When police officers break rules: Do personality traits matter in unethical decision-making? Open Access

Police work inherently involves a high degree of discretion in decision-making, allowing police officers room for professional judgement in complex and ambiguous situations (Gilleir, 2013; Kleinig, 1996; Lipsky, 2010; Verhage, 2022). While many of these decisions concern law enforcement, others relate to broader ethical norms and professional integrity. Police officers are frequently confronted with situations in which they can behave either ethically or unethically, and these choices have implications for both organisational functioning and public trust. Unlike more widely studied cases of overt misconduct, such as excessive use of force or criminal behaviour, subtler and sometimes invisible forms of unethical behaviour have received less attention. Examples include accessing police databases without legitimate grounds (e.g. Hutchings, 2025). Although such acts may appear less serious, they can undermine organisational legitimacy and erode public confidence. Understanding police officers' decision-making processes in such situations is therefore important for both theory and practice.

Research in psychology and organisational studies highlights that ethical decision-making is shaped by both individual and situational factors (e.g. Bazerman and Tenbrunsel, 2011; Seppala, 2021). Among these, personality traits have received growing attention as potential determinants of ethical or unethical conduct. The widely accepted Five-Factor Model – or Big Five – is one of the most robust and widely researched in personality psychology (Schmitt et al., 2007). This model consists of five broad personality dimensions: Openness to experience (curiosity and flexibility), Conscientiousness (self-discipline and reliability), Extraversion (sociability and energy), Agreeableness (compassion and cooperation), and Neuroticism (emotional instability). Some of these dimensions, such as Conscientiousness, Agreeableness, and Extraversion, may promote professional and ethical decision-making (Black, 2000; Grubb et al., 2015).

Although the Big Five captures broad and relatively stable personality dimensions and provides a useful lens for examining how personality shapes decision-making, it does not encompass all traits relevant to ethical decision-making in policing. Besides Neuroticism, high impulsivity and low empathy have for instance also been associated with problematic behaviour (Black, 2000; Gudjonsson and Adlam, 1983; Twersky-Glasner, 2005). Research highlights empathic concern as a determinant of prosocial and ethical behaviour (e.g. Terrill and Paoline, 2007), whereas impulsivity is often associated with reckless or unethical actions (e.g. Cojean et al., 2024). Although some research has examined the impact of personality characteristics on police decisions (see Feys, 2023b for an overview), the specific link between police officers' personality traits and unethical decision-making remains underexplored. To address this gap, the present study investigates the relationship between broad Big Five traits, empathic concern, and impulsivity in predicting unethical intentions among police officers. Empathic concern and impulsivity add important nuance beyond the Big Five and are particularly valuable in understanding individual differences in the ethical decision-making of police officers. By examining the unique and combined effects of the Big Five traits, empathic concern, and impulsivity, this study contributes to understanding the role of personality in (un)ethical decision-making in police work. This study contributes to the literature in three specific ways. First, it examines the extent to which broad personality traits are associated with unethical intentions among police officers. Second, it investigates whether empathic concern and impulsivity explain additional variance in unethical intentions beyond the Big Five dimensions. Third, it explores whether these predictive relationships differ across various types of unethical decision-making among police officers. The next section reviews literature on personality and (un)ethical decision-making in police work, followed by the study design and analyses.

The Big Five represents the five personality traits that form the basis of many studies on personality today: Openness to experience, Conscientiousness, Extraversion, Agreeableness and Neuroticism. Within personality psychology, the Big Five model is recognised as one of the most robust and widely researched theoretical frameworks (Schmitt et al., 2007). Several instruments have been developed to assess the Big Five, varying in length, detail, and psychometric properties. Table 1 presents a concise overview of the most commonly used instruments, summarising their abbreviations, full names, numbers of items, and brief descriptions.

