• 1 Introduction

  Victim-offender mediation (VOM) is an example of a restorative justice programme that offers offenders and victims of a crime the opportunity to contact each other and talk about the offence, under the guidance of a trained mediator (Claessen & Roelofs, 2020; Zehr, 2015). Victims can ask questions, tell their story, and it helps them to process the crime (Bazemore & Umbreit, 2001). Parties together can come to an agreement in which it can be stated what offenders can do to (further) restore the damage that has been inflicted. There is consensus in the literature that participation in VOM can have beneficial effects for both victims and offenders (e.g. Hansen & Umbreit, 2018; Zebel, 2012). It has been shown that both parties can feel more satisfied after a VOM process and experience it as fairer compared to the conventional criminal justice system in which VOM is not offered (Baldry, 1998; Hansen & Umbreit, 2018). In addition, participation in VOM is associated with reduced feelings of fear and anger among victims (Hansen & Umbreit, 2018; Zebel, 2012). VOM can also help to learn the offender what the actual impact of the crime is (Choi, Green & Gilbert, 2011), heighten victim empathy (Baldry, 1998; Jonas-van Dijk, Zebel, Claessen & Nelen, 2022b) and support the desistance process (Lauwaert & Aertsen, 2016).
  It is common (although not a given) that restorative justice programmes, such as VOM, are part of the criminal justice process (Claessen, Zeles, Zebel & Nelen, 2015). It can thus be argued that restorative justice programmes should elicit crime reduction effects that are better or at least not worse than similar cases that go through the conventional criminal justice system without restorative justice (Claessen et al., 2015). In line with this argument, multiple studies show that participation in VOM during the criminal justice process is related to a reduced risk of reoffending (Bergseth & Bouffard, 2013; Claessen et al., 2015; Jonas-van Dijk, Zebel, Claessen & Nelen, 2020; Stewart et al., 2018). On the other hand, other studies have observed no differences between the effects of VOM and the effects of the conventional criminal justice process without VOM on reoffending (Boriboonthana & Sangbuangamlum, 2013; Villanueva, Jara & García-Gomis, 2014).
  In a recent empirical study, the psychological outcomes of participation in VOM for offenders, which may explain in part a reduced risk of reoffending, were scrutinised (Jonas-van Dijk et al., 2022b). The researchers examined whether offenders who participated in VOM showed a different psychological change over time compared to offenders who did not participate. They cautiously concluded that offenders who did participate in VOM showed increased feelings of guilt, responsibility-taking, awareness of their moral failure, and victim empathy. These are outcomes that have also been put forward by different scholars (Choi et al., 2011; Marsh & Maruna, 2016; Miller & Hefner, 2015; Pabsdorff, Rytterbro, Sambou & Uotila, 2011). Moreover, these offenders felt less awkward about the prospect of meeting the victim in the future than those who did not participate. Victim empathy, responsibility-taking and guilt feelings, all have been linked previously to pro-social and less reoffending behaviour (De Hooge, Zeelenberg & Breugelmans, 2007; Gausel, Vignoles & Leach, 2016; Hosser, Windzio & Greve, 2008; Tangney, Stuewig & Martinez, 2014). With that, there seems to be indirect evidence that participation in mediation is related to a psychological impact that could possibly reduce the risk of reoffending. However, the question that remains unanswered is which mechanisms of the VOM process might bring about these outcomes. Therefore, the research question in this study is which working mechanisms of the VOM process might explain conducive psychological outcomes in offenders?
  A literature review reveals that two main categories of elements of the VOM process can be identified as candidates for impacting the quality of the VOM encounter and, therefore, fostering psychological outcomes among offenders (Jonas-van Dijk, Zebel, Claessen & Nelen, 2022a). The first category is related to fundamental conditions of a VOM process. These are requirements that should be present or taken into account in every VOM process, such as an elaborated choice of parties to participate based on informed consent (i.e. voluntariness of participation). These conditions are also closely connected to two of the three pillars or principles of mediation: voluntary participation and an impartial mediator (Mesmaecker, 2013). The second category contains working mechanisms that relate to the VOM encounter itself: what happens during this encounter that impacts the psychological outcomes? Importantly, the results of the aforementioned literature review suggest that there is an interplay between these two main categories (Jonas-van Dijk et al., 2022a). That is, the fundamental conditions might influence the impact of the working mechanisms on the psychological outcomes or might influence the presence of the working mechanisms directly. In the following, these fundamental conditions and working mechanisms will be elaborated upon.
  It is important to indicate that there might exist other working elements or fundamental conditions that have not been taken into account in this research. In the present research, we aimed to examine empirically these elements that have been put forward or shown to be related to the impact of mediation on offenders in terms of reoffending or psychological outcomes. These elements that have been proposed already by different scholars have been divided into fundamental conditions and working mechanisms.
  It is also important to point out the difference between the already proposed set of standards by Braithwaite (2002) (constraining standards, maximising standards, emergent standards) and the distinction made in the present research. Braithwaite approached these standards from a perspective coming from the values of restorative justice principles and human rights principles. To live up to the values of restorative justice, Braithwaite proposed to adhere to constraining standards, maximising standards and emergent standards. Constraining standards should be adhered to in every restorative process. Maximising standards should be encouraged by the facilitator in an active way, and emergent standards should be interpreted as a bonus within the process. The largest difference between these standards and the mechanisms that are under investigation in this research is the starting point taken. Braithwaite’s starting point is restorative values, whereas our starting point is the psychological impact that VOM can have. Potentially, these different standards proposed by Braithwaite could be a fundamental condition or working mechanism but have, in our view, not all been put forward by scholars to be related to a psychological change.
  To summarise, the aim of this study is to examine the relationship between the fundamental conditions and working mechanisms of VOM encounters identified in the literature and the psychological outcomes observed in offenders (Jonas-van Dijk et al., 2022b): increases in victim empathy, responsibility-taking, feelings of shame, guilt, heightened awareness of moral failure and reduced awkwardness for meeting the victim in the future. This is to our knowledge the first research in the field of VOM that aimed to examine empirically the relationship between the degree of elements of a VOM encounter and the degree of occurrence of psychological outcomes. In the remainder of this introduction, the different fundamental conditions and working mechanisms will be explained, resulting in a hypothesis of how these conditions and working mechanisms could be related to the psychological outcomes. With a combination of a qualitative and quantitative research approach, this research will examine these hypotheses in the context of a Dutch Mediation programme within criminal cases.

  1.1 Fundamental conditions of VOM

  1.1.1 Voluntariness of participation

  In the literature, three fundamental requirements of VOM were found that seem to be important fundamental conditions for VOM to produce psychological outcomes in offenders (Jonas-van Dijk et al., 2022a). The first one is voluntary participation in VOM. Participation in VOM is based on informed consent and parties have choice to participate. When offenders are well informed about VOM and are aware of the consequences of their voluntary choice to participate, this could influence the effectiveness of the conversation (Hansen & Umbreit, 2018). Consistent with Shapland et al. (2008: 41), who found that offenders who really wanted to meet the victim were observed to be more actively involved during a mediated encounter, we expect that offenders who perceive to participate voluntarily in VOM are more actively involved than offenders who feel (partly) forced to participate. Thus, we believe that this perceived voluntariness can have an (indirect) impact on the psychological outcomes among offenders, since it might influence the quality of the conversation.

  1.1.2 Preparation of parties

  When the mediator successfully manages expectations and explains what parties can expect to happen during a VOM meeting, this might have a positive impact on the course and perceived quality of the dialogue (Hansen & Umbreit, 2018). That is, research indicates that offenders who are well prepared also show to be more involved in the VOM encounter (Gerkin, 2009). Gerkin also postulates that when participants are not aware of the aim of restorative justice, they become subjects rather than participants in the process. In line with this, Hansen and Umbreit (2018) argue that proper preparation is one of the key elements to ensure the most optimal mediation outcomes. Hence, this suggests that proper preparation has an impact on the quality of the conversation and thus might impact the psychological outcomes in offenders.

  1.1.3 Professional competencies of the mediator

  Although every mediator might have a different style that could result in a variety of encounters, some tasks of the mediator are fundamental and should be present in every VOM process. Mediators play an important role in preparing offenders and victims during an intake meeting, as well as in the encounter itself: they open the conversation, set the tone, listen carefully, guide the conversation by asking open questions, summarise and make sure that parties are able to have a good conversation (Pabsdorff et al., 2011; Szmania, 2006). In all of this, the offender should perceive the mediator to be neutral and feel listened to (Choi, 2008).

