Transforming prison culture: supportive norms enhance prison officers’ well-being and prosociality toward detained people

The literature on social norms and prosociality could be applied to the prison context to generate insights clarifying the influence of prison institutional punitive norms on officers’ orientation, attitudes toward detained people and ultimately on their well-being. Social norms are unwritten rules and principles shared by ingroup members regulating behaviors distinguished between norms coding the perception of what others do (descriptive norms) and what is approved or disapproved (injunctive norms) (Cialdini and Trost, 1998). Prosocial norms have been shown to increase charitable behaviors (Lay et al., 2020), empathy and positive attitudes toward the outgroup (Murrar et al., 2020; Tarrant et al., 2009) and to reduce prejudice and discrimination toward peers with mental health issues (Silke et al., 2017). However, only a few papers addressed the effects of prosocial norms on well-being. A theoretical article envisioned that prosocial norms might improve happiness and well-being by pushing them to be prosocial (Helliwell, 2014). An empirical study showed that social norms prohibiting workplace mistreatment can increase work engagement and reduce burnout by reducing workplace conflict (Sliter et al., 2014).

The related literature on prosociality and empathy might help to fill the literature gap conceptually. Prosocial attitudes and behaviors increase eudemonic well-being by meeting needs for meaning (Midlarsky, 1991), self-actualization and fulfillment (Curry et al., 2018; Hui et al., 2020) and competence and relatedness (Aknin et al., 2013). Also, the literature on empathy depicts a more complex and ambivalent relationship between empathy, prosociality and well-being. On the one end, empathy could foster a sense of meaning, job and life satisfaction (Supervía et al., 2023; Zenasni et al., 2012). On the other, the empathic emotional closeness with (suffering) others can cause emotional exhaustion, compassion fatigue and burnout (Zenasni et al., 2012). Empirical findings support such theoretical ambivalence. Some suggest an overall positive relationship between empathy and well-being (Bourgault et al., 2015; Caro et al., 2017), whereas others found that empathic concern and emotional closeness might increase emotional exhaustion (Altmann and Roth, 2021; MacArthur et al., 2021).

We conducted a mixed correlational-experimental study to investigate if prisoner-supportive social norms could improve officers’ behaviors, attitudes, professional orientations, empathy and relationships toward detained people, ultimately fostering officers’ well-being. In the correlational section, we hypothesized a mediational path for which prisoner-supportive norms would be associated with improved attitudes (i.e. hostility and punitive–educative orientation) and increased emotional closeness (i.e. empathy, interpersonal and intergroup emotional closeness) toward detained people. In turn, we hypothesized that the improved attitudes would positively relate to officers’ psychological and organizational well-being (i.e. psychological distress, job-related burnout and satisfaction). We anticipated a similar effect of emotional closeness on officers’ well-being. However, this hypothesis is exploratory due to the mixed findings in the literature regarding the relationship between empathy and well-being.

The experimental part of the study aimed at conceptually replicating the expected correlational results, strengthening their reliability and robustness through triangulation from an experimental approach. We implemented a two-level single-factorial between-subject design to prime descriptive norms (supportive vs punitive) toward detained people. We hypothesized that activating a prisoner-supportive (vs punitive) norm would increase officers’ supportive behavioral intention toward detained people and, in turn, officers’ expected well-being and job satisfaction resulting from their intended behavior.

A convenience sample of 1,080 Italian prison police officers accessed the study (in the Italian system, correctional officers are a Police Corp). The 87.2% of the responders were men of 51.07 years old (SD = 6.05). The majority had high school education (63.8%), followed by middle school (19.7%), university (15.7%) and elementary school (0.8%). Most participants had a police rank of Constable (65.4%), followed by Sergeant (32.0%) and Inspector (2.6%). The average seniority was 29.0 years (SD = 6.5). None of the participants was excluded. In total, 861 participants fully completed the study, whereas the remaining presented missing values.