The five factor structure has been extensively studied in various populations (Mammadov, 2022; Meyer et al., 2023), among which police officers (e.g. Abrahamsen, 2006; Abrahamsen and Strype, 2010; Garbarino et al., 2012; Grubb et al., 2015; Petasis, 2020; Salters-Pedneault et al., 2010; TenEyck, 2024). Some of these studies found differences in police officers' personalities compared to non-police officers (e.g. Garbarino et al., 2012), whereas other studies did not find differences between police officers and other populations (e.g. Abrahamsen, 2006; Abrahamsen and Strype, 2010). Various studies explored the effects of police officers' personality characteristics. These characteristics were found to be related to job satisfaction (Ortega et al., 2006; Petasis, 2020; Spagnoli and Caetano, 2012); perceived workload (Chiorri et al., 2015) and perceived job stress (Garbarino et al., 2013a; Lau et al., 2006; Ortega et al., 2006); psychological well-being (Hart et al., 1995; Ortega et al., 2006) and mental health problems such as anxiety, depression, burnout and perceived frustration (Chiorri et al., 2015; Garbarino et al., 2012, 2013b; Louw, 2014; Papazoglou et al., 2019); PTSD symptoms (Madamet et al., 2018); suicide ideation (Pienaar et al., 2007); coping strategies (Bishop et al., 2001; Hart et al., 1995; Lau et al., 2006; Ortega et al., 2006; Wearing and Hart, 1996); and preferences for conflict resolution tactics (Abrahamsen, 2006; Abrahamsen and Strype, 2010). Personality traits were furthermore associated with police officer performance (Barrick et al., 2001; Cortina et al., 1992; Detrick and Chibnall, 2002; Hargrave and Hiatt, 1989; Varela et al., 2004) and police interview competencies (De Fruyt et al., 2006; Funicelli and Laurence, 2017; Smets, 2009, 2011).

Personality characteristics also influence police officers' decision-making. However, there is only little research that specifically examines this. A scoping review on police decision-making (see Feys, 2023b), which systematically studied the factors affecting police decisions, found only twelve studies in which personality was studied in relation to police decision-making (i.e. Corsianos, 2003; Fahsing, 2019; Girodo, 2007; Huhta et al., 2021; Landman et al., 2016; Leempoels, 2013; Noppe and Verhage, 2017; Spanoudaki et al., 2019; Terrill and Paoline, 2007; Verhage et al., 2018; Wortley, 1990; Zhang, 2020). Some of these studies only briefly mention police officers' personality characteristics (e.g. Corsianos, 2003; Fahsing, 2019; Spanoudaki et al., 2019; Terrill and Paoline, 2007; Zhang, 2020) and their potential impact on police decisions (e.g. empathy for the suspect, which may result in the decision not to arrest; Terrill and Paoline, 2007).

Three Belgian studies more specifically explored the potential impact of personality on police decision-making. Leempoels (2013) found that police officers who consider themselves sociable and extravert are more inclined to disclose information about themselves to interrogated persons (e.g. to connect with that person and receive information), whereas more reserved police officers will limit using self-disclosure as much as possible. Verhage et al. (2018) found that characteristics such as the ability to put things in perspective and self-confidence can impact the decision-making process during an intervention (e.g. as a confident police officer is less likely to make impulsive decisions). Based on the same research project, Noppe and Verhage (2017) argue that the turning point to use force is different for each police officer and is in large part the consequence of their personality.

A few other studies also explored the effects of personality traits on different aspects of police decision-making (e.g. Kim et al., 2020; Ponomarenko et al., 2022). These studies among others found that higher maximisation scores (i.e. the individual tendency to maximise outcomes by seeking the best possible choice, rather than settling for an acceptable choice; Shortland et al., 2020, p. 11) were positively associated with increased decision difficulty within a given scenario (Shortland et al., 2020), that police officers with a strong behavioural inhibition system took more time for their second shot when responding to pressure (Landman et al., 2016) and that time-urgent police officers, who perceive time to pass very quickly, had a larger reduction in the number of hypotheses they generated in the course of an investigation when placed under time pressure than non-time-urgent police officers (Alison et al., 2013).