  1.2 Working mechanisms of the VOM encounter

  1.2.1 A learning opportunity

  Previous research suggests mechanisms of the VOM encounter itself that may help to explain psychological outcomes among offenders. One such mechanism is offering a learning opportunity for offenders, by talking to the victim (Choi, 2008; Fellegi, 2008). Talking to victims and hearing the impact of the crime on their lives may contribute to psychological outcomes for offenders, such as feelings of victim empathy, remorse, accountability, shame and guilt (Abrams, Umbreit & Gordon, 2006; Meléndez, 2020; Miller & Hefner, 2015). In addition, talking to the victim during the encounter can help the offender to realise that there is an actual victim, which might explain victim empathy (Meléndez, 2020). As Choi et al. (2011) explain it, VOM can help put a human face on the crime. Therefore, in this research we try to find indications that by talking to the victim during an encounter, offenders become more aware of the impact of the crime on the victim and whether this might be related to psychological changes that might be found in the offender afterwards.

  1.2.2 Humanising impact

  The theory of reintegrative shaming postulates that the manner in which society reacts to offenders after crime with the intention to invoke remorse can influence the risk of subsequent deviant behaviour (Braithwaite, 1989). When offenders perceive to be responded to respectfully while their actions are disapproved, this is called reintegrative shaming. In contrast, when disapproval is not only focused on the act but also on offenders as a person of bad character, offenders can experience feeling labelled and stigmatised as criminal, which is called disintegrative shaming (Braithwaite, Braithwaite & Ahmed, 2018). Offenders might then act upon this labelling and will accordingly stand at risk to commit more crimes (Bernburg & Krohn, 2003). Also, Fellegi (2008) argues that when shaming is expressed in a reintegrative way, shaming can be productive and enable moral learning. Previous research shows that restorative justice programmes are experienced as less stigmatising compared to the conventional criminal justice system (Shapland et al., 2008). In particular, research from Abrams et al. (2006) and Baldry (1998) shows that during a VOM meeting offenders experience to be perceived more as human than offenders in the contemporary justice system. An open-minded and non-judgemental attitude of the victim and the mediator might foster this humanising effect of VOM (Lauwaert & Aertsen, 2015). It also helps when the victim is really motivated to listen to the offender (Lauwaert & Aertsen, 2015).

  1.2.3 Positive atmosphere and interaction with a cooperative victim

  Based on interactional ritual theory, a shared focus of attention and emotional mood characterises an effective conversation during a restorative justice programme, in which a mutual feeling of group membership and solidarity arises (Hausmann, Jonason & Summers-Effler, 2011; Rossner, 2011). A smooth conversation, synchronised bodily actions and power balance are elements of such an effective interaction ritual, which can increase offenders’ willingness to conform to social morality (Hausmann et al., 2011). Eventually, this can reduce the risk of reoffending (Sherman et al., 2005). We expect that a positive atmosphere and an open and cooperative attitude of the victim towards the offender are related to the psychological outcomes, since they might impact the course and quality of the conversation. One could argue that this mechanism can also be placed in the category of fundamental conditions of VOM, since preferably every VOM encounter takes place in a pleasant working atmosphere with a cooperative victim (Braithwaite, 2017). However, since these mechanisms are not completely controllable and can fluctuate naturally during the encounter itself, we consider it an element of the VOM encounter.
  As part of the interaction with the victim, we will also focus on the offenders’ apologies. An apology is considered to be an important element of a restorative justice process (Choi, Bazemore & Gilbert, 2012) and, for some parties, the main reason for participation (Dhami, 2016). However, it is not to be expected that in every VOM process an offenders offers a (sincere) apology (Braithwaite, 2017). Also, when offenders do apologise to victims during VOM encounters, the perceived sincerity of that apology is in ‘the eyes’ of the victims: they decide whether they perceive it as genuine and accept it (Bonensteffen, Zebel & Giebels, 2020). Whether or not victims respond positively towards a remorseful offender might impact the further course and impact of VOM (Rypi, 2016). For example, non-forgiveness on part of the victim might lead to feelings of victimhood within the offender and could hinder reconciliation subsequently (Thai, Wenzel & Okimoto, 2021). Since it cannot be expected or demanded from a victim to give a specific reaction to the apology (Suzuki & Jenkins, 2023), we are especially interested in the response of victims to an apology and how this is in turn perceived by offender. It is hypothesised that when offenders do not perceive victims to respond in a positive way to their apologies, this negatively influences to what extent offenders experience psychological outcomes such as remorse and empathising with the victim.

• 2 Method

  2.1 VOM in the Netherlands

  We examined a VOM programme in the Netherlands: mediation in criminal cases (in Dutch ‘Mediation in Strafzaken’ [MiS]). This programme is part of the criminal justice process, meaning that the criminal prosecutor or judge refers cases to VOM. When victims and offenders participate and come to an agreement during VOM, this agreement is communicated back to the referral agency, with consent of the parties. The criminal prosecutor or judge then decides what kind of punishment, if any, needs to be imposed.
  Two mediators are involved in every case. Before the actual encounter takes place, mediators first have separate meetings with victims and offenders to make sure that a meeting will be helpful for both parties. The entire duration of this VOM process (from referral to finalisation) is expected to take place within six weeks (Ministry of Justice and Safety, 2021).

  2.2 Research design

  Both qualitative and quantitative data were obtained. Offenders were asked to fill out a pre-test questionnaire at the start of the VOM process and a post-test questionnaire six to eight weeks after the VOM encounter took place. For the quantitative data, a correlational research design was used and the answers on the post-test questionnaire were analysed. We did not examine the differences between the pre- and post-test scores in this article. Previous research showed that these differences were very small (Jonas-van Dijk et al., 2022b). However, in that study differences did exist on the outcome variables between offenders who participated in mediation and those who did not. It is for this reason that we considered it key to examine how the working mechanisms and fundamental conditions correlate with the outcome variables on the post-test.
  In addition to these questionnaires, the mediators who handled the cases of these offenders were asked to fill out an observation form, right after the mediation encounter. This observation form included both Likert scale questions and open-ended questions; hence, these latter questions comprise the qualitative element of this research. The responses of the mediators were important to get a detailed description of what happened during the mediation process and how specific elements might have been related to a change within the offender. For the observed cases, we did look to what extent offenders showed a change in the outcome variables on the post-test compared to the pre-test. First, this offers a more detailed insight into the impact of VOM for individual offenders. In addition, we asked mediators to what extent they saw a change happening with the offenders in the outcome variables. We wanted to examine whether this observed change was in line with the change found between the pre- and post-test.

  2.3 Participants

  For the quantitative data in this study, a total of 55 offenders participated. Twenty-three (42 per cent) offenders filled in both the pre-test and the post-test. Twenty (36 per cent) offenders filled out the pre-test but dropped out and did not fill in a post-test. For twelve (22 per cent) offenders, the researcher was informed too late about their participation in VOM and, therefore, no pre-test could be completed. These offenders did fill out a post-test. Participants’ age varied between 15 and 67 (mean age 39).¹x The two minors received parental permission for participating. Table 1 provides an overview of the sample.

  Table 1 Overview of the sample

  		N	%
  Gender	Male	41	74
  	Female	14	26
  Highest education completed	Elementary school	1	2
  	High school	24	44
  	College	14	26
  	Missing	16	29
  Country of birth	Netherlands	40	73
  	Other	1	2
  	Missing	14	26
  Type of case	Personal	38	69
  	Property	11	20
  	Traffic	2	4
  	Missing	4	7
  District	Limburg	4	7
  	Amsterdam	10	18
  	Gelderland	3	6
  	Overijssel	14	26
  	Noord-Holland	9	16
  	Rotterdam	2	4
  	Den Haag	3	6
  	Oost-Brabant	2	4
  	Midden-Nederland	2	4
  	Zeeland-West-Brabant	6	11
  First time offender	No	9	16
  	Yes	31	56
  	Missing	15	27

  In nine of the cases in which offenders filled out the pre- and post-test questionnaires, mediators also filled out an observation form (directly) after the VOM session. Since (almost) every case involved two mediators, some cases had two observations. This resulted in thirteen different observations, for those nine cases. Four of those nine observed cases concerned conflicts between neighbours or were related to a conflicting neighbour issue. That is, two cases concerned physical abuse of neighbours; in one case, neighbours were threatening each other and destroyed each other’s property; and, in one case, the son of an elderly woman physically abused his mother’s neighbour. In two cases, a traffic controller (victim) tried to correct the offender, which in one case resulted in an offender dragging the victim for a few metres with his car and, in the other case, the offender pushing the victim, resulting in physical consequences. One case concerned a case of fraud; the offender asked victims at the door of their house to give him some money to visit his sick daughter. In that way, he made (according to the mediator) thousands of victims. Twenty-two victims filed a police report, and two of those victims were willing to talk to the offender. The eighth case concerned a man who threatened to hit two children with a machete, because the children scared his dog with fireworks. In the last case, one man was a suspect of physically abusing another man during a night-out. All mediations ended with a (signed) agreement and could therefore be considered as successful. Table 2 presents an overview of the characteristics of the nine observed cases by mediators.