Based on the most recent available data about the Italian prison police personnel (Parisi, 2022), the sample was 2.99% of the total prison police personnel population (n = 36,071). The sample’s gender distribution was not significantly different from the general population (population’s gender: 87.6% men, 12.4% women; Χ²(1) = 0.08, p = 0.780), whereas we oversampled sergeants compared to the other ranks (population’s ranks: Constable = 82.7%, Sergeant = 8.7%, Inspector = 5.8%, Commissioner and Executive = 2.8%; Χ²(3) = 601.4, p < 0.001). We believe the sample can be considered representative of the population as the sergeant distribution was biased because most participants were constables who had just been promoted to sergeants.

We invited to participate prison officers who were attending residential training programs to become Sergeants and Inspectors in various sites throughout Italy. We organized video calls to connect online in a structured meeting with all police officers attending the training programs. We presented the research to the classrooms of trainees and we projected a QR code redirecting to the online questionnaire and procedure embedded on Qualtrics. At first, participants completed the correlational part of the research. Then, they began the experimental part – embedded in the same Qualtrics survey –, where they were randomly assigned to one of the two experimental conditions by the Qualtrics platform. Participants concluded the study by answering the experimental measures and the sociodemographic questions. The researchers were available to address questions throughout the procedure, which lasted approximately 50 min.

The research was approved by the Ethics Committee of the University of Milano-Bicocca (ethics reference n. 756, May 25^th, 2023). Also, we obtained authorization from the Italian Ministry of Justice, which reviewed and approved the study’s aims, procedures and materials. To ensure officers did not feel pressured, we emphasized that participation was voluntary, anonymous and unrelated to the training courses; those opting out could leave the room for a break. Officers provided online informed consent before participating, and we also provided the first author’s contact information for any follow-up questions or concerns (none were received).

The correlational questionnaire included a list of measures from the broader research project investigating various risk and protective factors for prison officers’ well-being. From those, we considered all the available measures that matched the experimental manipulation and outcomes and, as such, were useful to test the present study’s hypotheses (see the complete measure package at the OSF link). Participants answered on a five-point Likert scale ranging from 1 (“not at all”) to 5 (“extremely”) and items were averaged into aggregate scores if not otherwise stated.

Injunctive and descriptive supportive norms toward detained people were measured with three items, each adjusted from the Cultural Empathy Subscale (Van der Zee et al., 2013). A principal component analysis showed that the six items loaded onto a single component (R² = 0.56, λs_range = 0.67–0.79; Cronbach’s α = 0.84).

Emotional closeness with detained people was operationalized using the three measures reported below. Empathy for Detained People was measured with an adjusted version of the full eight-item Cultural Empathy Subscale (Van der Zee et al., 2013). A principal component analysis showed that the eight items loaded onto a single component (R² = 0.51, λs_range = 0.62–0.76; Cronbach’s α = 0.86). Interpersonal Closeness with Detained People was measured by assessing the quantity and intimacy of interpersonal interactions with detained people using a listing procedure adjusted from Marinucci et al. (2022). Participants were asked to list the initials of up to ten detained people they had regularly interacted with in their work shifts and to rate how close they felt to each entry. The aggregate index was computed by summing the closeness scores: higher scores indicate more frequent and close relationships with detained people. Those who confirmed they had not had regular interactions with any received a final score of 0 (range = 0–35). Intergroup Closeness with Detained People was assessed with the single-item Inclusion of Others in the Self (Aron et al., 1992). Participants were asked to rate how emotionally close they felt to detained people. The two circles represented the participant and the group of detained people; the response scale ranged from 1 (least close) to 7 (most close).

Attitudes were operationalized by two measures. Hostility toward Detained People was assessed with three items from the Attitudes Toward Prison scale (Melvin et al., 1985; Mosser, 2022) measuring negative and hostile attitudes. A principal component analysis showed that the three items loaded onto a single component (R² = 0.53, λs_range = 0.68–0.77; Cronbach’s α = 0.55). Punitive–Educative Orientation was measured with an 11-point single bipolar continuum. Participants were asked in what percentages the punitive and educative mandates of the prison should rule the Italian prison system. The left end of the continuum coded the option “100% punitive – 0% educative” and the right end coded “0% punitive – 100% educative”. Each scale point measured the change of 10% in the professional orientations (e.g. “90% punitive – 10% educative” and so on). Higher scores indicated a higher educative orientation (range = 1–11).

