, 1990). Programmes that fully implement the principles of restorative justice (McCold & Wachtel, 2003) provide direct encounters between victims, offenders, their supporters and community members to discuss the crime and ways to repair the harm it caused (Marshall, 1999; Zehr, 2015). Successful restorative justice processes typically lead to reparation of harm, closure and healing (Bazemore, 1998; Harris, Walgrave, & Braithwaite, 2004).
  Among the range of practices referred to as restorative justice, the best-known models are victim-offender mediation, sentencing circles and conferences (Umbreit & Armour, 2011). Community-based conferences, a variant of conferencing, include the participation of community representatives in each process, stressing the social context of crime and reflecting a perception of the community as a primary stakeholder in the aftermath of the crime (Souza & Dhami, 2008).
  A relatively large body of literature considers the effectiveness of restorative justice programmes in reducing crime (Bonta, Jesseman, Rugge, & Cormier, 2006; Latimer, Dowden, & Muise, 2005; Sherman & Strang, 2007) and, to a lesser extent, its ability to address victims’ needs and wishes (Angel et al., 2014; Poulson, 2003). Other studies have examined specific characteristics of restorative justice, which are considered crucial in its effectiveness, such as procedural justice (Shapland et al., 2006; Tyler, 2006; Tyler, Sherman, Strang, Barnes, & Woods, 2007), reintegrative shaming (Harris, 2006; Sherman, Strang, Woods, 2000) and, focusing on victims’ well-being, cognitive behavioural therapy (Angel et al., 2014). Another strand in the literature considers specific populations and how they may or may not benefit from restorative justice processes. These include battered women (Hudson, 2002; Ptacek, 2009), child victims (Gal, 2011; Gal & Moyal, 2011), families of homicide victims (Walters, 2015) and youth offenders (Crawford & Newburn, 2002). Nevertheless, knowledge is lacking about many characteristics of restorative justice processes, particularly a systematic measurement of an array of multiple traits that characterise the process.

  2.2 The restorativeness of restorative justice

  The restorative justice literature includes some standards of restorativeness (Braithwaite, 2002). On the basis of findings regarding family group conferences practised in New Zealand, Maxwell et al. (2004) found several elements that promote the restorativeness of such processes. Regarding the outcome characteristics, these have been defined on the basis of two variables: accountability (whether the sanctions are designed to be restorative or restrictive) and the enhancement of well-being (whether the sanctions are intended to be reintegrative or rehabilitative). Other variables include participation (by offenders, victims and family members), involvement (understanding the process and being able to influence the decision), consensus decision-making (joint decision-making process and avoidance of coercion) and cultural responsiveness (informal processes that facilitate cultural traditions).
  Daly (2003) listed a set of desired activities and behaviours that are expected in youth restorative justice conferences on the basis of an analysis of the South Australia Juvenile Justice (SAJJ) RJ Conferences. The conference process should bring together victims, offenders and supporters to discuss the offence, its consequences and appropriate outcomes. Conferences are expected to include a sincere apology, and all stakeholders must be treated fairly and respectfully. The discussions should be inclusive, non-hierarchical, with maximum decision-making power in the hands of private stakeholders. Outcomes should not be excessive. With respect to the legal context, conferences should be understood by the young offender and voluntary. With respect to outcomes, reparation plans should include reparative and restitutive measures. With respect to effects, victims are expected to feel better after the conference, and offenders are expected to be less inclined to recidivate.
  Presser and Van Voorhis (2002) identified three core processes that distinguish the three dominant restorative justice models (victim-offender meditation, family group conferences and circles) from other responses to crime. These are an inclusive and authentic dialogue, relationship building and the communication of moral values. These core processes can be evaluated and measured on the basis of data collected by various means (Kuo, Longmire, & Cuvelier, 2010).
  Zernova (2006) focused on four elements of the restorativeness ideal: achieving restorative rather than retributive outcomes, focusing on victims’ needs while promoting offender’s accountability, the relative voluntariness of the process and empowering private stakeholders to assert decision-making authority.
  Recently, Van Ness and Strong (2014) stressed that restorative justice processes must (a) be inclusive of all those affected by the crime, (b) involve a voluntary encounter between stakeholders to discuss the harm that was caused and the appropriate responses, (c) involve the offender accepting responsibility and repairing the harm and (d) promote reintegration rather than stigma and isolation.
  Community restorative justice involves additional model characteristics that highlight the centrality of the community as both a responsible party to the problem and a resource for its solution (Walgrave, 2013). Envisioning a restorative community justice model, White (2003) suggests four core principles: (a) social inclusion (involving victims, offenders, and communities in negotiating compensation for everybody); (b) communal objectives (extending the discussion beyond individual responsibility to address societal injustice, inequalities and needs); (c) communities of support (strengthening the social ties of offenders and victims within their communities); and (d) enhancement of community resources (fostering the creative use of existing institutions and people). Ideally, CRJ processes reflect the community ownership of the process. This can be achieved through the active involvement of community members, the use of community resources and attendance to the needs of victims and offenders to prevent future harm and promote harmony and peace (Gal, 2016; Dhami & Joy, 2007; Zehr & Mika, 2003).
  This review of the attempts to capture the restorativeness of restorative justice highlights the need to integrate them into a unified list of measurable characteristics, so that specific programmes and policymakers are able to evaluate whether restorativeness is actually achieved in practice. Indeed, some scholars have already identified a discrepancy between the restorativeness ideal and restorative justice practice. Daly (2003) found that although SAJJ conferences met the ideals of active participation of stakeholders, fair and respectful treatment by professionals and restrained penalties, four types of discrepancies emerged. The first concerns the containment of justice ideals by budgetary limitations and professional interests that limit the ability of restorative justice processes to produce fully restorative outcomes. The second is the limited ability of private stakeholders to live up to the restorative ideal because of their limited preparation and familiarity with the restorative principles. A third and related discrepancy concerns the limited ability of the direct stakeholders to behave restoratively – listen empathetically, express remorse and accept other participants’ feelings – elements that are outside the control of the professionals (Daly, 2006). The fourth discrepancy is between the hope for ‘absolute’ success in participant satisfaction, sense of fairness and restorativeness and the typical findings suggesting that restorative justice conferences are successful part of the time, for some participants, on most parameters (Daly, 2006).
  Similarly, Zernova (2006) found that the restorative ideals often remain unfulfilled, for example, the aspiration to provide a completely voluntary process for offenders, which is at the same time also victim-oriented. The family group conferencing programme on youth offenders that she analysed, which was embedded in the punitive criminal justice system, demonstrates the risk of restorative justice programmes being co-opted by the criminal justice system and its traditional goals.
  Aiming to suggest a method to ‘mind the gap’ (Daly, 2003) between the restorativeness ideal and practice, we offer here an integrated list of characteristics that, when fully adhered, can reflect a practice that is ‘fully restorative’: The process promotes victim-offender dialogue, and is flexible, non-hierarchical, inclusive and relatively voluntary; private stakeholders have decision-making power, the discussion is both victim- and offender-sensitive, there is community leadership and involvement, there is an emotional discourse that addresses the stakeholders’ needs, the process facilitates constructive dialogue and it promotes the use of community resources. In terms of the outcomes, restorativeness is achieved when the outcomes are future-oriented, restorative and rehabilitative, and promote global social justice.
  As explained in the Methodology Section, these measurable characteristics are organised as a list of scales in the analytical instrument we use. Table 2 summarises the conceptual links between the characteristics that restorative justice scholars have identified as reflecting restorativeness and the characteristics included in our analytical instrument.