  Table 2 Overview of the nine cases that mediators observed

  Case name	Place of encounter	Duration of encounter	Persons present	Number of observations
  Conflicting neighbours	Police station	Unknown	Offender (f) Victim (f)	2
  Threatening neighbours	Court of law	90 minutes	Offender (f) Victim (m) Husband victim Daughter offender	1
  Physical violence between neighbours	Mediator’s office	60 minutes	Offender (m) Victim (m)	1
  Fence fight between neighbours	Court of law	120-135 minutes	Offender (m) Victim (m) Mother victim Brother offender	2
  Violence in traffic	Court of law	70 minutes	Offender (m) Victim (m)	1
  Driving away from traffic accident	Unknown	60 minutes	Offender (m) Victim (m) Caregiver victim	2
  Case of fraud	Court of law	75 minutes	Offender (m) Victim (f)	1
  Man threatening two children	Court of law	45 minutes	Offender (m) Victim (m) Mother victim	1
  Violence while going out	Court of law	60-75 minutes	Offender (m) Victim (m) Girlfriend victim	2

  Note: m = male, f = female.

  2.4 Measurements

  2.4.1 Fundamental conditions

  Voluntariness. To measure to what extent offenders felt that they were able to voluntarily participate in VOM and were not forced, three items asked about the voluntariness to participate. The items were derived from research from Shapland (2006). An example of an item was: ‘My participation in VOM was voluntary.’ After reverse-coding one item, a factor analysis showed that the three items measured one underlying factor with an Eigenvalue higher than 1, explaining 53 per cent of the variance. All items loaded highly on that factor (factor loadings > 0.62). However, the reliability analysis showed a Cronbach’s alpha of 0.53, which is questionable.
  Preparation. To measure to what extent offenders considered to be well prepared for participation in VOM, a scale was formed based on items derived from Shapland et al. (2007). This scale consisted of five items, measured on a 7-point Likert scale with responses that ranged from strongly agree to strongly disagree. Example items were to what extent offenders thought they received information about how the process of mediated contact worked, what the possible result of VOM could be and what the effects could be on the criminal case. A factor analysis showed one underlying factor with an Eigenvalue higher than 1, explaining 62 per cent of the variance. All items loaded highly on this factor (factor loadings > 0.71). The scale showed to be reliable (α = 0.83).
  Role mediator. To examine the role of the mediator we have asked offenders two questions: to what extent they thought the mediator to be neutral and to what extent they thought the mediator took them seriously. Both questions were measured on a 5-point Likert scale and were derived from Shapland et al. (2007). The two questions were separately analysed.

  2.4.2 Working mechanisms

  Learning opportunity. To our knowledge, it has not been examined to what extent during a VOM process the offenders experience to learn and become aware of the rules and norms they have violated. Therefore, a new scale had to be formed. With three items the scale measured to what extent (1) offenders became more aware of the rules they had broken, (2) offenders understood better why it was wrong what they did and (3) offenders were made aware that their behaviour did not fit the norms and values of society. An exploratory factor analysis showed that these three items measured one underlying factor with an Eigenvalue higher than 1. All items loaded highly on this factor (item loadings > 0.86). In addition, this scale also showed to be highly reliable with a Cronbach’s alpha of 0.89.
  Mediators were also asked to indicate to what extent they thought the violated rules and norms were discussed during VOM and to what extent they thought offenders became more aware of how their behaviour violated these rules and norms. These two items were again measured with 5-point Likert scale with responses ranging from not at all to very much.²x Given that there were thirteen observations and we used these data of mediators qualitatively, factor analysis and reliability analysis are not reported for the scales filled in by the mediators. In addition, they were able to give other remarks about the discussion of rules and norms in an open-ended question.
  Humanising impact. To measure to what extent offenders thought they were stigmatised or reintegrated, items were adopted from research from Harris (2006). Five items measured reintegration, and four items measured stigmatisation. Harris (2006) found that these two scales were valid and reliable. However, since this study examined VOM and not conferencing as in Harris’ study, and because the items had to be translated to Dutch to fit the sample, it was decided to do another factor analysis and reliability analysis. An exploratory factor analysis confirmed the two scales and showed two factors with an Eigenvalue higher than 1. The items measured the intended scales (factor loadings > 0.70). An example of an item measuring reintegrative shaming was: ‘During the mediation session you have learned that people care about you.’ The item ‘During the VOM session you felt that you were treated as criminal’ measured stigmatisation. Both scales showed to be reliable as well (reintegrative shaming, α = 0.82; stigmatisation, α = 0.89).
  Mediators were asked to indicate to what extent they thought victims stigmatised the offender during VOM. In addition, they were asked about reintegrative shaming: whether the victim was looking at the offender more positively during VOM, whether victims indicated towards the offender that they thought it was a positive thing that the offender participated in VOM, and if not the offender, but only the act was disapproved. In addition, it was asked whether the mediators thought offenders had the feeling that the victim looked at them more positively. These questions were administered using a 5-point Likert scale, with responses ranging from not at all to very much. Mediators also had the option to add remarks about stigmatising and reintegration in an open-ended question.
  Positive and effective interaction with victim and general atmosphere. To examine whether the atmosphere of the VOM encounter was associated with the impact of VOM on the offender, we asked offenders to rate eight different elements of the mediation process based on the research of Shapland et al. (2007): (1) their satisfaction with the process, (2) to what extent they were treated with respect, (3) to what extent they were able to speak freely, (4) to what extent they felt safe during VOM, (5) to what extent they were listened carefully to, (6) to what extent they thought the victim took them seriously, (7) to what extent they thought everyone had even possibilities to speak and (8) to what extent they thought to have a say in the outcome. Every item was measured on a 7-point Likert scale. We have not examined these items combined into one scale (as one construct) for multiple reasons. First of all, the factor analysis for these eight items showed two distinct factors with an Eigenvalue higher than 1. In addition, the mediators were also asked specifically about elements concerning the atmosphere and how parties treated each other. To connect the outcomes of the quantitative research to the qualitative research we aimed to have comparable measures. And lastly, for mediators it would be informative to know for their work how each of the eight elements would be related to the psychological outcomes.
  Mediators were also asked about the atmosphere. On a 5-point Likert scale it was asked how actively involved both parties were, whether they treated each other and the mediator(s) with respect, whether the parties showed provocative behaviour and whether the mediation ended positively. With an open-ended question, mediators were also asked to give a description of the complete mediation process. We urged them to describe how the VOM encounter started and how it ended, what the atmosphere was like, what the attitudes of the parties were like and whether it was an emotional dialogue. Mediators also had the possibility to give any other remarks about the atmosphere of the VOM process.
  Victims’ responses to offenders’ apologies. First, it was asked whether offenders offered their apology to the victim (sixteen offenders did not offer their apology). If so, we examined whether the offenders felt the victim accepted their apology, whether the offenders thought the victim perceived the apology as sincere and whether the offenders thought the victim appreciated the apology, which was measured on a 5-point Likert scale. All items loaded high on one factor (factor loadings > 0.90) and formed one reliable scale (α = 0.92). This scale was named apology acceptance.
  The mediators were asked the same questions: whether they thought the apology was sincere, whether the victim perceived it as sincere and whether the victim accepted the apology, with responses measured on 5-point Likert scales. In addition, mediators were asked to give a brief summary of the content of the apology and to give any other notable remarks about the apology.

  2.4.3 Psychological change variables (outcomes after the VOM process)

  Psychological change. In this research, we will examine six variables that previous research suggested were impacted by participation in VOM (Jonas-van Dijk et al., 2022b): increased victim empathy, responsibility-taking, feelings of guilt and shame, reduced experience of moral failure, and awkwardness for meeting the victim in the future. Jonas-van Dijk et al. (2022a) showed that offenders who participated in VOM experience more victim empathy and feelings of responsibility, guilt and shame and experience a higher moral failure in comparison to offenders who did not participate in VOM and had their cases dealt with solely through the criminal justice system. For a complete description of these variables, we refer to the original research of Jonas-van Dijk et al. (2022a).
  Mediators were also asked to indicate to what extent offenders showed those variables at the beginning of the mediator encounter, on a 5-point Likert scale with responses ranging from not at all to very much. In addition, we asked the mediators to what extent these feelings that the offender showed decreased or increased during the encounter, again measured on a 5-point Likert scale with responses ranging from showing much less to showing much more. By asking mediators to reflect on these changes we hoped to be better able to connect mechanisms of the VOM process to psychological outcomes from the perspective of the mediator.
  Finally, to gauge the relationship between the victim and the offender, we have asked mediators also to indicate to what extent the relationship between the parties has been restored during mediation.