Professional and personal well-being was measured by considering burnout, job satisfaction and overall psychological distress. Burnout was assessed with the 4-item short Burnout scale measuring exhaustion and disengagement components (Mitropoulou and Zampetakis, 2022). A principal component analysis showed that the four items loaded onto a single component (R² = 0.67, λs_range = 0.79–0.85; Cronbach’s α = 0.84). Job Satisfaction was measured with three items assessing liking their job and perceived job satisfaction (Lawler et al., 1979) and three items measuring intentions to quit their job (Wayne et al., 1997). After removing two poor-fitting items, the principal component analysis suggested that the four items (2 from each scale) loaded onto a single component (R² = 0.63, λs_range = 0.74–0.85; Cronbach’s α = 0.80). Psychological Distress was assessed by measuring the Resignation Stage that is a psychological condition characterized by depression, alienation, helplessness and unworthiness derived by the threat posed by chronic social exclusion to the needs for belonging, self-esteem, control and recognition (Williams, 2009). We adopted the same 12-item scale from Marinucci and Riva (2021) to measure depression, alienation, unworthiness and helplessness. A principal component analysis showed that the 12 items loaded onto a single component (R² = 0.48, λs_range = 0.31–0.84; Cronbach’s α = 0.87). See Table 1 for descriptive statistics and correlations between the measures.

After the correlational measures, participants were randomly shown a mock online newspaper article describing a fictitious Slovenian prison (the Ledano prison) either adopting a friendly, open-cell, prisoner-supportive and educative-oriented regime (the supportive norm prime) or a harsh, closed-cell, prisoner-restrictive and custody-oriented regime (the punitive norm prime). The supportive norm condition was inspired by the online article from Benezic (2016) describing the Halden prison in Norway. The punitive condition matched the supportive norm one in terms of text length and the prison’s features described and was based on a typical custody/punitive-oriented prison. The author I.T., (a prison police commissioner), advised the generation of the stimuli, confirming their face validity.

After being randomly presented with one of the two priming conditions, participants were instructed to think about their typical work shift and to imagine themselves in the following scenario: “About the end of your work shift in the detention section, an inmate gets closer to talk to you. He tells you that he is struggling with some of the other detained people and that he is going through a tough period of suffering. He asks if you could listen to him in a quiet place, away from prying ears, maybe having a walk together in the courtyard. You immediately notice that he is suffering for real and he is really in need.” Then, they were asked the following.

Behavioral intentions in response to the scenario were measured with a single item: “In the described situation, how likely would you accept the request from the inmate?”.

Participants were instructed to answer thinking about how they would feel after their intended behavior in the presented scenario. Expected well-being was assessed by measuring anticipated satisfaction of the basic needs for belonging, self-esteem, control and recognition with one item for each need selected from the Need-Threat scale from the social exclusion literature (Williams, 2009). We opted for such operationalization of well-being to be consistent with the correlational measure of psychological distress. A principal component analysis showed that the four items loaded onto a single component (R² = 0.65, λs_range = 0.70–0.87; Cronbach’s α = 0.81).

Expected job satisfaction was assessed with a single item: “At the end of the work shift, I would be satisfied with my job.” As for expected well-being, participants were instructed to answer in relation to the scenario and their intended behavior.

Finally, participants answered three manipulation-check items measuring supportive descriptive norms toward detained people referred to the prison context described in the fictitious prison described in the newspaper article (e.g. “In the Ledano prison, officers sympathize with prisoners”). A principal component analysis showed that the three items loaded onto a single component (R² = 0.87, λs_range = 0.91–0.95; Cronbach’s α = 0.93).

We report how we determined our sample size, all manipulations and all measures (see the supplementary materials). We reported all the experimental conditions and measures. The study was not preregistered. Study materials, data, the analysis code for results’ reproducibility and supplementary materials are publicly available (https://osf.io/e5f6c/?view_only = 6b45ed3455334a719726ef9f24eef20b).