Only a few studies regarding the impact of police officers' personalities on their decision-making included the Big Five personality traits in their research design (e.g. Girodo, 2007; Huhta et al., 2021; Tejeiro et al., 2024). Huhta et al. (2021) for instance demonstrated that police officers who scored lower in Extraversion were better able to adjust their behaviour in response to a changing situation. Police officers who shot the target before it was lawfully justified scored higher in Extraversion. Furthermore, more Extraverted persons were more likely to go after the target in a way that compromised their safety. Police officers with higher emotionality scores (i.e. more emotionally dependent) were more likely to avoid the encounter. Girodo (2007) found that personality dimensions, among which Extraversion and Neuroticism, did not predict who had shot someone on the job. Most comparable to the study reported in this article is the research of Tejeiro et al. (2024), which focused on students in the final week of their 5-year training to become police commanders in Spain. They used a personality questionnaire and two vignettes requiring decision-making in their survey. Contrary to expectations, they did not find that individuals with high Extraversion responded more quickly in uncertain or high-pressure scenarios or that highly Conscientious respondents exhibited deliberate and cautious decision-making in complex and ambiguous situations. They also did not find that individuals high in Neuroticism applied a more cautious and delayed approach in uncertain and risky situations. In line with their expectations, Openness and Agreeableness were not significantly related to the timing of action.

The literature study demonstrates that the direct impact of personality traits on police rule-breaking behaviour remains poorly understood. Moreover, the number of studies that included the Big Five personality traits, especially in combination with empathy and impulsivity, is very limited. This is why the present study focuses on the impact of the Big Five personality traits, empathic concern and impulsivity on police officers' rule-breaking intentions.

Given the limited research on personality's direct role in unethical decision-making among police officers, the present study adopts an exploratory approach to investigate three main questions:

1. To what extent do the Big Five personality traits, as measured with the BFI-44, predict unethical intentions among police officers?

2. Do empathic concern and impulsivity explain additional variance in unethical intentions beyond the Big Five traits?

3. Do the predictive effects of personality traits (Big Five, empathic concern, impulsivity) differ across various types of unethical intentions?

Although the BFI-44 assesses the five broad personality dimensions, it does not include specific measures of empathy and impulsivity. To address RQ2, we therefore measured empathic concern and impulsivity separately, allowing us to examine their incremental predictive value.

An empirically testable structural path model (see Figure 1) was developed in which the Big Five traits, empathic concern, and impulsivity served as exogenous predictors, and two forms of unethical intentions (i.e. unauthorised access to police databases and theft) were modelled as endogenous outcomes. Including both behaviours in the same model enabled the partitioning of shared variance and the estimation of the unique contribution of each predictor. This approach allows exploration of whether personality traits exert consistent influences across different ethical contexts (RQ3) and offers a more comprehensive understanding than analysing each trait–outcome link in isolation.

The present study extends previous enquiries by building on a large-scale PhD study on police decision-making in Belgium (see Feys, 2023a). The PhD study, which has been approved by the Ethical Committee of the Faculty of Law and Criminology of Ghent University, specifically focused on factors that (do not) affect police decisions, among which personality traits. For the present article, we use only part of the PhD study: the survey responses of 501 Belgian police officers. The online survey, programmed in LimeSurvey and accessible by means of a URL, consisted of four sections: Introduction and factors, Hypothetical scenarios and factors, Personality traits, and Background characteristics. The survey included 42 broader questions, often consisting of multiple sub-questions (e.g. statements). Each personality scale (see below) was counted as one question. The survey was sent to the chiefs of police of all Dutch and French speaking police forces in Belgium on 11 February 2019. An important disadvantage of this distribution method is that chiefs of police could also choose not to distribute the survey within their police force, which may have impacted the response rate. Three reminders have been sent (in March, April and May respectively) and the survey was closed on 31 May 2019. All respondents actively gave their consent after having received general information on the research project. Filling out the survey took approximately 30–45 min. No incentives were given for participation in the survey.

Two types of hypothetical scenarios were included in the survey: general scenarios that are applicable to all police officers, regardless of their position; and function-specific scenarios (e.g. focusing on emergency calls, investigative work, order maintenance). The latter type of scenarios are not used for the current analyses as they were only presented to the respondents in that respective function and thus filled out by a lower number of respondents. As such, only the general scenarios are used for this article, allowing us to explore the impact of personality characteristics on the intentions to steal (Heinz dilemma, based on previous literature, see Donenberg and Hoffman, 1988) and the intentions to access police databases without proper authorisation (scenario developed by the first author in close cooperation with Belgian police officers). The Heinz dilemma is a classic moral dilemma that is frequently used in psychological research to measure moral reasoning and moral intentions. This dilemma is useful for the present study as we do not only look at the impact of personality traits on police work, but on intentions more generally within a sub-population, namely police officers. As such, the Heinz dilemma allows to create a general baseline of moral intentions of police officers, regardless of their specific function. By combining both a police-specific scenario and a general scenario, we can explore whether police officers' personality traits consistently correlate with moral intentions in both general and professional contexts.