  2.5 Procedure

  When an offender was referred to mediation and after the mediation officer of the mediation bureau had contact with the offender, a research intern explained the research to the offender by phone. When the offender was willing to participate, the intern then sent the contact details to the researcher. The online questionnaire was then sent to be filled out before the VOM encounter.³x The online platform Qualtrics was used to register the responses. The post-test questionnaire was sent out six to eight weeks after the VOM encounter took place. Reminders were sent by email and via telephone. An incentive was offered. All offenders participating in the study could win one out of five gift cards. Mediators were asked to fill out the observation form right after the VOM encounter and were offered a gift card worth 20 euros as incentive for participation in the study. This study received ethical approval from the relevant academic authorities, as well as from the public prosecution office.

  2.6 Data analysis

  To examine which conditions and mechanisms of VOM might be associated with psychological outcomes, correlation analyses were used. In the analyses, we have correlated the proposed fundamental conditions and working mechanisms with the scores on the psychological outcomes. To examine whether the fundamental conditions predict the working mechanisms, correlational analyses were again used. Multiple linear regression analyses were then used to examine whether the fundamental conditions work as moderator and impact the correlation between the working mechanisms and the psychological outcome variables.
  After the correlational analyses, the principal researcher analysed the observational data collected among mediators after the VOM encounter. Using ATLAS.ti the data were structured, which made it possible to look for patterns. For coding the qualitative data, codes were used that were related to the independent and dependent measures of this research. In this way, patterns and codes could be identified that underpin and may complement the observed correlations. In addition, the observations were used to examine whether any mechanisms were observed that were not included or shown in the correlation analyses. The observations were used to further deepen the quantitative analyses.

• 3 Results

  3.1 Multiple imputation

  Due to missing data multiple imputation was used, to maintain a sample of 55 offenders for the quantitative analyses. With multiple imputation a complete version of a dataset can be formed, based on an incomplete dataset. Using a regression model and a random error term, missing data points in the dataset are replaced by predicted values (Little & Rubin, 1989). The advantage of this method in comparison to alternatives, such as listwise or pairwise deletion, is that the sample remains intact. This is especially an advantage in the context of smaller samples, such as in this study. In addition, listwise or pairwise deletion could result in a bias when data are not completely missing at random (van Ginkel, Linting, Rippe & van der Voort, 2020).
  Imputation was done using default settings in IBM SPSS statistics 25. The minimum and maximum constraints were set according to the Likert scale and rounded to the nearest integer. Almost 40 per cent of all data of the offenders used in this study was missing, for which 40 imputations were necessary. That is, White, Royston and Wood (2011) suggest that the number of imputations should at least be equal to the percentage of missing data. The rounded means of the original data were compared to the imputed data; these showed to be highly similar (Table 3). Therefore, the imputed data were used in the analysis and pooled outcomes were interpreted. The pooled parameter estimates are calculated by taking the average of the parameters from all imputed outcomes. Standard errors are pooled by combining variance within the imputation and the between-imputation variance (Eekhout, n.d.).

  Table 3 Means and standard deviations of the original and imputed variables

  	Mean (SD) original data (N = 33-35)		Mean (SD) imputed data (n = 55)
  Fundamental conditions
  Preparation	4.1 (0.64)		4.1 (0.58)
  Voluntariness	4.1 (0.67)		4.1 (0.65)
  Professional competences mediator	4.1 (0.82)		4.1 (0.95)
  Working mechanisms
  Satisfaction	4.0 (0.94)		3.9 (1.0)
  Being treated with respect	4.2 (0.92)		4.2 (1.0)
  Being able to speak freely	4.3 (0.84)		4.3 (0.95)
  Feel safe	4.3 (0.77)		4.3 (0.80)
  Be listened to	4.0 (1.1)		3.9 (1.2)
  Taken seriously by victim	3.0 (1.4)		3.0 (1.4)
  Having equal possibilities to speak	4.1 (1.1)		4.0 (1.2)
  Have a say in outcome	3.6 (1.2)		3.6 (1.2)
  Apology acceptance	3.9 (0.98)		3.9 (0.85)
  Stigmatising	2.1 (1.2)		2.2 (1.0)
  Reintegrative shaming	3.3 (0.90)		3.3 (0.81)
  Awareness broken rules and norms	2.8 (1.6)		2.8 (1.0)
  Psychological outcomes
  Empathy	3.3 (1.1)		3.3 (0.97)
  Guilt	3.0 (0.91)		3.0 (0.79)
  Shame	2.7 (1.4)		2.7 (1.5)
  Responsibility	2.9 (1.0)		2.9 (0.92)
  Moral failure	2.8 (1.1)		2.8 (1.0)
  Awkwardness meeting the victim	2.2 (1.1)		2.2 (1.2)

  3.2 Examination of the mean scores

  Table 3 shows that this sample overall perceived a positive atmosphere; the scores for satisfaction, being treated with respect, feeling able to speak freely, being listened to, professional competences of the mediator, and having equal possibilities to speak were all between 4.0 and 4.3. Only the score for being taken seriously by the victim is a bit lower, at 3.0. So, it seems that the offenders in the sample did not always experience to be taken seriously by the victim, as this score was neutral.
  The sample showed to be well prepared and to participate voluntarily (both mean scores 4.1). The experience of stigmatisation was low (mean score 2.1) and that of reintegrative shaming was neutral (mean score 3.3), just as the score on awareness of broken rules and norms (2.8). This sample of offenders on average indicated that the victim accepted the apology (mean score 3.9).
  Looking at the psychological outcome variables, offenders scored rather neutral. The scores for empathy, guilt, shame, responsibility-taking and awareness of moral failure were all between 2.7 and 3.3. Only awkwardness to meeting the victim in the future was lower, at 2.2.

  3.3 Correlation analyses and observations

  3.3.1 Empathy

  The correlational analyses showed that offenders’ overall satisfaction with the mediation encounter, the offender experiencing to be taken seriously by the victim, experiencing equal possibilities to speak during VOM and having a say in the outcome were all positively and significantly correlated to empathy. The fundamental condition professional competencies of the mediator also positively correlates with empathy. Table 4 offers an overview of all correlations.
  Turning to the qualitative data, we coded different observations that fit the correlations. In cases in which the offenders showed a (small) increase in empathy, the mediator observed that: ‘She [the victim] actually wanted to get to know the offender behind the monster and give him a second chance’ and ‘both the victim and the offender were interested in each other stories,’ which suggests that the victim took the offender seriously. Remarkably, in a case in which the offender showed a small decline in his feelings of empathy, the mediator observed that ‘the minor victim did not really care about it [the offence] anymore’ and did not show to take more empathy or perspective. The offender in this case also did not feel to be taken seriously by the victim, which could be due to the victim not caring about the offence anymore. In this same case, the mediator observed that ‘the mother of the victim accepted the apology, but could not forgive the offender.’ This might explain why this offender felt as if the victim did not experience his apology as sincere and why this offender scored neutral on the questions to what extent the victim accepted and appreciated the apology. This is in accordance with the positive correlations found between acceptance of a sincere apology and higher empathy.