• 3. Data and methodology

  3.1 Case study approach

  The investigation of a single case is particularly useful when exploring difficult-to-reach variables ‘that are yet to be evaluated empirically by researchers’ (Walters, 2015: 1212). A case study approach based on in-depth data sources enables a ‘thick description’ and information-rich’ study (Patton, 2002). Several case studies have been used to explore the application of restorative justice for various complex crimes, such as sexual violence (McGlynn, Westmarland, Godden, 2012), domestic violence (Miller & Iovanni, 2013) and homicide (Walters, 2015). These studies offered a nuanced understanding of the potential benefits and challenges of restorative justice processes for the victims of serious crime.
  To produce a rich and detailed account of the nature of CRJ, which remains under-researched within the restorative justice literature, we investigated 21 well-documented files of completed CRJ processes that have been conducted by Mosaica, a Jerusalem-based non-government organisation promoting conflict resolution and peaceful communication between multiple communities in the city. The study is based on archival content analysis of these CRJ cases to explore attributes relating to the structure, dynamics, substance and agreed-upon outcomes of CRJ conferences. Analysing the detailed documentation of each of the completed Mosaica cases expands our knowledge of the procedural and substantive traits that characterise the operation of CRJ, which have not yet been empirically explored in the literature.
  Although the case study method is limited in the general applicability of its results, it provides valuable in-depth data on the characteristics of CRJ conferences and makes tentative recommendations about the practical implementation of restorative justice ideals. In this sense, the findings of the study should be treated as a contribution to ‘the collective process of knowledge accumulation’ in the CRJ field (Flyvbjerg, 2006: 227).

  3.2 Case study description: the Mosaica programme

  Restorative justice has been applied in Israel since the 1990s, but mainly in cases involving juvenile offenders and outside the mainstream of criminal procedures (Yanay & Borowski, 2013). In 2008, Mosaica established a CRJ programme, in accordance with a model developed by the Israeli Ministry of Justice.²x The Ministry of Justice developed a detailed protocol for preparing, organising and facilitating restorative justice conferences. The documents are with the authors (Hebrew). The Mosaica CRJ programme is the only programme currently operating in Israel in accordance with the original government model, and continues to train, supervise and monitor the work of its volunteer facilitators on the basis of that model.
  Since its inception, in 2008, the programme received approximately 50 referrals from prosecutors, defence attorneys, judges, probation officers and community members requesting to conduct a CRJ conference following a criminal indictment. Of the 50 referrals, only 27 cases involved documented preparations, such as meetings with victims and offenders. The other 23 referrals were discontinued without any documentation of substantial activities, either because of stakeholders’ refusal to speak with programme representatives or because the case moved on to the legal process before any actions took place. Of the 27 documented referrals, 21 restorative justice processes were completed between 2008 and the beginning of 2015 and were, therefore, included in this study. Indeed, the six files that did not mature into full restorative justice encounters can potentially reveal the restorative nature of the preparatory phase, and future studies should examine the restorativeness of both complete and incomplete processes. But the small number of incomplete processes in this specific programme and the laconic terminology that was used to describe the preparatory steps in some of them led us to decide to focus only on the completed cases. Despite the relatively small number of completed files, Mosaica is an ideal case study because of the adherence of the programme to the original model designed by the Ministry of Justice and the extremely detailed documentation of each completed case.
  The 21 completed files were delivered to the researchers in anonymised format for content analysis. Stakeholders were informed, and did not object to the use of data for research purposes. Most cases involved indictable misdemeanours and lower-level felonies, such as assaults, threats and property crimes. Each file contains a detailed description of phone calls, preparatory meetings and correspondence with the stakeholders, the indictment that was the basis for the CRJ conference, and the reparation agreement, if one has been reached. The files also include court protocols of relevant hearings, protocols of the CRJ conferences and facilitators’ notes and reflections.