Due to partial nonresponse on the personality questionnaire (missing answers to one or more items), four respondents were excluded, resulting in a final analytic sample of 497 police officers. The majority of respondents were male (75.9%). Age distribution was fairly balanced across the 18–35, 36–45 and 46–55 age groups (each approximately 30%), with 11.9% aged 56–65. Nearly half of the sample (47.9%) reported more than 20 years of professional experience. Further methodological details and sample information are provided in Feys (2023a).

Police officers were asked how they would respond in two hypothetical scenarios. For each scenario, they could select one of three response options: “Yes, I would” (e.g. steal or look up the information), “No, I would not”, or “I do not know”. For analytical purposes, responses were dichotomised into “Yes” (indicating an intention to engage in the unethical behaviour) versus “Not yes” (a combined category including both No and Don't know responses).

This coding decision was driven by both methodological and conceptual considerations. First, excluding Don't know responses entirely would have led to a reduction in sample size of approximately 34% for the stealing scenario and 12.3% for the police-specific scenario, thereby limiting statistical power. Second, from a theoretical standpoint, Don't know was interpreted as an absence of a clear behavioural intention, and therefore more closely aligned with No than with Yes. This approach is conservative, treating uncertain responses (Don't know) as if the police officer does not intend to behave unethically, to preserve sample size and measurement integrity (e.g. Denman et al., 2018).

The Big Five Inventory (BFI; John et al., 1991) was used to assess personality traits, covering Extraversion, Agreeableness, Conscientiousness, Neuroticism, and Openness to experience. We applied the Dutch version of the BFI (Denissen et al., 2008), a 44-item self-report questionnaire rated on a five-point Likert scale (1 = strongly disagree, 5 = strongly agree). The BFI is widely recognised for its reliability and validity, demonstrating consistent internal consistency, test-retest reliability, and a clear factor structure (John and Srivastava, 1999). Higher scores indicate stronger expression of each trait. Descriptive statistics and reliability estimates for the current sample are provided in the supplementary materials. Confirmatory factor analysis (CFA) confirmed that the five-factor model adequately fits the data from Belgian police officers. Further details on the psychometric properties and CFA results can also be found in the supplementary materials.

The seven-item Empathic Concern (EC) subscale of the Interpersonal Reactivity Index (IRI; Davis, 1983, 1994) was used to measure empathy. This scale captures other-oriented feelings of sympathy and concern and is widely regarded as a valid measure of emotional empathy. We used the Dutch version (De Corte et al., 2007), with items rated on a five-point Likert scale (1 = strongly disagree, 5 = strongly agree). Higher total scores indicate greater empathic concern. Descriptive statistics and internal consistency estimates for the EC scale are provided in the supplementary materials.

The Dickman Impulsivity Inventory (DII; Dickman, 1990; Dutch version by Claes et al., 2000) was used to measure impulsivity. This 23-item self-report measure distinguishes two relatively independent dimensions: dysfunctional impulsivity (12 items), which is linked to negative behavioural outcomes, and functional impulsivity (11 items), reflecting more adaptive impulsive tendencies. Both dimensions describe tendencies to act quickly but differ in their consequences. Whereas dysfunctional impulsivity reflects hasty responding in situations where speed is likely to be detrimental (e.g. acting rashly and making unnecessary errors), functional impulsivity refers to rapid responding when such speed is advantageous (e.g. seizing opportunities or acting decisively under time pressure). This distinction allowed us to examine whether advantageous and detrimental forms of impulsivity relate differently to police officers' intentions in the scenarios. In this study, a modified five-point Likert scale was used (1 = completely disagree, 5 = completely agree), with higher scores indicating greater impulsivity within each domain. Psychometric properties and descriptive statistics for the DII are reported in the supplementary materials. The DII has demonstrated good validity in prior research and correlates with other impulsivity measures (Ernst-Linke et al., 2023; Lange et al., 2017).

Data analysis was conducted using Structural Equation Modelling (SEM) in Mplus version 7.11 (Muthén and Muthén, 2017). Given the binary nature of the two outcome variables—intention to steal and intention to access police databases without authorisation—the Weighted Least Squares Mean and Variance Adjusted (WLSMV) estimator was used. This estimator is suitable for categorical data and provides robust parameter estimates under conditions of non-normality (Kline, 2023).