  3.3.2 Guilt

  Three elements were positively correlated with feelings of guilt: satisfaction, being taken seriously by the victim and being made aware of the broken rules and norms. Previous research (Jonas-van Dijk et al., 2022b) already indicated that offenders who participated in VOM had higher feelings of guilt in comparison to non-participating offenders, not because of increases in their feelings of guilt but because of consolidation of these feelings where guilt decreased among non-participants. In the observed cases, we did not see an increase in the offenders’ feelings of guilt either. The changes that were observed were very small decreases. At an item level, we saw that this was not due to offenders feeling less guilty or remorseful but due to them feeling less tensed and afraid. This could be considered a positive outcome of VOM as well. In two cases, the mediators also observed offenders to be relieved, which might indicate that offenders felt less tensed and afraid. As one mediator indicated, both parties were, as it were, ‘overwhelmed’ by the positive effect of the mediation encounter. This turned out to be a huge relief for all, because as a result of the incident there was fear of facing each other and fear of anger and repetition. Only one offender showed a significant increase in his feelings of guilt as he scored low on the pre-test and high on the post-test.
  What is noteworthy is that in one case it seemed helpful for the VOM process that the offender showed to be feeling guilty: ‘There was immediate full acknowledgment of guilt by the offender and a willingness to compensate all damage suffered. The case was therefore concluded relatively easily with a settlement agreement.’ In a different case, in which both parties were suspects, the mediator observed that only one party acknowledged guilt, whereas the other party did not. Remarkably, the mediator reported that

  Table 4 Correlations between perceived fundamental conditions, working mechanisms and psychological outcomes among offenders who participated in VOM (n = 55)

  	Empathy	Guilt	Shame	Responsibility-taking	Moral failure	Awkward-ness meeting victim
  Required conditions
  Preparation	0.24	0.22	0.18	0.20	0.15	−0.32*
  Voluntariness	0.21	0.22	0.12	0.29*	0.13	−0.23
  Professional competencies mediator	0.30*	0.25	0.20	0.04	0.16	0.03
  Working mechanisms
  Satisfaction	0.36*	0.31*	0.26	0.21	0.30*	−0.17
  Being treated with respect	0.21	0.11	0.002	0.15	0.11	−0.23
  Being able to speak freely	0.24	0.20	0.16	0.09	0.16	−0.03
  Feeling safe	0.19	0.20	0.09	0.11	0.03	−0.17
  Being listened to	0.24	0.08	0.15	0.10	0.13	−0.24
  Taken seriously by victim	0.55**	0.30*	0.26	0.36**	0.36*	−0.46**
  Having equal possibilities to speak	0.32*	0.28	0.23	0.14	0.26	−0.07
  Have a say in outcome	0.31*	0.22	0.005	0.16	0.28	−0.38*
  Perceived acceptance of the apology by the victim	0.32**	0.15	0.08	0.28*	0.22	−0.53*
  Stigmatising	−0.21	−0.07	−0.05	−0.02	−0.09	0.41**
  Reintegrative shaming	0.29*	0.19	0.08	0.20	0.26	−0.34*
  Increased awareness broken rules and norms	0.34*	0.30*	0.34*	0.37**	0.36**	−0.02

  Note: *correlation is significant at the 0.05 level (2-tailed).
  **correlation is significant at the 0.01 level (2-tailed).

  There was some discussion with V2 [the offender who did acknowledge guilt] about the content of the first proposed agreement, in particular about the request to the public prosecutor and to what extent he seemed more guilty than the other party.

  Based on these quotes, it seems important that the right person(s) acknowledge(s) guilt for the quality of the VOM process.

  3.3.3 Shame

  Feelings of shame only correlated positively with being made aware of the broken rules and norms. In the cases observed by the mediators, it was hard to find specific examples of how the mediation encounter could influence feelings of shame. First, we saw that offenders who were already highly ashamed of their wrongdoings at the start of the VOM process continued to score high on these feelings afterwards. The same applied to offenders who scored low on feelings of shame, since they continued to score low. In addition, the mediators also did not report anything specific about how ashamed offenders showed to be or how this changed during VOM. Lastly, in all cases, the mediators observed that the broken rules and norms were discussed, which might explain why nothing specific can be said about the variable shame.

  3.3.4 Responsibility-taking

  Taking responsibility was positively correlated with being taken seriously by the victim and acceptance of the apology. One mediator observed that:

  The offender took responsibility, did not think it was chic of herself, nor did she think it was a good example for her child. But to call it that way happened in a very quick breath, which made the victim doubt the credibility.

  In this case, the offender did not feel to be always taken seriously or that the victim accepted her apology. Although the mediator indicated that the offender took responsibility, the scores on the questionnaire did not correspond to this observation: The offender did not show to feel more responsible after participation; possibly because the victim doubted their credibility. This fits the correlation found.
  Being made aware of the broken rules and norms also positively correlated with responsibility-taking. In one case, the offender showed a clear increase in his feelings of responsibility, which might be due to the offender becoming more aware of what rules and norms have been broken, as shown in the observation from a mediator: ‘the fact that the act was qualified as a threat was an eye-opener and that behaviour was subsequently also qualified as incorrect.’ This quote indicates that the due to the conversation the offender became aware that her behaviour was wrong and not in accordance with existing rules and norms.
  Interestingly, responsibility-taking is the only psychological variable that positively correlated with experiencing to participate voluntarily. In eight of the observed cases, the offenders scored high on voluntariness. Only one offender scored neutral. The mediator in this case reported the following about the offenders’ intentions:

  Here too, the dividing line between offender and victim was thin. The suspect herself believed that she [the victim] should not have filed a report, which made her now a suspect. Might as well have been the other way around.

  Since the offender in this case felt as if she was a victim too, and that in her opinion the victim should be treated as offender as well, this might have forced her to participate. Noteworthy is that this mediation was ‘very difficult, minimal visible result, but for both [parties] this was the maximum achievable,’ which might be related to the offender feeling not to be participating voluntarily. Possibly, this has negatively impacted the dialogue.

  3.3.5 Moral failure

  Being aware of moral failure positively correlated with being taken seriously by the victim, perceived acceptance of the apology and being made aware of the broken rules and norms. In one case in which the offender’s awareness of moral failure increased, the mediator observed that:

  He [the offender] said he was very sorry and looked at her [the victim]. The victim saw that he meant it. This was reinforced by the fact that he himself asked what she needed from him. He also immediately refunded the money to her.

  In this case, both the mediator and the offender perceived that the victim experienced the apology as sincere and accepted it, what underpins the found correlation.
  In the observations, we saw that in cases in which offenders became more aware of their moral failure the mediator specifically observed that due to the story of the victim, the offender became aware of the impact of his actions: ‘He only now heard about the impact of his actions and then had to cry’ and ‘The offender did not expect that what happened had so much impact on the victim.’ This also indicates that hearing the story of the victim can contribute to becoming more aware of one’s actions.
  We observed one offender to be less aware of her moral failing after participation, which corresponded with the observation of the mediator:

  Parties mainly wanted to prove themselves right and tell the other person what she did wrong. Little self-reflection, but this was more due to powerlessness and incompetence, than due to unwillingness.

  When offenders are not willing or able to self-reflect, it might also be harder to make them aware of the broken rules and norms, which might explain in this case why the offender became less aware of her moral failing. In this same case ‘… the victim doubted the credibility [of the offender]’ and ‘the victim reluctantly accepted the apology as it was quickly overturned by defence,’ which could have resulted in the offender experiencing the victim to not take her seriously and to not accept the apology. This is also in accordance with the found correlations.

  3.3.6 Awkwardness meeting the victim

  Interestingly, most of the correlations between the working elements and awkwardness to meet the victim were negative; preparation, being taken seriously by the offender, having a say in the outcome, reintegrative shaming, and the acceptance of the apology, all negatively correlated with this psychological variable. Stigmatising, on the other hand, correlated positively with awkwardness to meet the victim again.
  In multiple cases, mediators noticed that the relationship between parties was improved, which also might have resulted in the offender feeling less awkward to meet the victim again. In one case the mediator said:

  The best proof of coming closer together was that, while writing down the agreement, they [the victim and the offender] drank a cup of coffee together and talked further in a relaxed atmosphere.

  In another case, the offender still scored high on awkwardness for meeting the victim again in the future. Typical in this case is that parties showed to have little confidence in each other: ‘the victim had no intention of getting a different opinion [about the offender],’ which might explain the feelings of stigmatisation of this offender. This underpins the positive correlation between feeling stigmatised and awkwardness to meet the victim again.

  3.4 The influence of the fundamental conditions

  Interestingly, most fundamental conditions did not correlate with the psychological outcome variables. However, the fundamental conditions might impact the working mechanisms directly or the relationship between the working mechanisms and the psychological outcome variables.
  To examine whether the fundamental conditions predict the working mechanisms, we again did correlational analyses (Table 5). These analyses showed that preparation and the professional competencies of the mediator have a positive correlation with most of the working mechanisms: satisfaction, being treated with respect, being able to speak freely, feeling safe, being listened to, having equal possibilities to speak, perceived acceptance of the apology, and reintegrative shaming. Preparation is also positively correlated with being taken seriously by the victim. This means that when these fundamental conditions are adhered to in a greater extent, the working mechanisms are also present to a higher extent, according to the experience of the offenders. In addition, two negative correlations were found with the working mechanism stigmatisation. If offenders reported more strongly that they felt well prepared and perceived to be taken more seriously by the neutral mediator, they experienced to be stigmatised to a lesser extent. Finally and importantly, the perceived voluntariness of participation did not correlate with any of the working mechanisms.
  To examine whether the fundamental conditions work as a moderator and impact the association between the working mechanisms and the psychological outcome variables, multiple linear regression analyses were used. The psychological outcome variables were the dependent variables. The working mechanisms and conditions were the predictor variables and were all centred. The interaction terms between the four conditions and mechanisms were added in the analyses. This means that we looked at 216 interactions of which 4 were significant (p < 0.05) and 7 were marginally significant (p = 0.5 − 0.06). Only 5 per cent of the interactions were (close to being) significant, and hence there is a high risk of type 1 error (false positives). However, we do think it is well worth reporting these outcomes.