Correlational and experimental hypotheses were tested by conducting Structural Equation Models using the lavaan package (Rosseel, 2012) on RStudio (Posit team, 2024). We applied full-information maximum likelihood estimation to manage random missing data (Schafer and Graham, 2002). Model fit was assessed considering root mean square error of approximation (RMSEA) (acceptable < 0.08), Tucker Lewis index (TLI) and comparative fit index (CFI) (acceptable > 0.90) and standardized root mean squared residual (SRMR) (acceptable < 0.07) (Marsh et al., 2005; Satorra and Bentler, 2001).

A confirmatory factor analysis (CFA) confirmed that empathy, interpersonal and intergroup closeness loaded onto the emotional closeness factor (λs = 0.40–76), whereas punitive-educative orientation (λ = 0.55) and hostility (λ = −0.53) loaded onto the supportive attitude factor (Model fit, optimal: Χ²(4) = 10.95, p = 0.027; CFI = 0.98, TLI = 0.95; RMSEA = 0.043, SRMR = 0.019). Then, we estimated the main parallel mediation model (Figure 1).

In the measurement part, we specified emotional closeness and supportive attitude as latent factors measured by the respective indicators (as in the above-described CFA). In the structural part, we specified the latent factors as parallel mediators of the relations between the predictor (supportive norms) and the three outcomes of burnout, job satisfaction and psychological distress. A sensitivity power analysis showed that the sample size allowed the detection of even a small size (β = 0.10) of all the direct effects with at least 79% of power (see the Supplementary Materials and Figure 1S on the OSF platform).

The model had good fit (Χ²(16) = 89.02, p < 0.001; CFI = 0.95, TLI = 0.90; RMSEA = 0.071, SRMR = 0.038). The results showed that supportive norms were associated with a higher supportive attitude toward detained people, measured by a higher educative orientation and reduced hostility. In turn, supportive attitudes related to lower psychological distress, burnout and increased job satisfaction.

The results also showed that supportive norms related to higher emotional closeness, measured by empathy and interpersonal and intergroup closeness with detained people. Differently from supportive attitudes, the effects of emotional closeness on the well-being indicators were more complex. While emotional closeness was related to higher burnout, it was also related to higher job satisfaction, whereas it was not significantly associated with psychological distress. The nonsignificant effect of emotional closeness on psychological distress might be due to a zero-sum of the opposing effects for burnout and job satisfaction, resulting in a null effect on overall psychological well-being.

Overall, the test of the indirect effects (Table 2) showed that supportive norms might have nuanced effects on the different dimensions of well-being via the parallel mechanisms of supportive attitudes and emotional closeness. Supportive norms were positively associated with well-being (i.e. reduced burnout and psychological distress) by increasing officers’ supportive attitudes. Also, supportive norms indirectly related to higher job satisfaction by increasing supportive attitudes and emotional closeness. However, the findings also showed that supportive norms might backlash on well-being via emotional closeness, as it related to higher burnout.

A two-sample t-test on the manipulation check scores confirmed that participants in the supportive (vs punitive) prison condition perceived the prison officers from the fictitious prison as more friendly, empathic and supportive with detained people (t(866) = 20.95, p < 0.001, Cohen’s d = 1.42; Supportive condition: M = 3.00, SD = 1.03, Punitive condition: M = 1.70, SD = 0.79). Preliminarily, we tested the direct effect of the conditions on officers’ behavioral intention, anticipated well-being and job satisfaction while also exploring the overall correlations between the constructs (Table 3).

The supportive prime significantly increased supportive behavioral intentions toward detained people. Differently, the priming procedure did not directly influence expected well-being and job satisfaction following the intended behavior. As the expected well-being and job satisfaction referred to participants’ behavioral intention in the hypothetical scenario, it could be that the outcomes were only conditioned by the intended behavior (as confirmed by the experimental mediation), not by the priming conditions.

Then, we conducted a mediation model specifying the condition as the predictor, the behavioral intention as the mediator and expected well-being and job satisfaction as separate outcomes. We included the latent factors of emotional closeness and supportive attitude from the main correlational analysis as control variables of the effect of the prime condition on both the mediator and the outcomes (Figure 2). The inclusion of these control variables allowed us to assess the effect of the priming condition net of baseline individual differences on emotional closeness and supportive attitude toward detained people, also ruling out the possibility that such individual differences could confound the priming effect. Note that the results did not change without the control variables.