In the WLSMV framework, each dichotomous outcome is treated as a thresholded manifestation of an underlying continuous latent response variable, estimated via a probit link function (Muthén, 1993; Muthén and Muthén, 2017). This allows for the joint estimation of multiple correlated binary outcomes within a single multivariate structural model, capturing shared unexplained variance between them.

To address the study's research questions, we estimated two nested multivariate SEM models ^[1]:

1. Model 1 (RQ1) tested the extent to which the Big Five personality traits, as measured by the BFI-44, predicted unethical intentions.

2. Model 2 (RQ2) expanded the model by adding empathic concern and impulsivity as additional predictors, to examine whether they explained incremental variance in unethical intentions beyond the Big Five.

In the present study, all predictors—the Big Five traits, empathic concern, and impulsivity—were treated as observed variables using summed scale scores. Consequently, no latent measurement component was included, and measurement error is not explicitly modelled. Analytically, this means the model can be considered a path model, though the SEM framework was retained for several reasons:

1. SEM allows simultaneous modelling of multiple endogenous variables (here, intention to steal and intention to access police databases without authorisation) and estimation of their residual correlation, capturing shared variance not explained by the predictors.

2. SEM provides a coherent framework for testing complex relationships even when using observed variables, in line with methodological guidance (Kline, 2023; Muthén and Muthén, 2017).

Initially, a latent measurement model using multiple-item indicators for all constructs was specified. However, it proved too complex given the sample size (N = 497) and led to poor model fit. Using summed scale scores, based on prior psychometric validation (De Buck et al., 2025), improved model stability while maintaining measurement reliability. This validation included exploratory and confirmatory factor analyses and demonstrated:

1. Acceptable fit for the five-factor structure of the Big Five, with factor loadings between 0.42 and 0.78. Factor loadings of 0.32 or higher are generally considered the minimum acceptable threshold (Costello and Osborne, 2005; Tabachnick and Fidell, 2001), no items exhibited cross-loadings ≥0.30, indicating that each item loaded primarily on its intended factor and that the measurement structure is robust.

2. Construct validity, including convergent validity (e.g. strong correlations between Agreeableness and Empathic Concern) and discriminant validity (e.g. low correlations with unrelated traits such as Openness to experience).

3. General psychometric stability across sex, with minor differences that did not compromise overall scale validity.

All predictors were entered simultaneously with direct effects on each outcome. Residual covariance between the two outcomes was freely estimated to account for shared method variance and within-individual consistency. No interaction or indirect effects were modelled. Conventional global fit indices (CFI, TLI and RMSEA) are not informative for models with observed predictors and binary outcomes analysed with WLSMV (Bentler, 1990; Byrne, 2012; Hu and Bentler, 1999). Therefore, the evaluation focused on parameter estimates, their significance and direction, providing valid and interpretable results within this analytical framework.

Overall, although the model is a path model, SEM remains methodologically sound for addressing the research questions, given the simultaneous estimation of multiple outcomes, their residual covariance, and the proper handling of binary outcomes via latent response variables.

Two nested structural equation models were estimated to examine the associations between personality traits and unethical intentions. Model 1, which included only the Big Five traits, indicated that higher Openness to experience was positively associated with intention to steal (β = 0.279, p < 0.001). In contrast, Agreeableness (β = −0.165, p = 0.006) and Conscientiousness (β = −0.211, p = 0.001) were significant negative predictors of stealing intentions. For the intention to access police databases without authorisation, only Agreeableness was a significant negative predictor (β = −0.174, p = 0.007).

Model 2 extended Model 1 by adding empathic concern and impulsivity variables. The effects of the Big Five traits on stealing intentions remained significant and similar in magnitude (Openness to experience β = 0.264, p < 0.001; Agreeableness β = −0.163, p = 0.028; Conscientiousness β = −0.197, p = 0.003). Agreeableness continued to significantly predict lower intentions to access databases without authorisation (β = −0.214, p = 0.004). Neither empathic concern nor impulsivity significantly predicted unethical intentions; functional impulsivity showed a negative relationship to the intention to access databases, which approached but did not reach statistical significance (β = −0.138, p = 0.060). Overall, Model 2 explained slightly more variance than Model 1, with R² increasing from 12.7% to 13.5% for stealing intentions, and from 4.3% to 5.8% for intentions regarding unauthorised database access ^[2]. Results of the extended Model 2 are shown in Figure 2. Model 1 results are included in the supplementary materials. Table 2 summarises the explained variance (R²) for unethical intentions across both models.