  Table 5 Correlations between the fundamental conditions and working mechanisms

  	Preparation	Voluntariness	Professional competencies mediator
  Satisfaction	.40**	.16	.42**
  Being treated with respect	.40**	.19	.55**
  Feeling safe	.61**	.14	.51**
  Being listened to	.49**	.08	.51**
  Taken seriously victim	.44**	.23	.29
  Having equal possibilities to speak	.42**	.18	.63**
  Have a say in the outcome	.27	.28	.25
  Perceived acceptance of the apology by the victim	.50**	.15	.34*
  Stigmatising	-.48**	-.04	-.34*
  Reintegrative shaming	.41**	.11	.33*
  Increased awareness broken rules and norms	.07	-.05	.07

  In the correlational analyses above (Table 5), it became clear that voluntariness as one of the three conditions did not correlate with any of the working mechanisms. Interestingly, almost all, that is ten of the eleven (marginal) significant interactions we observed, involved the condition of voluntariness (Table 6). The results show that for six of the eleven (marginal) significant interactions, the correlations were positively moderated by the degree of voluntariness that offenders experienced: between feeling listened to and empathy, between feeling safe and empathy, between being able to speak freely and guilt, between feeling safe and guilt, between reintegrative shaming and responsibility, and between feeling safe and responsibility-taking. This means that the more offenders perceived to be able to participate voluntarily, the stronger the association between these working mechanisms and the psychological outcome variables. This matched our expectations.
  The other five interactions were not as expected. The correlations between being taken seriously by the victim and empathy, between having a say in the outcome and empathy, between being taken seriously by the victim and guilt, and between being made aware of the broken rules and norms and guilt were all negatively moderated by voluntariness. This means that when offenders to a greater extent felt they could participate voluntarily, the correlation between these working mechanisms and the psychological outcome variables became weaker. This also applies to the correlation between having a say in the outcome and responsibility-taking, since this correlation is negatively moderated by preparation.

• 4 Discussion

  Recent research suggests that offenders who participated in VOM show a number of beneficial psychological changes afterwards that have been linked to a lower risk of reoffending in previous research (Hosser et al., 2008; Vaish, Carpenter & Tomasello 2016). However, it has not been examined to date which mechanisms of the VOM might be related to these psychological outcomes. This research aimed to examine this. Research that points to key mechanisms of VOM and how these are related to psychological outcomes might help considerably to further best practices and could be used to optimise VOM in the future.

  Table 6 (Marginally) significant outcomes of the multiple linear regression analyses with the interaction terms between the fundamental conditions, working mechanisms and psychological outcomes

  Fundamental condition	Psychological outcome	Working mechanism	Nonstandardized B	t	p	CI	Slope low (−1 SD) fundamental condition	Slope high (+1 SD) fundamental condition
  Voluntariness	Empathy	Feeling safe	0.60	1.90	0.058	−0.20-1.22	−0.41	0.37
  		Feeling listened to	0.50	1.90	0.058	−0.02-1.01	−0.16	0.48
  		Taken seriously by victim	−0.36	−2.29	0.026	−0.68 to −0.04	0.56	0.13
  		Say in outcome	−0.36	−1.89	0.059	−0.73-0.01	0.48	−0.003
  	Guilt	Speak freely	0.53	1.89	0.059	−0.20-1.09	−0.30	0.39
  		Feeling safe	0.60	2.25	0.025	0.08-1.13	−0.31	0.46
  		Taken seriously by victim	−0.30	−1.99	0.046	−0.60 to −0.01	0.26	−0.09
  		Awareness about broken rules and norms	−0.37	−2.01	0.045	−0.73 to −0.01	0.43	−0.01
  	Responsibility-taking	Feeling safe	0.56	1.88	0.061	−0.03-1.15	−0.34	0.39
  		Reintegrative shaming	0.53	1.96	0.50	0.000-1.06	−0.10	0.59
  Preparation	Responsibility-taking	Say in outcome	−0.47	−1.97	0.050	−0.93 to −0.001	0.31	−0.23

  4.1 Working mechanisms of the mediator encounter

  In line with previous research (Jonas-van Dijk et al., 2022a), the present research indicates that a constructive interaction with the victim during VOM is key for the beneficial psychological impact VOM can have on offenders. It is important for offenders that the victim takes them seriously, as this is related to higher victim empathy, higher feelings of guilt, higher responsibility-taking and higher awareness of moral failure. In line with this, it is also important that offenders experience the victim to accept their apology and perceive it as sincere. Such a receptive and accepting response by the victim might be hard to bring about, however, as it is the victim who decides to accept the apology and whether or not the apology is perceived as sincere (Bonensteffen et al., 2020). The results of this study indicate that it is most conducive for offenders when the victim has an open and cooperative attitude – although this by no means can be a requirement or demand for victims to participate.
  The present research confirms that VOM can function as a learning process. Being made more aware of the broken rules and norms during VOM is related to more victim empathy, higher feelings of guilt and shame, more responsibility-taking and a higher moral failure. Considering that we found this correlationwith five of the six examined variables, this might serve as indication that talking about the broken rules and norms is an important element of VOM. Research of Abrams et al. (2006) and Miller and Hefner (2015) also indicate that moral learning of the offender can lead to feelings of remorse, accountability, shame and guilt. Our qualitative findings corroborate this: two clear observations indicate that due to the story of the victim, the offender became more aware of the impact of the crime. This is in accordance with previous research in which it was shown that talking to the victim makes offenders more aware of the impact of the crime (Choi et al., 2011; Lauwaert & Aertsen, 2016; Marsh & Maruna, 2016).
  Unexpectedly, no correlation was found between reintegrative shaming and stigmatising and feelings of shame. However, according to this research, reintegrative shaming is linked positively and significantly to more victim empathy. Since in previous research empathy has been related to a lower risk of reoffending (Schalkwijk, Stams, Stegge, Dekker & Peen, 2016), this suggests that one of the ways reintegrative shaming during VOM might influence the risk of reoffending is through victim empathy. This might be a reason for mediators to safeguard (and facilitate as best as they can) a reintegrative shaming atmosphere during an encounter. Future research could examine which elements of mediation are experienced as reintegrative. Another reason to foster reintegrative shaming and minimise stigmatising during the encounter is that more reintegrative shaming and less stigmatisation seem to be related to feeling less awkward to meet the victim in the future. When this awkwardness is reduced, it might be easier to talk to that person when a problem arises again in the future. This might also prevent further escalations and, therefore, reoffending.
  One last important finding of this research is the change that was observed in offenders’ feelings of guilt. Previous research already indicated that offenders who participated in VOM showed higher feelings of guilt afterwards compared to offenders who did not participate (Jonas-van Dijk et al., 2022b). However, this was not due to an increase of feelings of guilt but due to consolidation of those feelings (whereas the level of guilt declined for offenders who did not participate in mediation). The data of the case study show that most offenders did show a small decrease in their feelings of guilt. However, this decline was mostly due to them feeling less tensed and bad, which could be seen as a positive outcome of VOM. Offenders are still aware that they are guilty but seem to feel less negative about it.

  4.2 Impact of the fundamental conditions

  A unique feature of the research reported here is that it examined the impact of fundamental conditions of the VOM process on (the relationship between) the working mechanisms and psychological outcomes for offenders. Consistent with our reasoning in the introduction, this research suggests that adherence to the fundamental conditions of the VOM process is likely to affect the working mechanisms of the VOM encounter itself. That is, when offenders feel better prepared and experience the mediator to be neutral and to take them seriously, they also report higher experience of multiple working mechanisms.
  Importantly, we observed that the fundamental condition perceived voluntariness of participation did not have a direct relationship with any of the working mechanisms. It did however operate as a moderator for some of the relationships between working mechanisms and psychological outcomes. When offenders reported more strongly that their participation was voluntary, this strengthened the association between feeling listened to and empathy, between feeling safe and empathy, between being able to speak freely and guilt, between feeling safe and guilt, between reintegrative shaming and responsibility-taking, and between feeling safe and responsibility-taking.
  Unexpectedly, however, for the other associations between working mechanisms and psychological outcomes a lower voluntariness seemed to strengthen the association. This was true for the association between being taken seriously by the victim and empathy and guilt, and between being made aware of the broken rules and norms and guilt. What might explain these unexpected findings is that when offenders feel that their participation was not entirely voluntary this may have lowered their positive expectations regarding the interaction with their victim during the VOM encounter. It might be that the actual interaction with the victim during the encounter positively surpassed these negative expectations, which as a result may have boosted the association between these working mechanisms and psychological outcomes. To examine this speculation, future research could ask offenders after their participation to what extent the actual encounter was (in)consistent with their expectations and in what way.