  3.3 The Mosaica restorative justice process

  In accordance with the philosophy of community-based restorative justice and the model developed by the Israeli Ministry of Justice, the Mosaica CRJ programme is guided by the following principles: voluntariness, confidentiality, honest and respectful dialogue and attentiveness. Ideally, each process involves the victim and the offender, their supporters and at least one community representative. Upon referral, two facilitators assigned to every process meet in person at least once before the conference with each participant, to ensure that they understand the process, its principles and their role in it. The preparatory phase is completed only after the facilitators have ensured that the offender accepts full responsibility for the offence and that the victim and the offender are willing to meet. The preparatory meetings are also used to agree on a mutually convenient location and timing for the conference, and on the identities of the supporters and of the community representative. Conferences typically last for several hours and conclude with an agreed-upon reparation plan, which normally includes symbolic and material reparation for the victim, the community or both. The facilitators are responsible for monitoring the implementation of the reparation plan. After its completion, they report back to the referring entity. The reparation plan becomes the basis for all subsequent legal decisions: the prosecution may withdraw the indictment, and the court may reduce the sentence or accept the reparation plan as complete replacement for punishment.
  Karen’s case, described above, provides an illustration of how each case is processed by Mosaica. In January 2013, a public defence attorney contacted the Mosaica office, talked with the programme coordinator about Karen’s case and transferred the indictment filed against her in the Jerusalem Magistrate’s court to Mosaica. Mosaica received most of its referrals after an indictment had been filed in court, although police prosecutors referred some cases after the completion of the investigation but before indictment. In Karen’s case, the defence attorney informed Mosaica that she had reached an agreement with the district police prosecutor that if a restorative justice process were successfully settled, they would withdraw the indictment. Karen naturally wanted the restorative justice process to succeed, as she was anxious about her case going to criminal court.
  After Mosaica took the case, two facilitators were assigned to it, and met Karen as a first step, according to the recommended default practice (Kirkwood, 2010). Karen came to the meeting with her psychologist as supporter. She told the facilitators that her mother had died recently after a long struggle with cancer. It was implicit that Karen had been going through an intense change since her mother’s death, which was the climax of a deterioration process in her life. Karen expressed her willingness to apologise to her employers. The next step was a preparatory meeting with the victim, Jacob, who came alone. He was angry and impatient at the meeting, telling facilitators that he suffered severe financial damage as a result of Karen’s behaviour: he had to sell his car to cover his losses. Emotionally, Karen’s action undermined his trust in people. ‘I am still licking my wounds,’ he told the facilitators, and said that he expected a restitution sum of no less than USD 13,000 (Mosaica, Casefile D, report of preparatory meeting with the victim). Despite his anger, Jacob expressed willingness to attend a restorative justice conference to be paid the amount to which he was entitled. He objected offers of being supported by his wife or anyone else at the conference, and wanted it to be quick and simple.
  After the preparations were completed, the conference took place. Karen arrived with her father, an ultra-Orthodox Jewish-American immigrant. Jacob came unescorted, and no community representative attended, following Jacob’s request. The conference began with Karen and her father stating that they knew about Jacob’s expectation for restitution and were willing to sign an agreement. Karen apologised, expressed her shame and said that there was no justification for her actions. Jacob was impatient, refused to discuss the emotional harms he had suffered and rejected Karen’s attempts to discuss her past, the death of her mother and her efforts to change her ways. ‘What he saw before him was a thief, not a remorseful young woman’ (Mosaica, Casefile D, facilitators’ minutes of the conference). But while the parties were waiting for the agreement to be printed, after he was reassured that he would receive the full amount he expected, Jacob seemed more open to listen. Encouraged by the facilitators, Karen told the participants about her past as a rebellious child who left her Orthodox family and was beset by chaos, internal conflicts and alcohol use. According to the facilitators’ report, ‘[Karen] touched Jacob’s angry, hurt, and unmoved heart. There was a magical moment in the room… [Jacob] left the room a different person’ (Mosaica, Casefile D, facilitators’ minutes of the conference).
  The reparation plan was signed by the parties. In the following months, Karen paid the sum in full, using her deceased mother’s insurance money. Her defence attorney and the police prosecutor were notified. The indictment was withdrawn, Karen was never summoned to court and her file was closed, without a criminal record. If she had been indicted, she might have been required to serve some community service, be under probation, receive a suspended sentence and bear a criminal conviction for the rest of her life. The victim may have not received the agreed-upon amount of restitution (the estimated value of the theft, according to the police report, was much lower) and Karen may have been left with some of her mother’s life insurance money to assist her in building her future.

  Table 1 Criminal Law Taxonomy (Dancig-Rosenberg & Gal, 2014 )

  Process-related parameters	1.	Lack of victim-offender dialog	0	1	2	Victim-offender dialog
  	2.	Formal	0	1	2	Flexible
  	3.	Hierarchical	0	1	2	Non-hierarchical
  	4.	Coercive	0	1	2	Voluntary
  Stakeholder-related parameters	5.	Professional-centered	0	1	2	Lay-centered
  	6.	Not victim-oriented	0	1	2	Victim-oriented
  	7.	Not offender-oriented	0	1	2	Offender-oriented
  	8.	Exclusive	0	1	2	Inclusive
  	9.	State-managed	0	1	2	Community-managed
  Substance-related parameters	10.	Rights-based terminology	0	1	2	Needs-based terminology
  	11.	No emotional discourse	0	1	2	Emotional discourse
  	12.	Process as obstacle	0	1	2	Process as vehicle
  	13.	Libertarian	0	1	2	Communitarian
  Outcome-related parameters	14.	Past-oriented	0	1	2	Future-oriented
  	15.	Retributive requital	0	1	2	Restorative requital
  	16.	Incapacitative	0	1	2	Rehabilitative
  	17.	Conflict resolution	0	1	2	Justice making

  3.4 The research instrument: criminal law taxonomy

  A recent study by the authors proposed a methodology for mapping, comparing and characterising diverse criminal justice mechanisms (Dancig-Rosenberg & Gal, 2014). The criminal law taxonomy (CLT) includes seventeen characteristics, presented as a set of scales extending between two opposing ends. To streamline the characteristics, they were divided into four clusters: (a) process-related characteristics, including elements that describe the format, structure and procedural specifics of each mechanism; (b) stakeholder-related characteristics, focusing on the participants and the nature of their involvement in the various justice mechanisms (layperson-centred vs. professional-centred, victim-oriented vs. not victim-oriented and so on); (c) substance-related characteristics, referring to the core dynamics at the heart of each mechanism (needs-based terminology vs. rights-based terminology, emotional discourse vs. no emotional discourse and so on); and (d) outcome-related characteristics, typifying criminal justice mechanisms on the basis of their designated outcomes (including future-oriented vs. past-oriented, retributive vs. restorative requital) (Table 1).
  The CLT was not designed specifically for analysing restorative justice mechanisms, but for analysing all mechanisms operating within criminal justice systems worldwide. Nevertheless, the right-hand side of each item of the CLT reflects the restorative justice scholarship regarding the ‘ideal’ restorative justice process, that is, the characteristics that underlie the restorativeness of a process. Accordingly, a coding that awards the highest score in all CLT characteristics reflects a fully restorative programme, a highly desirable outcome in the case of restorative justice. Other justice mechanisms, in contrast, might aspire to achieve different scores on the basis of their own desired goals and values. A discussion of the ideals of a range of other justice mechanisms is beyond the scope of this article. Focusing on restorative justice, Table 2 shows the links between the CLT characteristics and the restorativeness concepts and principles identified in the literature as discussed earlier, which they reflect.
  The CLT serves several purposes. One of them is to evaluate the adherence of individual programmes to the principles, goals and values of a given justice mechanism. It does not aim to replace existing measurements for evaluating programme effectiveness in reducing recidivism, promoting victim well-being or achieving participant satisfaction. Rather, it aims to help programme managers and law enforcement officials to assess model fidelity (Miller & Miller, 2015). In the context of the present article, the CLT is used to assess the level of restorativeness that the Mosaica CRJ conferences achieved.