The model showed optimal fit (Χ²(18) = 87.03, p < 0.001; CFI = 0.95, TLI = 0.91; RMSEA = 0.066, SRMR = 0.035). The results confirmed that the supportive (vs punitive) norms prime increased participants’ intentions to act supportively toward detained people in a hypothetical scenario realistically set in participants’ working context. In turn, higher supportive behavioral intentions significantly increased participants’ expectations to be satisfied with their job and well-being at the end of their hypothetical work shift. Besides, the results showed that baseline levels of emotional closeness with detained people led participants to anticipate higher supportive behavioral intentions toward detained people and to expect to be more satisfied with their job and well-being. Differently, baseline supportive attitude did not influence anticipated behaviors, well-being or job satisfaction.

The indirect effects (Table 4) indicated that the supportive norms prime led participants to expect significantly better well-being and job satisfaction at the end of the hypothetical work shift, mediated by increased intentions to behave supportively toward detained people. In other words, the results suggested that exposing prison officers to prisoner-supportive norms might prompt supportive behaviors toward detained people, with ultimate benefits for officers’ psychological and occupational well-being.

This correlational-experimental research investigated whether fostering prosocial norms among prison officers could promote two benefits: improve relationships with and attitudes toward detained people, which could deter officers from abusing and mistreating them, and enhance officers’ professional and psychological well-being. The correlational part of the study found that descriptive and injunctive norms supporting detained people were associated with increased job satisfaction and reduced burnout and psychological distress indirectly via a higher supportive attitude toward detained people, characterized by an educative orientation and reduced hostility. The results are consistent with the literature suggesting that officers’ supportive attitudes might generate a less tense and anxious climate in the relationships with detained people, positively impacting their well-being (Maculan, 2022). Also, the positive effects can be understood as officers whose working contexts are imbued with supportive norms might feel fulfilled in their twofold mission of ensuring security while contributing to prisoner reeducation (D’angelo et al., 2018; Tewksbury and Higgins, 2006). Besides, the findings speak to the broader literature about prosociality, showing that holding prosocial attitudes might contribute to well-being by promoting a sense of meaning and fulfillment in life and satisfying the basic needs for competence, relatedness and meaning (Aknin et al., 2013; Hui et al., 2020; Midlarsky, 1991).

The results also showed that supportive norms related to higher emotional closeness, measured by empathy and interpersonal and intergroup closeness with detained people. However, the mediating effects of emotional closeness were mixed as supportive norms indirectly increased both burnout and job satisfaction. Such ambivalent effects align with the theoretical propositions from Zenasni et al. (2012), suggesting that empathy might carry emotional costs resulting in emotional exhaustion while also providing a sense of meaning and satisfaction in one’s job. More related to the prison context, it could be that higher emotional closeness, while fostering a prosocial sense of meaning and fulfillment, would induce a role conflict on an operational level as officers emotionally close to detained people might experience distress when called to execute custody-related orders (Molleman and Leeuw, 2012). Such conflict would not occur in officers with a highly supportive attitude but less emotional involvement, for whom custody-related tasks could be conceived as part of a higher re-educative purpose and, as such, less distressing.

The interpretations of the correlational results should be taken with care due to their methodological limitations. Mediational models on cross-sectional data cannot claim causations of the tested effects nor rule out alternative explanations based on an opposite direction of the effects, from the outcome to the main predictor and mediators. For example, concerning our finding that officers’ supportive attitudes were associated with lower burnout, it could also be that higher workplace burnout could lead to more negative attitudes toward detained people. Also, job satisfaction and psychological well-being could influence officers’ professional orientation, as shown by a previous correlational study (Ferdik, 2018). Hence, the current correlational results should not be intended to claim the directionality or causality of the effects but rather to suggest possible interpretations of the associations between the considered variables based on the study hypotheses and their theoretical rationale.