  4.3 Limitations and implications for future research

  Although this research offers both practical and theoretical insights, it is important to interpret these outcomes with a number of shortcomings in mind. First, no causal relationship between any of the mechanisms and psychological outcomes can be drawn from the results. The sample is too small to infer any causality and, thus, the analysis might be underpowered. Therefore, based on this research it is only able to offer and interpret patterns.
  It is also not clear whether the working mechanisms explain the psychological outcomes or whether it is other factors that do so, such as the willingness of offenders to desist from crime or motivations to make things right. For example, it could be that offenders who participate in mediation are highly motivated to make things right and, therefore, take more responsibility or feel more guilt. High responsibility-taking or feelings of guilt are then not due to working mechanisms but due to a selection bias.
  More research is needed to understand how mechanisms of VOM can cause psychological outcomes within offenders. We would, therefore, suggest using qualitative research designs to more specifically examine the working mechanisms by means of (participant) observation. For example, proper preparation seems important, but what entails proper preparation? Interviewing offenders and observing mediation encounters would offer valuable insights into what constitutes these working mechanisms and fundamental elements. With outcomes of such a qualitative research design, valid and reliable scales could be formed which could subsequently be used to examine the impact of working mechanisms in a larger sample of offenders. If patterns are reoccurring in multiple studies, best practices could eventually be drawn. Therefore, this research offers unique data, insights and starting points for future research, since it is the first research that tries to examine which mechanisms of VOM might be related to psychological outcomes in offenders.
  What should also be taken into account is that the unravelled elements were mainly based on the offenders’ experience. For example, offenders were asked to what extent mediators were neutral. However, this was not confirmed by a third party or the researchers. This means that a bias might have impacted the outcomes: possibly, offenders were in general positive about mediation and, therefore, experienced the mediator to be neutral. In future research, a researcher could observe mediation encounters as a third party to confirm the offenders’ experiences.
  It would also be better to observe and code during the encounter. However, this could influence the conversation, since participants might consciously experience that they are observed and, therefore, not feel comfortable to speak freely. This could be solved by video-recording the encounter, which would enable a researcher to code the encounter while observing and give an opportunity to include a second observer to check for inter-rater reliability.
  Another important factor that should be examined in future research is the impact of the encounter on the victim. In this research, the focus was on the offender. Therefore, the patterns found and implications drawn from the findings only account for the offender. It is unclear what the working mechanisms of VOM are for the victim. Future research could adopt a broader study design, in which both victims and offenders are observed and for both parties the psychological changes are measured.

  4.4 Conclusion

  With these limitations in mind, it can be concluded that this research has found indications that three working mechanisms might be related to offenders’ psychological outcomes after their participation in VOM. The first mechanism is a constructive dialogue with an open and cooperative victim, in which the victim takes the offender’s perspective, shows empathy and the offender is being taken seriously by the victim. The second mechanism to take into consideration is maximise reintegrative shaming and minimise stigmatising. For example, this can be strived for during the pre-meetings before the actual encounter by making sure that victims are not (solely) focused on revenge. The last mechanism is using VOM as a learning process. For VOM to be a learning process, it is important that the victim is able to explain the actual impact of the offence and offenders become (more) aware of their moral failure. When practitioners take these mechanisms into account, this might positively influence the psychological impact of VOM on offenders and, hence, might reduce the risk of reoffending.
  This research also showed that three fundamental conditions might have a direct influence on the working mechanisms or indirect influence on the relationships between the mechanisms and psychological outcomes: a neutral mediator who takes the offender seriously, proper preparation of parties and voluntary participation. Future research could use the recommendations given to further examine the relationship between the fundamental conditions, the working mechanisms and the psychological outcomes for offenders as well as victims – which may help to develop further restorative justice practices.

References

• Abrams, L.S., Umbreit, M. & Gordon, A. (2006). Young offenders speak about meeting their victims: implications for future programs. Contemporary Justice Review, 9(3), 243-256. doi: 10.1080/10282580600827835.

• Baldry, A.C. (1998). Victim-offender mediation in the Italian Juvenile Justice System: the role of the social worker. British Journal of Social Work, 28(5), 729-744. doi: 10.1093/oxfordjournals.bjsw.a011388.

• Bazemore, G. & Umbreit, M. (2001). A comparison of four restorative conferencing models. Juvenile Justice Bulletin. Office of Juvenile Justice and Delinquency Prevention, U.S. Department of Justice. Retrieved from https://www.ojp.gov/pdffiles1/ojjdp/184738.pdf

• Bergseth, K.J. & Bouffard, J.A. (2013). Examining the effectiveness of a restorative justice program for various types of juvenile offenders. International Journal of Offender Therapy and Comparative Criminology, 57(9), 1054-1075. doi: 10.1177/0306624X12453551.

• Bernburg, J.G. & Krohn, M.D. (2003). Labeling, life chances, and adult crime: the direct and indirect effects of official intervention in adolescence on crime in early adulthood. Criminology, 41(4), 1287-1318. doi: 10.1111/j.1745-9125.2003.tb01020.x.

• Bonensteffen, F., Zebel, S. & Giebels, E. (2020). Sincerity is in the eye of the beholder: using eye tracking to understand how victims interpret an offender’s apology in a simulation of victim–offender mediation. Frontiers in Psychology, 11. doi: 10.3389/fpsyg.2020.00835.

• Boriboonthana, Y. & Sangbuangamlum, S. (2013). Effectiveness of the restorative justice process on crime victims and adult offenders in Thailand. Asian Journal of Criminology, 8(4), 277-286. doi: 10.1007/s11417-013-9160-8.

• Braithwaite, J. (1989). Crime, shame and reintegration. Cambridge, UK: Cambridge University Press.

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

• Braithwaite, J., Braithwaite, V. & Ahmed, E. (2018). Reintegrative shaming 1. In The Essential Criminology Reader (pp. 286-296). Henry, S, New York.

• Choi, J.J. (2008). Opening the ‘black box’: a naturalistic case study of restorative justice. (Doctoral dissertation). University of Kansas. ProQuest LLC, Michigan

• Choi, J.J., Bazemore, G. & Gilbert, M.J. (2012). Review of research on victims’ experiences in restorative justice: implications for youth justice. Children and Youth Services Review, 34(1), 35-42. doi: 10.1016/j.childyouth.2011.08.011.

• Choi, J.J., Green, D.L. & Gilbert, M.J. (2011). Putting a human face on crimes: a qualitative study on restorative justice processes for youths. Child and Adolescent Social Work Journal, 28(5), 335-355. doi: 10.1007/s10560-011-0238-9.

• Claessen, J. & Roelofs, K.J.M. (2020). Herstelrecht(voorzieningen) en mediation in strafzaken [Restorative justice (services) and mediation within criminal cases]. In J. Boksem, P.A.M. Mevis, D.J.M.W. Paridaens, C. Waling & H.D. Wolswijk (eds.), Handboek Strafzaken – online. Wolters Kluwer. The Netherlands

• Claessen, J., Zeles, G., Zebel, S. & Nelen, H. (2015). Bemiddeling in strafzaken in Maastricht II. Onderzoek naar de samenhang tussen bemiddeling en recidive [Mediation within criminal cases in Maastricht II. Examining the relation between mediation and recidivism]. Nederlands Juristenblad, 29, 2015-2025.

• De Hooge, I.E., Zeelenberg, M. & Breugelmans, S.M. (2007). Moral sentiments and cooperation: differential influences of shame and guilt. Cognition and Emotion, 21(5), 1025-1042.

• de Mesmaecker, V. (2013). Victim-offender mediation participants’ opinions on the restorative justice values of confidentiality, impartiality and voluntariness. Restorative Justice, 1(3), 334-361. doi: 10.5235/20504721.1.3.334.

• Dhami, M.K. (2016). Apology in victim–offender mediation. Contemporary Justice Review: Issues in Criminal, Social, and Restorative Justice, 19(1), 31-42. doi: 10.1080/10282580.2015.1101686.

• Eekhout, I. (n.d.). Multiple imputation. Retrieved August 16, 2022, from https://www.missingdata.nl/missing-data/missing-data-methods/multiple-imputation/.

• Fellegi, B. (2008). De Impact van herstelrecht verklaard. De wisselwerking tussen moraliteit, neutralisatie, schaamte en sociale bindingen. Tijdschrift Voor Herstelrecht, 8(3), 37-51.