To test whether the predictive effects of personality traits differed between the two types of unethical intentions, two SEM models were compared. The first was an unconstrained model allowing regression paths from personality traits to differ across outcomes. The second was a constrained model in which the regression coefficients were forced to be equal across both outcomes.

Model fit for the constrained model was acceptable (χ²(8) = 13.81, p = 0.087; RMSEA = 0.038, 90% CI [0.000, 0.071]). The critical test was the DIFFTEST, a chi-square difference test that compared the constrained model to the unconstrained model. The DIFFTEST result was not statistically significant (Δχ ²(8) = 14.39, p = 0.072), indicating that constraining the regression paths to be equal did not significantly worsen model fit. This suggests that the same set of personality predictors influences both types of unethical intentions similarly.

Standardised regression estimates from the constrained model showed that Openness to experience positively predicted unethical intentions (β = 0.17, p = 0.001), whereas Agreeableness (β = −0.19, p = 0.001) and Conscientiousness (β = −0.16, p = 0.003) were negative predictors. Empathic concern and impulsivity did not significantly predict either outcome. The residual correlation between the two unethical intentions was moderate (r = 0.25, p = 0.002), indicating shared variance beyond that explained by the predictors, likely reflecting unmeasured individual or contextual factors. Together, these findings suggest that personality traits similarly influence police officers' intentions to engage in two distinct types of unethical behaviour.

This study aimed to examine to which extent the Big Five personality traits, along with empathic concern and impulsivity, predict police officers' intentions to engage in two distinct types of unethical behaviour: stealing and unauthorised access to police databases. Results showed that higher Openness to experience and lower Agreeableness and Conscientiousness significantly predicted intentions to steal, while for unauthorised database access, only Agreeableness was a significant negative predictor. Empathic concern and impulsivity did not explain significant additional variance beyond the Big Five traits for either type of unethical intentions. Furthermore, testing whether these predictive effects differed across the two behaviours revealed no significant differences, indicating that personality traits influence these distinct unethical intentions similarly.

Our findings support the predictive validity of the BFI-44 in policing contexts while suggesting that additional traits like empathic concern and impulsivity add limited incremental value, but the model demonstrated limited explanatory power, accounting for only a modest proportion of variance in unethical intentions. Among the Big Five traits, Agreeableness emerged as the most consistent characteristic associated with lower intentions to engage in unethical behaviour, across both scenarios. Within our sample, police officers scoring higher on Agreeableness were generally less inclined to consider rule-breaking. This personality domain captures tendencies such as forgiveness, gentleness, flexibility, and patience (Ashton et al., 2014) — all of which may act as dispositional restraints on unethical decision-making. In an organisational context such as the police force, these traits may serve as protective factors, discouraging ethical transgressions. Police officers who are more tolerant, cooperative, and less prone to anger or moral condemnation may be more inclined to uphold ethical norms, especially in challenging situations. These results align with prior research showing that Agreeableness is associated with prosocial and ethical behaviour in personal and organisational contexts. A recent comprehensive meta-analysis by Wilmot and Ones (2022) highlighted the central role of Agreeableness in predicting desirable social and workplace outcomes across a wide range of variables and samples. They synthesise eight general themes describing Agreeableness' functioning, including social norm orientation and relational investment, which emphasise its role in fostering cooperative, ethical, and norm-abiding behaviour. Although their review is not specific to policing, these insights underscore the potential importance of assessing Agreeableness in police recruitment and training, given its relevance to maintaining ethical standards in complex social interactions. This may even be more important in the context of police organisations as the impact of solidarity, leading by example by police leaders and the behaviour of colleagues are seen as strong elements in police (sub)culture (Quick, 2023). Adding the assessment of Agreeableness in recruitment and training may also have an effect on police culture – it may aid in the retention of police officers in the organisation and make the police organisation more attractive for specific groups of individuals with comparable traits.