  Table 2 CLT and Accepted Restorativeness Characteristics for CRJ

  CLT Characteristics	Restorativeness Characteristics	Source
  1. Victim-offender dialogue	Active participation by stakeholders; direct encounter	Maxwell et al., 2004; Van Ness and Strong, 2014; Daly, 2003; Presser and Van Voorhit, 2002
  2. Flexibility	Informal processes that facilitate cultural traditions	Maxwell et al., 2004
  3. Non-hierarchy	Joint decision-making	Maxwell et al., 2004; Daly, 2003, 2006
  4. Voluntariness	Avoidance of coercion; voluntary encounters	Maxwell et al., 2004; Van Ness and Strong, 2014; Zernova, 2006
  5. Laymen centered	Active involvement of community members	Zehr and Mika, 2003; Daly, 2003, 2006; Zernova, 2006; Presser and Van Voorhit, 2002; White, 2003
  6. Victim orientation	Reparation of harm to victims	Van Ness and Strong, 2014; Daly, 2003; Zernova, 2006
  7. Offender orientation	Adherence to offenders’ needs	Maxwell et al., 2004; Dhami and Joy, 2007
  8. Inclusiveness	Participation of victims, offenders and family members; Inclusiveness	Maxwell et al., 2004; Van Ness and Strong, 2014; Daly, 2003; Presser and Van Voorhit, 2002; White, 2003
  9. Community leadership	Community ownership of the process	Zehr and Mika, 2003; White, 2003
  10. Needs discourse	Reintegrative outcomes	Maxwell et al., 2004
  11. Emotional discourse	Discussing the harm	Van Ness and Strong, 2014; Daly, 2003; Presser and Van Voorhit, 2002
  12. Process as enabling	Direct encounter that facilitates constructive dialogue	Van Ness and Strong, 2014; Daly, 2003; Zernova, 2006
  13. Communitarian	Use of community resources	Zehr and Mika, 2003; Dhami and Joy, 2007; Presser and Van Voorhit, 2002; White, 2003
  14. Future oriented	Promotion of reintegration rather than stigma	Van Ness and Strong, 2014; Daly, 2003
  15. Restorative requital	Restorative outcomes; offender takes responsibility and repairs harm	Maxwell et al., 2004; Van Ness and Strong, 2014; Daly, 2003; Zernova, 2006
  16. Rehabilitative	Rehabilitative outcomes	Maxwell et al., 2004; Daly, 2003
  17. Justice making	Promotion of harmony and peace	Zehr and Mika, 2003; White, 2003

  3.5 Study design

  Our content analysis involved the following stages: First, we assigned an ordinal number (0, 1 or 2) to each CLT characteristic (Table 1). For each characteristic, 0 represents full agreement with the left-hand side of the scale, reflecting a non-restorative implementation; 2 represents full agreement with the right-hand side, reflecting fully restorative implementation; and 1 represents a mid-way position along each scale.³x The justification for using a 3-point ordinal scale was based on our assumption that it would make it easier for the raters to make unequivocal decisions in a vague and subjective context. Our coding related to each completed Mosaica conference. Karen’s case is representative of these conferences and their level of documentation.
  In the second stage, four coders, trained in the CLT and in the operational definitions of its characteristics, coded separately each of the 21 files according to the seventeen characteristics.⁴x Two of the coders were the first two authors, who had developed the CLT. The other two coders were research assistants with legal and social-science academic background (one was a former prosecutor, who had extended experience in working within restorative justice settings, while the other had no prior knowledge of restorative justice). The result was four separate tables of 21 × 17. Third, we used rWG (James, Demaree, & Wolf, 1984, 1993) to test the inter-rater agreement (IRA).⁵x The rWG is one of the most commonly used statistics for assessing IRA (LeBreton & Senter, 2008), and it is suitable for studies with two or more coders and a single or multiple targets. LeBreton and Senter (2008) provided commonly cited cut-offs for qualitative ratings of agreement on the basis of the logic presented by Nunnally and Bernstein (1994). According to these cut-off ratings, a lack of agreement produces IRA levels of .00 to .30; weak agreement, levels of .31 to .50; moderated agreement, levels of .51 to .70; strong agreement, levels of .71 to .90; and very strong agreement, levels of .91 to 1.00. As Table 3 indicates, the overall IRA was either strong or very strong across coders and case files. In other words, when coding 21 cases, four raters were at least in strong agreement about all the seventeen characteristics of the CRJ conferences.

  Table 3 Inter-rater Agreement among 4 Coders: rWG Assessment

  Characteristic #	Characteristic Name	rWG(J) 4 coders
  1	Victim-offender dialogue	1.00
  2	Flexibility	1.00
  3	Lack of hierarchy	0.99
  4	Voluntariness	0.99
  5	Laymen centered	0.99
  6	Victim orientation	0.99
  7	Offender orientation	0.99
  8	Inclusiveness	0.99
  9	Community leadership	0.99
  10	Needs discourse	0.99
  11	Emotional discourse	0.98
  12	Process as enabling	0.99
  13	Communitarian	0.98
  14	Future oriented	0.99
  15	Restorative requital	0.98
  16	Rehabilitative	0.99
  17	Justice making	0.98
  Average rWG(J)		0.99

• 4. Findings: characteristics of the Mosaica CRJ conferences

  Figure 1 presents the average coding of the 21 case files by the four coders. The strong and very strong IRA levels suggest that the analysis is reliable.
  As Figure 1 indicates, the Mosaica restorative justice conferences showed a highly restorative orientation by most CLT characteristics. Below, we present each finding, together with a short explanation of the characteristic as originally provided in the CLT (Dancig-Rosenberg & Gal, 2014). The finding is then compared with the expectations of restorativeness, as described above and summarised in Table 2.

  Average Coding for Restorative Justice across 17 Parameters
  

  4.1 Process-related characteristics

  1. Level of victim-offender dialogue. Typically, the Mosaica CRJ conferences facilitated direct victim-offender dialogue, with an average coding of 1.95 on a scale of 0-2. This finding is consistent with the restorativeness idea, which includes a victim-offender encounter as one of its basic and common features (Daly, 2003; Maxwell et al., 2004; Presser & Van Voorhis, 2002; Van Ness & Strong, 2014) and an important expectation among victims (Kirkwood, 2010; Van Camp, 2017).

  2. Formality vs. flexibility. The Mosaica CRJ conferences were characterised as highly flexible, receiving the maximum score in all 21 cases, by all coders (M = 2). Conferences were custom-tailored and not formalised in their symbolic representations such as language, attire and gathering appearances. The flexibility of the conferences enabled the parties to participate in ways that suited their cultural norms and preferences, in consistence with the restorativeness principle suggested by Maxwell et al. (2004).

  3. Hierarchical vs. flat process. The Mosaica CRJ conferences were characterised as non-hierarchical (M = 1.93). Participants spoke directly to each other, used first names and were not subjected to any rules of hierarchy. Most important, the participants engaged in joint decision-making, without anyone dictating the conference outcomes, as suggested in the restorativeness principles (Daly, 2003, 2006; Maxwell et al., 2004).

  4. Coercion vs. voluntariness. The Mosaica CRJ conferences were characterised as highly voluntary (M = 1.93). Although no mechanism can guarantee absolute voluntariness on the part of offenders because the threat of being referred back to the formal criminal justice process is always present, the Mosaica facilitators generally refrained from pressuring the parties to continue their engagement in the process, in accordance with the principles of restorativeness (Van Ness & Strong, 2014; Zernova, 2006).