The experimental part of the research aimed at strengthening the robustness and reliability of the correlational results through conceptual replication and triangulation with an experimental methodology. Results showed that priming prisoner-supportive norms led participants to anticipate themselves as more behaviorally supportive toward detained people and to expect higher needs and job satisfaction following their intended behavior. As such, the findings confirmed that fostering prisoner-supportive social norms could enhance officers’ well-being by improving their relationships with detained people. However, it must be noted that the experimental part of the study only induced a temporary activation of supportive norms and considered anticipated behaviors and well-being as outcomes. The findings cannot tell if the shift in behavioral intentions would translate into an observable change in officers’ behaviors and well-being, and, if so, for how long. Also, we cannot exclude that the experimental results were biased by a demand effect that would nudge officers to answer the experimental part complying with the prison’s organizational climate described in the journal article they read, rather than based on how prosocial they actually intend to behave toward detained people and expect to feel accordingly.

Despite the limitations described above, we believe that adopting a mixed correlational-experimental method on such a large and representative sample of Italian prison officers enhanced the robustness and reliability of the results. The findings from the correlational part, in terms of associations between variables, were confirmed and strengthened by the experimental part, specifically designed to assess the directions of the links connecting the constructs of interest. Indeed, the experimental part placed social norms as antecedents of intended prosocial behavioral intentions, that, in turn, were framed as antecedents of anticipated well-being and job satisfaction. Although the limitations described above impede the conclusion of any causal relations, at least the experimental findings confirm that the direction of the correlation effects is reasonable, sound and robust.

The findings advance the theory in several significant ways. First, the research informs the literature on prosocial norms, emphasizing how these can influence well-being. Second, it explained the mechanisms conveying such effects. It clarified the mixed literature about the link between empathy and well-being by emphasizing the nuanced dynamics and backlash effects on burnout, job satisfaction and overall psychological well-being. Third, the study advanced the criminology literature on the prison context by showing how prison norms can reduce officers’ hostility, improve their professional orientations and relations toward detained people and, ultimately, their well-being. These multiple theoretical contributions are enriched by the novelty, the methodological triangulation of the results and the large representative sample from a hard-to-reach population.

Future research should replicate and extend the present findings by implementing a more rigorous methodology able to overcome the current limitations. For instance, research could compare professional orientations, relationships with and attitudes toward detained people and psychological and professional well-being between officers coming from institutions with norms varying on how supportive vs punitive they are toward incarcerated people. Besides, longitudinal studies could investigate if more punitive norms and orientation would lead to lowered well-being over time. In Addition, future research should consider possible gender differences in social norms, empathy and relationships with detained people, as well as psychological and professional well-being. Indeed, the current predominantly male sample (reflecting the Italian prison officers’ gender composition) prevented us from identifying such differences or generalizing our findings to women officers. Finally, future research should consider developing and testing applied interventions to improve prosocial institutional norms toward detained people and assess whether the intervention would improve officers’ orientation and attitudes toward prisoners and, ultimately, their well-being.

In conclusion, this research provides timely, evidence-based guidance for addressing the issues of prison officers’ mental health with important repercussions for the issue of prisoner mistreatment. Scholars have warned that the issue of officers’ mistreatment and hostility toward incarcerated people is of growing concern as such abuses are common practices worldwide (Albano et al., 2023). A large-scale US study pointed at officers as the most frequent perpetrators of violence against detained people (Wolff et al., 2007). In recent years, hundreds of Italian prison officers have been accused of violent behavior or even torture toward incarcerated persons (ANTIGONE, 2024). More than 50 years after the Stanford prison experiment, many prison systems still reflect its findings, which suggest that violence and illness in prisons are rooted in their culture, management and policies (Haney et al., 1973; Specter, 2006). In this challenging context, the current research indicates that a prosocial shift in the social norms of prison officer culture could be a significant step toward resolving the critical issues of police violence and mental illness in prisons worldwide.

The authors did not receive fundings to conduct the research. Authors thank Dott. Pietro Buffa and Dott.ssa Paola Gubbiotti from the Department of Penitentiary Administration of the Italian Ministry of Justice for their advice and support to the research.

The supplementary materials for this article can be found online.