• Gausel, N., Vignoles, V.L. & Leach, C.W. (2016). Resolving the paradox of shame: differentiating among specific appraisal-feeling combinations explains pro-social and self-defensive motivation. Motivation and Emotion, 40(1), 118-139. doi: 10.1007/s11031-015-9513-y.

• Gerkin, P.M. (2009). Participation in victim—offender mediation: Lessons learned from observations. Criminal Justice Review, 34(2), 226-247. doi: 10.1177/0734016808325058

• Hansen, T. & Umbreit, M. (2018). State of knowledge: four decades of victim-offender mediation research and practice: the evidence. Conflict Resolution Quarterly, 36(2), 99-113. doi: 10.1002/crq.21234.

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

• Hausmann, C., Jonason, A. & Summers-Effler, E. (2011). Interaction ritual theory and structural symbolic interactionism. Symbolic Interaction, 34(3), 319-329. doi: 10.1525/si.2011.34.3.319.

• Hosser, D., Windzio, M. & Greve, W. (2008). Guilt and shame as predictors of recidivism: a longitudinal study with young prisoners. Criminal Justice and Behavior, 35(1), 138-152. doi: 10.1177/0093854807309224.

• Jonas-van Dijk, J., Zebel, S., Claessen, J. & Nelen, H. (2020). Victim–offender mediation and reduced reoffending: gauging the self-selection bias. Crime and Delinquency, 66(6-7), 949-972. doi: 10.1177/0011128719854348.

• Jonas-van Dijk, J., Zebel, S., Claessen, J. & Nelen, H. (2022a). How can the victim-offender mediation process contribute to a lower risk of reoffending? A synthesis literature review. The International Journal of Restorative Justice, 6(2), 207-234. doi: 10.5553/TIJRJ.000144.

• Jonas-van Dijk, J., Zebel, S., Claessen, J. & Nelen, H. (2022b). The psychological impact of participation in victim-offender mediation on offenders: evidence for increased compunction and victim empathy. Frontiers in Psychology. doi: 10.3389/fpsyg.2021.812629.

• Lauwaert, K. & Aertsen, I. (2015). Desistance and restorative justice mechanisms for desisting from crime within restorative justice practices. European Forum for Restorative Justice, Leuven.

• Lauwaert, K. & Aertsen, I. (2016). With a little help from a friend: desistance through victim-offender mediation in Belgium. Restorative Justice- an International Journal, 4(3, SI), 345-368. doi: 10.1080/20504721.2016.1245913.

• Little, R.J.A. & Rubin, D.B. (1989). The analysis of social science data with missing values. Sociological Methods & Research, 18(2-3), 292-326. doi: 10.1177/0049124189018002004.

• Marsh, B. & Maruna, S. (2016). Desistance and restorative justice: learning from success stories of Northern Ireland’s Youth Justice Agency. Restorative Justice, 4(3), 369-387. doi: 10.1080/20504721.2016.1243855.

• Meléndez, A. (2020). The role of criminogenic needs and emotions in restorative justice: Offenders’ experiences in victim–offender mediation. European Journal of Probation, 13(2). doi: 10.1177/2066220320948374.

• Miller, S. & Hefner, M. (2015). Procedural justice for victims and offenders?: exploring restorative justice processes in Australia and the US. Justice Quarterly, 32(1), 142-167. doi: 10.1080/07418825.2012.760643.

• Ministry of Justice and Safety (2021). Informatieblad Herstelrecht na een strafbaar feit: mediation in Strafzaken en Herstelbemiddeling. [Information sheet Restorative Justice after an offence: mediation in criminal cases and restorative mediation.] Retrieved from https://www.reclassering.nl/media/1955/211210_informatiebladplusherstelrechtplusnapluseenplusstrafbaarplusfeit-1.pdf.

• Pabsdorff, M.-L., Rytterbro, L.-L., Sambou, S. & Uotila, E. (2011). Victim-offender mediation: observations from Scandinavia. Oñati Socio-Legal Series, 1(2).

• Rossner, M. (2011). Emotions and interaction ritual: a micro analysis of restorative justice. The British Journal of Criminology, 51(1), 95-119. doi: 10.1093/bjc/azq075.

• Rypi, A. (2016). The feeling rules of victim offender mediation. International Journal of Work Organisation and Emotion, 7(2), 83-97. doi: 10.1504/IJWOE.2016.078074.

• Schalkwijk, F., Stams, G.J., Stegge, H., Dekker, J. & Peen, J. (2016). The conscience as a regulatory function: empathy, shame, pride, guilt, and moral orientation in delinquent adolescents. International Journal of Offender Therapy and Comparative Criminology, 60(6), 675-693. doi: 10.1177/0306624X14561830.

• Shapland, J. (2006). Restorative justice in practice: the second report from the evaluation of three schemes. University of Sheffield, Centre for Criminological Research, Sheffield, UK.

• Shapland, J., Atkinson, A., Atkinson, H., Chapman, B., Dignan, J., Howes, M., Johnstone, J., Robinson, G. & Sorsby, A. (2007). Restorative justice: the views of victims and offenders. Ministry of Justice Research Series, 3, 1-60.

• Shapland, J., Atkinson, A., Atkinson, H., Dignan, J., Edwards, L., Hibbert, J., Howes, M., Johnstone, J., Robinson, G. & Sorsby, A. (2008). Does restorative justice affect reconviction? The fourth report from the evaluation of three schemes. Ministry of Justice, London

• Sherman, L., Strang, H., Angel, C., Woods, D., Barnes, G., Bennett, S., & Inkpen, N. (2005). Effects of face-to-face restorative justice on victims of crime in four randomized, controlled trials. Journal of Experimental Criminology, 1(3), 367-395. doi: 10.1007/s11292-005-8126-y.

• Stewart, L., Thompson, J., Beaudette, J.N., Buck, M., Laframboise, R. & Petrellis, T. (2018). The impact of participation in victim-offender mediation sessions on recidivism of serious offenders. International Journal of Offender Therapy and Comparative Criminology, 62(12), 3910-3927. doi: 10.1177/0306624X17752274.

• Suzuki, M. & Jenkins, T. (2023). Apology–forgiveness cycle in restorative justice, but how? International Review of Victimology, 29(2), 259-276. doi: 10.1177/026975802210799.

• Szmania, S.J. (2006). Mediators’ communication in victim offender mediation/dialogue involving crimes of severe violence: an analysis of opening statements. Conflict Resolution Quarterly, 24(1), 111-127. doi: 10.1002/crq.161.

• Tangney, J.P., Stuewig, J. & Martinez, A.G. (2014). Two faces of shame: the roles of shame and guilt in predicting recidivism. Psychological Science, 25(3), 799-805. doi: 10.1177/0956797613508790.

• Thai, M., Wenzel, M. & Okimoto, T.G. (2021). Turning tables: offenders feel like ‘victims’ when victims withhold forgiveness. Personality and Social Psychology Bulletin, 49(2), 233-250. doi: 10.1177/01461672211062401.

• Vaish, A., Carpenter, M. & Tomasello, M. (2016). The early emergence of guilt-motivated prosocial behavior. Child Development, 87(6), 1772-1782. doi: 10.1111/cdev.12628.

• van Ginkel, J.R., Linting, M., Rippe, R.C.A. & van der Voort, A. (2020). Rebutting existing misconceptions about multiple imputation as a method for handling missing data. Journal of Personality Assessment, 102(3), 297-308. doi: 10.1080/00223891.2018.1530680.

• Villanueva, L., Jara, P. & García-Gomis, A. (2014). Effect of victim-offender mediation versus dispositions on youth recidivism: the role of risk level. Journal of Forensic Psychology Practice, 14(4), 302-316. doi: 10.1080/15228932.2014.950508.

• White, I.R., Royston, P. & Wood, A.M. (2011). Multiple imputation using chained equations: issues and guidance for practice. Statistics in Medicine, 30(4), 377-399. doi: 10.1002/sim.4067.

• Zebel, S. (2012). Een quasi-experimenteel onderzoek naar de effecten van de Nederland slachtoffer-dadergesprekken [A quasi-experimental study into the effects of Dutch victim-Meetings]. Bemiddeling Na Strafbare Feiten: De Nederlandse Slachtoffer-Dadergesprekken, 21-44.

• Zehr, H. (2015). The little book of restorative justice: revised and updated. Intercourse, PA: Good Books.

Noten

• 1 The two minors received parental permission for participating.

• 2 Given that there were thirteen observations and we used these data of mediators qualitatively, factor analysis and reliability analysis are not reported for the scales filled in by the mediators.

• 3 The online platform Qualtrics was used to register the responses.