  4.2 Stakeholder-related characteristics

  1. Professional- vs. lay-centred. Despite the involvement of trained facilitators, the Mosaica CRJ conferences were characterised as highly laymen-centred (M = 1.96). The private stakeholders held exclusive decision-making power, and the facilitators intervened only to direct or help advance the discussions. Particularly meaningful was the contribution of the community representatives, who often raised suggestions regarding the reparation plan and helped in its execution, reflecting strong community involvement in the process (Daly, 2003, 2006; Presser & Van Voorhis, 2002; White, 2003; Zehr & Mika, 2003; Zernova, 2006).

  2. Level of victim orientation. Conferences were highly victim-oriented (M = 1.87). Victims were always included in the process, and they usually expressed their needs, wishes and sense of justice. The reparation plans typically reflected victim-sensitive considerations. Victim orientation, or high sensitivity to victims’ wishes, interests and needs, is at the heart of the restorativeness principles, and is one of the characteristics that differentiates restorative justice from other criminal justice mechanisms (Daly, 2003; Van Ness & Strong, 2014; Zernova, 2006). For example, in case file K, an observing Muslim entered a supermarket store during the holy month of Ramadan. After filling a large shopping cart, he asked permission to exit through the emergency door, to shorten his path. The guard, an elderly Jew of Ethiopian origin, stopped him. Frustrated and hungry, the customer approached the cashier, who referred him to the manager, who in turn sent him back to the guard. When the guard refused again to let him through, the offender threw a large water bottle at the guard’s head. The offender fled, and the store manager filed a criminal complaint, which was later referred to Mosaica. At the conference, the offender’s father offered to take the victim and his family out to a restaurant. Instead, the victim said that he preferred to be repaid for the medicine he purchased and the loss of working days. The offender agreed and immediately paid the requested sum of approximately USD 200.

  3. Level of offender orientation. As expected, the CRJ conferences were found to be highly offender-oriented (M = 1.92). Processes were usually sensitive to the interests and wishes of the offenders, custom-tailored to their needs, as suggested by the principles of restorativeness (Dhami & Joy, 2007; Maxwell et al., 2004). In case file E, for example, a security manager at a Jerusalem mall attacked a journalist who refused to leave the mall and stop filming, despite the manager’s request. The journalist fell off the stairs and his camera was damaged. At the conference, the parties discussed reparation options, and the offender talked about his experience in training security guards. The parties agreed that he would attend several training sessions and would talk with the trainees about the ‘dos’ and the ‘don’ts’ of securing public malls and about his own involvement in the criminal justice system as a result of his behaviour. The plan, which required significant and continued activities on the part of the security manager (reflecting accountability), was designed and scheduled in consultation with him.

  4. Level of inclusiveness. The CRJ model welcomes the active participation of relatives and supporters in restorative justice encounters and considers them to be an important resource in the process of reaching a reparation plan (Daly, 2003; Maxwell et al., 2004; Presser & Van Voorhis, 2002; Van Ness & Strong, 2014; White, 2003). In our study, the conferences were found to be only partially inclusive (M = 1.40). Some conferences did not include any supporters or community representatives and involved only the victim, the offender and the facilitator, deviating from the ideal restorativeness model (Van Ness & Strong, 2014).

  5. State vs. community leadership. CRJ aspires to promote and reflect the community ownership of the process (White, 2003; Zehr & Mika, 2003). The Mosaica conferences were characterised as generally community-managed (M = 1.78). Most of the conferences were led by the participants themselves, their supporters and the community representatives. Case file R, for example, involved a young pub owner who had physically attacked a policeman attempting to enforce noise regulation at night. The CRJ conference brought together the pub owner and his mother, the policeman and his direct officer and a local community centre worker in charge of local business. After the offender expressed his remorse and sorrow for his actions, the parties discussed reparation options. The community representative suggested that the offender would work with children placed at a local emergency centre for battered women, an offer that he accepted enthusiastically. Eventually, the offender continued his voluntary work there long after the reparation terms were met. The community acted both as a resource and as a target for restoration. In contrast to this example, in several cases, no community representative attended the conference. Furthermore, in a few cases, the State agents (prosecutors, judges) were more dominant, intervening significantly or rejecting a reparation plan.

  4.3 Substance-related characteristics

  1. Needs vs. rights discourse. The terminology used in the Mosaica CRJ conferences was mainly needs-based (M = 1.82), focusing on the concrete needs or interests of the participants and setting aside their procedural rights and formal entitlements. This is not surprising, given that restorativeness means, among other things, an effort to address the underlying needs of offenders, victims and the community (Maxwell et al., 2004). In a few conferences, however, the process overlooked some of the stakeholders’ rehabilitative needs and focused more on procedure. In case file L, for instance, two neighbours, a young woman and a young man, engaged in mutual attacks following an ongoing dispute. They were referred to a CRJ conference and were considered both victims and offenders. The conference was very short and laconic and involved no supporters. It seemed as if the neighbours had already resolved their conflict and attended the process only to avoid criminal proceedings. No discussion about their feelings or rehabilitative needs took place. They agreed to perform community work, which they never did. The criminal case was nevertheless closed.

  2. Level of emotional discourse. Conferences were characterised as quite emotional (M = 1.64). Stakeholders expressed various emotions during the reparation meetings, reflecting the aspiration of restorativeness to provide a platform for an open discussion about the crime and its outcomes (Daly, 2003; Presser & Van Voorhis, 2002; Van Ness & Strong, 2014). Emotions included excitement, repentance, shame, sorrow, regret, anger, fear, resentment, hope, empathy and compassion. Several cases, however, were coded as lacking emotional discourse, as demonstrated in case file L.

  3. Process as vehicle vs. process as obstacle. The Mosaica CRJ conferences were characterised as largely enabling the pursuit of the restorativeness ideal in general, in particular the insistence on a direct encounter that enables constructive dialogue between victims and offenders (Daly, 2003; Van Ness & Strong, 2014; Zernova, 2006) (M = 1.96).

  4. Communitarian vs. liberal approach. CRJ programmes aspire to implement a communitarian approach, which stresses the importance of the community and its role both as a regulatory agent and as a stakeholder, by virtue of being a victim (Dhami & Joy, 2007; Presser & Van Voorhis, 2002; White, 2003; Zehr & Mika, 2003). Mosaica CRJ conferences were characterised as only partially communitarian (M = 1.67). At times, the discourse emphasised the private conflict, ignoring the social context in which the crime had been committed and the other community members whose interests were tied to those of the direct stakeholders. In case file E, for example, the victim became increasingly concerned about the welfare of two neighbouring children who lived with their divorced mother after seeing them searching for food in garbage cans. He had approached social services, but according to his statement, nothing had changed. One day, the mother’s ex-husband attacked the victim verbally and threatened his life should he ever come near his family again. The victim filed a complaint. Three years later the parties met at a restorative justice conference. The offender apologised and the victim explained that he acted out of concern for the children’s well-being. To repair the harm, the parties agreed that the offender would invest a sum of USD 130 to spend with his children on ‘quality fun time’ (Mosaica, Case file E, minutes of the conference). On the one hand, this case file seemed as a ‘missed opportunity’, because the children (who were now in their late teens) and their mother were excluded from the process. On the other hand, the reparation plan did involve the children, thus representing a ‘sufficiently communitarian’ approach.