The present study has some limitations. One limitation relates to the relatively small sample size, which may limit the generalisability of the results beyond our sample of Belgian police officers. Furthermore, the study used self-reported intentions in (a limited number of) hypothetical scenarios rather than actual behaviour, which may compromise ecological validity (Eifler and Petzold, 2019). Moreover, personality traits explain only part of why certain persons are inclined to behave unethically. Earlier research has demonstrated the variety of factors that affect police decision-making, in addition to police officers' personality (see Feys, 2023b), and has emphasised that the effects of personality traits are context-dependent and interact with situational factors (e.g. the moral ambiguity of the act, social norms, and perceived consequences; Deaux and Snyder, 2018; Larsen and Buss, 2021). Future research should therefore include other factors (e.g. working conditions, stress, organisational justice), for instance by means of factorial surveys (see Hembroff, 1987) in which contextual variables can be systematically manipulated (e.g. the context of the action or organisational pressures) to examine how these situational factors interact with personality characteristics in predicting ethical or unethical behaviour, providing a more nuanced understanding of the trait's role across different scenarios.

Future studies could also explore the motivation behind the (un)ethical behaviour. For instance, the motivation behind the act of stealing a medicine can be financial need, altruistic intent or personal gain, and such motivations can interact with personality traits and potentially influence the likelihood of unethical behaviour. Helping a loved one could, in some circumstances, be more likely among agreeable individuals due to altruistic tendencies. Similarly, in professional contexts such as access to police databases, the behavioural outcome could vary depending on who benefits from the action and the perceived consequences. In our data, the net effect of Agreeableness was protective, but it is important to recognise that this effect could differ in situations with different contextual details or targets. Our design does not manipulate or measure such motivational factors, thus not allowing us to empirically test how different motivations might alter the relationships observed between personality traits and unethical intentions. Experimental designs, such as scenario-based experiments that systematically vary the motivation behind ethically ambiguous behaviour, could help disentangle the conditions under which traits such as Agreeableness, Conscientiousness, or impulsivity exert stronger or weaker effects.

It would furthermore be interesting to investigate the role of additional personality dimensions such as the Dark Factor (D). D captures core antisocial tendencies—selfishness, manipulativeness, callousness, deceitfulness, and lack of empathy—that underlie a range of “dark” traits (e.g. Machiavellianism, psychopathy, narcissism). As a unifying construct of antisocial propensity, D predicts unethical behaviour above and beyond what the Big Five or empathic concern can explain (e.g. Moshagen et al., 2018, 2020). Therefore, including D could increase predictive power for unethical decisions. Besides including additional personality dimensions, it would also be beneficial to use longitudinal or panel data to assess the stability of personality effects and track changes in unethical intentions over time. Finally, comparing models across different police units or ranks could help identify differences between subsets of police officers.

This study has important added value in regard to the existing scholarship on the impact of personality traits on police decision-making and more specifically unethical intentions. This study focused on more subtle forms of unethical behaviour, which are rarely studied in a policing context. Such decisions can seriously hamper trust in police and perceived legitimacy of the police organisation. As real-life decisions are shaped by more than just personality, it is not recommended to focus the selection procedure too narrowly on recruits' personality. However, mapping police officers' personality might help to spot police officers who are more at risk for unethical decisions and help to develop appropriate training and support. Such training and support could focus on how to deal with temptations of the police profession. Insights into police officers' personality profiles and how personality traits affect (unethical) behaviours can also help to develop ethical policies and may assist in moulding police culture. Practically, our findings highlight the consistent and significant role of Agreeableness in reducing unethical intentions—specifically stealing and unauthorised database access—in police officers. We can reasonably infer that Agreeableness may also reduce the likelihood of unethical decisions in other contexts. However, policing often involves situations that challenge or suppress Agreeableness (e.g. aggressive confrontations). Police work can be demanding, and police officers are systematically confronted with both psychologically and physically impactful events. The repetitive exposure to strain can affect police officers' attitudes and behaviour. Therefore, police organisations should explore ways to foster and support Agreeableness among officers on a structural basis, throughout their career, through organisational culture, leadership, and training programs.

The PhD project for which these data were collected has been approved by the Ethical Committee of the Faculty of Law and Criminology of Ghent University.

The supplementary material for this article can be found online.