  4.4 Outcomes-related characteristics

  1. Future vs. past orientation. Past-oriented mechanisms typically emphasise the blameworthiness of the offender and the severity of the offence as criteria for proportional punishment, reflecting the theoretical and moral basis of retributivism. By contrast, future-oriented mechanisms seek to achieve utilitarian outcomes that maximise wellness in society, such as rehabilitation, incapacitation and deterrence. The Mosaica conference produced agreed-upon outcomes that were coded as moderately tilted towards the future (M = 1.56). This coding reflects, at least to a certain degree, the aspiration of restorativeness to promote reintegration rather than stigma (Daly, 2003; Van Ness & Strong, 2014). The restorativeness ideal, however, was not always implemented, as attested to by the efforts to ascribe sums of reparation proportional with the severity of the offence. In Karen’s case, for example, the coding reflects duality: whereas the restitution sum of approximately USD 13,000 was interpreted as a retributive, past-oriented act of the victim, who wanted the young offender to ‘pay for what she did’, the process outcome was perceived, to some extent, as future-oriented, because it provided Karen with a ‘second chance’ to pay her debt without the labelling consequences of a criminal trial.

  2. Restorative vs. retributive requital. Even past-oriented outcomes can be defined as either retributive or restorative. Retributive outcomes are imposed as an act of vengeance, justified by their intrinsic moral value for being proportional to the severity of the offence and the blameworthiness of the offender. Restorative outcomes may impose a similar burden on the wrongdoer, but out of motivation to ‘right the wrong’ by repairing the harm caused to the victim. The Mosaica CRJ conferences were coded as more restorative than retributive, as expected according to the restorativeness ideal (Daly, 2003; Maxwell et al., 2004; Van Ness & Strong, 2014; Zernova, 2006) (M = 1.70). Case file N, for example, involved a young man who attacked his direct supervisor at the local municipal authority, after the supervisor removed him from his position. Both parties belonged to the ultra-Orthodox community and had known each other long before the incident took place (the offender had been a young client of the victim, in his role as a youth worker). The reparation plan involved a letter of apology and a decision that the offender would contribute 50 working hours at the municipality; the parties decided that the work would be conducted hand-in-hand with the victim, to rebuild trust.

  3. Rehabilitative vs. incapacitative. The Mosaica conferences were coded as largely rehabilitative (M = 1.88), in accordance with the restorativeness model (Daly, 2003; Maxwell et al., 2004).

  4. Justice-making vs. conflict resolution. This characteristic creates a continuum between processes that are limited to conflict-resolution outcomes and others that aspire to achieve a broader concept of justice, derived from universal norms of human rights and truth-finding. The Mosaica CRJ conferences tilted towards justice-making (M = 1.76), reflecting the restorative justice aspiration to reach solutions that promote harmony, peace and social justice (White, 2003; Zehr & Mika, 2003). At the same time, in some cases the outcomes were somewhat ambivalent and illustrated the different ways of perceiving similar outcomes. In Karen’s case, for instance, the coding reflects this ambivalence: on the one hand, Karen had an opportunity to tell her story in full and eventually to obtain Jacob’s forgiveness and understanding; on the other hand, the restitution money came from the deceased mother’s social security benefits, reflecting a narrow conflict-resolution outcome.

• 5. Discussion and conclusion

  The goal of this study was to expand the current knowledge about the implementation of the ideals of restorativeness in real-life restorative justice conferences. Our analysis of the cases handled by the Mosaica CRJ programme, using a detailed, theoretically robust list of seventeen scale-based characteristics, offers a method for identifying gaps between the restorative ideal and its daily application.
  Typical evaluation tools used in the restorative justice literature seek to measure the success of the programmes. By contrast, our CLT-based analysis uncovered the success, and at times the challenges of the Mosaica programme, in meeting the standards of restorativeness. The study pinpointed programme characteristics that make the process restorative, as well as the elements of restorativeness that were more difficult to attain in CRJ conferences. The findings suggest that the Mosaica conferences were particularly successful in achieving the ideal of restorativeness (M ≥ 1.9) with respect to the following variables: facilitating offender-oriented processes, providing for victim-offender dialogue, voluntariness, flexibility and the absence of hierarchy. These characteristics can be implemented by restoratively oriented professionals and are less contingent on private stakeholders, who typically attend conferences without prior restorative inclinations (Daly, 2006). The Mosaica facilitators used effectively the procedural elements of the restorative justice model to promote restorative values (process as vehicle). The conferences were also largely successful (1.7 ≤ M ≤ 1.89) in being victim-oriented and community-managed, in facilitating a needs-based discourse, in achieving restorative and rehabilitative outcomes and in generally promoting justice-making.
  Although community conferences were found to be both victim- and offender-oriented, supporting the win-win argument (Strang, 2002: 63), the process was overall rated as more offender- than victim-oriented. This finding suggests that although restorativeness means sensitivity to both victims’ and offenders’ needs, interests and wishes, it may be easier for restorative justice programmes operating in collaboration with the criminal justice system to fulfil the offender-orientation ideal. This is not surprising regarding the Mosaica programme, which receives its referrals from the formal criminal justice system. After all, offender orientation is closer to the rehabilitative lens (Zehr, 1990), already in place when addressing low- and even mid-level offences. Further studies focusing on other restorative justice programmes are warranted to examine the extent of the victim orientation challenge.
  Even more instructive are the findings concerning the characteristics of restorativeness that the programme was less successful in achieving (1.4 ≤ M ≤ 1.69). The conferences were not as inclusive as desired, they allowed for only partial emotional discourse, were not fully communitarian and did not always offer future-oriented outcomes. Community representatives were not always present, supporters did not always attend, discussions were at times shallow and laconic and the suitability of several cases for restorative justice was questionable, even in the retrospective evaluation of programme facilitators.
  These may be the most important findings, as they reflect an inherent challenge in fulfilling the ideal of restorativeness. The Mosaica CRJ programme was carefully designed by senior professionals at the Ministry of Justice, based on well-established and researched community conferencing programmes in other jurisdictions. The work of the facilitators was intensively supervised. The extensive preparatory phase for each conference, attested to in the documentation, further illustrates its strong adherence to the theory and philosophy of restorative justice (model fidelity) (Merkel-Holguin & Marcynyszyn, 2015; Rauktis, Bishop-Fitzpatrick, Jung, & Pennell, 2013). If these conferences were only partially successful in implementing features of the ideal of restorativeness, perhaps these features are too difficult to achieve, and a more realistic set of expectations should be formulated (Daly, 2006; Zernova, 2006). At the same time, the fact that the Mosaica conferences did meet most of the features of restorativeness in most of the case files paints an optimistic picture of the ability of well-constructed restorative justice programmes to live up to expectation. Further studies, based on similar methodologies, are needed to explore specific gaps between theory and practice, in particular in relation to those restorative justice characteristics that were found harder to implement in the current case study. The Mosaica staff and other restorative justice programmes may use these findings to improve their practice and make it more nuanced, in accordance with their own stated goals.
  Some of the findings deserve further consideration in light of the restorative justice theory itself. For example, the finding regarding the level of future orientation of CRJ conferences provides a first evidence of our attempt to capture restorative justice outcomes as both restorative and, at least partially, past-oriented. If the same case files were graded by the same coders as mostly restorative and only partially future-oriented, there is basis for our recent claim (Dancig-Rosenberg & Gal, 2013) that restorative justice outcomes may be past-oriented and restorative at the same time. Reparation plans agreed upon in restorative justice encounters typically relate to the past actions of offenders, but include restorative, rather than retributive, measures. A deontological approach that sees merit in burdening the offender for past actions, as our coding demonstrates, may be compatible with a restorative form of requital, such as the one that restorative justice promotes. Karen’s payment of a significant amount of money to her victim was definitely a past-oriented burden that was designed to ‘make her pay’, but it was also restorative because it repaired the harm she caused. This finding may contribute to the debate about the position and meaning of retribution in restorative justice (Dancig-Rosenberg & Gal, 2013).

• 6. Study limitations and future directions

  Arguably, the number of specific cases used in this study is a limitation. Our case study approach, however, involved an analysis of all of the completed files that were handled by the selected restorative justice programme. Furthermore, the coding of 21 files by four coders across seventeen parameters yielded a matrix containing 1,428 units of analysis, which proved sufficient for addressing our research questions.
  One of the questions emerging from the study is whether it is possible to extend our findings beyond this particular programme. While it is reasonable to consider the programme as representing, at the very least, the community conferencing model, the Mosaica CRJ programme does not necessarily represent restorative justice practice in general. Many restorative justice programmes involve different and often more serious crimes. It is reasonable to assume that in such cases at least some of the characteristics would produce different results, such as victim orientation, emotional discourse and retributive outcomes. Even so, our findings may be used as a starting point for characterising other processes representing the same or other restorative justice models, dealing with the same or other offences.
  Finally, the Mosaica case study is unique in that the documentation of each case file was, for the most part, complete. It included reports on every meeting and discussion with each of the participants, the reflections of facilitators and programme directors, as they worked on each case, and detailed descriptions of the restorative justice encounters. This level of detail made it easier to reach a strong IRA and consequently conclusive findings regarding the characteristics of the conferences. We can expect less well-documented case studies to produce somewhat weaker levels of IRA. One possible way of addressing this challenge is to use other methods, such as observations, as a basis for multi-rater coding.

References

• Angel, C.M., Sherman, L.W., Strang, H., Ariel, B., Bennett, S., Inkpen, N. & 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(3), 291-307.

• Bazemore, G. (1998). Restorative justice and earned redemption communities, victims, and offender reintegration. American Behavioral Scientist, 41(6), 768-813.

• Bonta, J., Jesseman, R., Rugge, T. & Cormier, R. (2006). Restorative justice and recidivism: promises made, promises kept. In D. Sullivan & L. Tifft (eds.), Handbook of restorative justice: a global perspective (pp. 108-120). New York: Routledge.

• Braithwaite, J. (2002). Setting standards for restorative justice. British Journal of Criminology, 42(3), 563-577.

• Crawford, A. & Newburn, T. (2002). Recent developments in restorative justice for young people in England and Wales: community participation and representation. British Journal of Criminology, 42(3), 476-495.

• Daly, K. (2003). Mind the gap: restorative justice in theory and practice. In A. Von Hirsch, J. Roberts, A.E. Bottoms, K. Roach & M. Schiff (eds.), Restorative justice and criminal justice: competing or reconcilable paradigms? (pp. 219-236). Oxford: Hart Publishing.

• Daly, K. (2006). The limits of restorative justice. In D. Sullivan & L. Tifft (eds.), Handbook of restorative justice: a global perspective (pp. 134-145). New York: Routledge.

• Dancig-Rosenberg, H. & Gal, T. (2013). Restorative criminal justice. Cardozo Law Review, 34(6), 2313-2346.

• Dancig-Rosenberg, H. & Gal, T. (2014). Criminal law multitasking. Lewis Clark Law Review, 18(4), 893-934.

• Dhami, M.K. & Joy, P. (2007). Challenges to establishing volunteer-run, community-based restorative justice programs. Contemporary Justice Review, 10, 9-22.

• Flyvbjerg, B. (2006). Five misunderstandings about case-study research. Qualitative Inquiry, 12, 219-245.

• Gal, T. (2011). Child victims and restorative justice: a needs—rights model. New York: Oxford University Press.

• Gal, T. (2016). ‘The conflict is ours’: community involvement in restorative justice. Contemporary Justice Review, 19(3), 289-306.

• Gal, T. & Moyal, S. (2011). Juvenile victims in restorative justice: findings from the reintegrative shaming experiments. The British Journal of Criminology, 51(6), 1014-1034.

• Harris, N. (2006). Reintegrative shaming, shame, and criminal justice. Journal of Social Issues, 62(2), 327-346.

• Harris, N., Walgrave, L. & Braithwaite, J. (2004). Emotional dynamics in restorative conferences. Theoretical criminology, 8(2), 191-210.

• Hudson, B. (2002). Restorative justice and gendered violence: diversion or effective justice?. British Journal of Criminology, 42(3), 616-634.

• James, L.R., Demaree, R.G. & Wolf, G. (1984). Estimating within-group interrater reliability with and without response bias. Journal of Applied Psychology, 69, 85-98.

• James, L.R., Demaree, R.G. & Wolf, G. (1993). rWG: an assessment of within-group interrater agreement. Journal of Applied Psychology, 78, 306-309.

• Kirkwood, S. (2010). Restorative justice cases in Scotland: Factors related to participation, the restorative process, agreement rates and forms of reparation. European Journal of Criminology, 7(2), 107-122.

• Kuo, S., Longmire, D. & Cuvelier, S.J. (2010). An empirical assessment of the process of restorative justice. Journal of Criminal Justice, 38, 318-328.

• Latimer, J., Dowden, C. & Muise, D. (2005). The effectiveness of restorative justice practices: a meta-analysis. Prison Journal, 85, 127-144.

• LeBreton, J.M. & Senter, J.L. (2008). Answers to 20 questions about interrater reliability and interrater agreement. Organizational Research Methods, 11, 815-852.

• Marshall, T.F. (1999). Restorative justice: an overview. London: Home Office.

• Maxwell, G., Kingi, V., Robertson, J., Morris, A., Cunningham, C. & Lash, B. (2004). Achieving effective outcomes in youth justice. Wellington: Ministry of Social Development.

• McCold, P. & Wachtel, T. (2003). In pursuit of paradigm: a theory of restorative justice. XIII World Congress of Criminology, 10-15.

• McGlynn, C., Westmarland, N. & Godden, N. (2012). ‘I just wanted him to hear me’: sexual violence and the possibilities of restorative justice. Journal of Law and Society, 39(2), 213-240.

• Merkel-Holguin, L. & Marcynyszyn, L.A. (2015). The complexity of fidelity in measuring system change: the case of family group decision making. British Journal of Social Work, 45(2), 724-736.

• Miller, S.L. & Iovanni, L. (2013). Using restorative justice for gendered violence: success with a postconviction model. Feminist Criminology, 8, 247-268.

• Miller, J.M. & Miller, H.V. (2015). Rethinking program fidelity for criminal justice. Criminology and Public Policy, 14(2), 339-349.

• Nunnally, J.C. & Bernstein, I.H. (1994). Psychometric theory (3rd ed.). New York: McGraw-Hill.

• Patton, M.Q. (2002). Qualitative research and evaluation methods (3rd ed.). Thousand Oaks: Sage.

• Poulson, B. (2003). A third voice: a review of empirical research on the psychological outcomes of restorative justice. Utah Law Review, 167-203.

• Presser, L. & Van Voorhis, P. (2002). Values and evaluation: assessing processes and outcomes of restorative justice programs. Crime and Delinquency, 48(1), 162-188.

• Ptacek, J. (Ed) (2009). Restorative justice and violence against women. Oxford: Oxford University Press.

• Rauktis, M.E., Bishop-Fitzpatrick, L., Jung, N. & Pennell, J. (2013). Family group decision making: measuring fidelity to practice principles in public child welfare. Children and Youth Services Review, 35(2), 287-295.

• Shapland, J., Atkinson, A., Atkinson, H., Colledge, E., Dignan, J., Howes, M. & Sorsby, A. (2006). Situating restorative justice within criminal justice. Theoretical Criminology, 10(4), 505-532.

• Sherman, L.W. & Strang, H. (2007). Restorative justice: the evidence . Smith Institute.

• Sherman, L.W., Strang H. & Woods, D.J. (2000). Recidivism patterns in the Canberra Reintegrative Shaming Experiments (RISE). Canberra: Centre for Restorative Justice, Research School of Social Sciences, Australian National University.

• Souza, K.A. & Dhami, M.K. (2008). A study of volunteers in community-based restorative justice programs. Canadian Journal of Criminology and Criminal Justice, 50, 31-57.

• Strang, H. (2002). Repair or revenge: victims and restorative justice. Oxford: Clarendon Press.

• Tyler, T.R. (2006). Restorative justice and procedural justice: dealing with rule breaking. Journal of Social Issues, 62(2), 307-326.

• Tyler, T.R., Sherman, L., Strang, H., Barnes, G.C. & Woods, D. (2007). Reintegrative shaming, procedural justice, and recidivism: the engagement of offenders’ psychological mechanisms in the Canberra RISE drinking-and-driving experiment. Law and Society Review, 41(3), 553-586.

• Umbreit, M.S. & Armour, M.P. (2011). Restorative justice and dialogue: impact, opportunities, and challenges in the global community. Washington University Journal of Law and Policy, 36, 65-89.

• Van Camp, T. (2017). Understanding victim participation in restorative practices: looking for justice for oneself as well as for others. European Journal of Criminology, 14, 679-696.

• Van Ness, D. & Strong, K. (2014). Restoring justice: an introduction to restorative justice (5th ed.). Waltham: Elsevier.

• Walgrave, L. (2013). Restorative justice, self-interest and responsible citizenship. Oxon: Willan.

• Walters, M.A. (2015). ‘I thought “he’s a monster”… [but] he was just… normal’: Examining the therapeutic benefits of restorative justice for homicide. British Journal of Criminology, 55(6), 1207-1225.

• White, R. (2003). Communities, conferences and restorative social justice. Criminal Justice 3(2), 139-160.

• Yanay, U. & Borowski, A. (2013). From a court orientation to a victim orientation: the paradigm shift in Israel’s Juvenile Probation Service. European Journal of Criminology, 10(6), 675-689.

• Zehr, H. (1990). Changing lenses: a new focus for crime and justice. Scottdale: Herald Press

• Zehr, H. (2015). The little book of restorative justice: revised and updated. New York: Skyhorse Publishing, Inc.

• Zehr, H. & Mika, H. (2003). Fundamental concepts of restorative justice. In E. Mclaughlin, R. Fergusson, G. Hughes & L. Westmarland (eds.), Restorative justice: critical issues (pp. 40-43). London: Sage.

• Zernova, M. (2006). Working under the aegis of the criminal justice system: implications for restorative justice practice. British Journal of Community Justice, 4(3), 67-79.

Noten

• 1 We use pseudonyms throughout this article, to maintain the anonymity of participants in the study.

• 2 The Ministry of Justice developed a detailed protocol for preparing, organising and facilitating restorative justice conferences. The documents are with the authors (Hebrew).

• 3 The justification for using a 3-point ordinal scale was based on our assumption that it would make it easier for the raters to make unequivocal decisions in a vague and subjective context.

• 4 Two of the coders were the first two authors, who had developed the CLT. The other two coders were research assistants with legal and social-science academic background (one was a former prosecutor, who had extended experience in working within restorative justice settings, while the other had no prior knowledge of restorative justice).

• 5 The rWG is one of the most commonly used statistics for assessing IRA (LeBreton & Senter, 2008), and it is suitable for studies with two or more coders and a single or multiple targets. LeBreton and Senter (2008) provided commonly cited cut-offs for qualitative ratings of agreement on the basis of the logic presented by Nunnally and Bernstein (1994). According to these cut-off ratings, a lack of agreement produces IRA levels of .00 to .30; weak agreement, levels of .31 to .50; moderated agreement, levels of .51 to .70; strong agreement, levels of .71 to .90; and very strong agreement, levels of .91 to 1.00.