The “Civilizing” Effect of Body-Worn Cameras on Police-Civilian Interactions: Examining the Current Evidence,Potential Moderators,and Methodological Limitations

Self-Awareness Theory

Wicklund and Duval (1971) introduced Self-Awareness Theory, describing the psychological conditions in which the self becomes the focus of human consciousness. According to this theory, people ebb and flow through two states of self-awareness: subjective and objective self-awareness. Subjective self-awareness is the state where a person is unaware that they are a separate entity from others and blurs their sense of being with others (Wicklund & Duval, 1971). Like a child lacking theory of mind, they do not consider another person's viewpoint. As a result, they do not perceive themselves as a potential object of evaluation, which prevents them from assessing the consequences of their actions. On the other hand, objective self-awareness is the state where an individual perceives themselves as a distinct entity and can take on the viewpoint of others. As a result, the self becomes an object of evaluation. This allows an individual to reflect on how their behaviors align with their attitudes (Wicklund & Duval, 1971). If their behaviors do not align with their attitudes, they experience cognitive dissonance—a psychological discomfort felt when one's behavior is inconsistent with one's beliefs (Festinger, 1957). People who experience cognitive dissonance are motivated to relieve their discomfort by taking measures to adjust their behavior to fit their attitudes (Stone & Fernandez, 2008) or vice versa (Elliot & Devine, 1994). Hence, individuals engaged in objective self-awareness tend to act according to their attitudes. People typically develop prosocial self-schemas, consisting of positive attitudes about the self (e.g., I am a nice person). When these schemas are activated through self-awareness, people usually behave in prosocial ways (Froming et al., 1998).

The most obvious way in which self-awareness can be triggered is in the presence of onlookers (Carver & Scheier, 1978; Froming et al., 1982). An individual can become objectively self-aware when they believe an audience sees them as an object of evaluation. However, the physical presence of human observers is not a necessary condition of objective self-awareness. Inanimate objects, like mirrors, can trigger self-awareness and self-scrutiny as well (Heine et al., 2008). Cameras, in particular, have the potential to reach billions of viewers at one time on a live feed and can be reviewed again later. Hence, it is unsurprising that cameras can invoke objective self-awareness (Davis & Brock, 1975) and may promote prosocial behavior (Van Rompay et al., 2009; Van Bommel et al., 2014). Within the context of policing, public scrutiny of BWC footage is common. BWC tapes can be released to the community upon public record requests (Zansberg, 2021), and supervisors can view officer footage without restriction (e.g., New York Police Department, 2020). Thus, given the potential for BWC footage to be viewed later by a wide human audience, police and civilians should, theoretically, resort to prosocial behavior in the presence of a BWC.

Deterrence Theory

Deterrence theory originated from essays written by two philosophers, Beccaria and Bentham, in the 18^th century (Tomlinson, 2016). The theory was later developed to make parsimonious predictions about crime and punishment. That is, punishment would discourage criminal behavior insofar as its certainty, swiftness, and severity exceed the benefits of committing a crime (Piquero et al., 2011). Of particular interest is the element of certainty—the probability of being caught and punished. Research has shown that punishment is most likely to have a deterrent effect when the certainty of punishment is high (Feess et al., 2018). In addition, a meta-analysis found that the element of certainty had the most consistent inverse relationship with criminal behavior found in the literature (Pratt et al., 2006). Although, this association is attenuated when it is placed from bivariate to multivariate models containing other relevant control variables, such as gender and age (Pratt et al., 2006).

Further, BWC recordings can serve as direct evidence in court (Zamoff, 2019). When officers are granted limited discretion on activating BWCs and department policy dictates that nearly every civilian encounter is recorded, BWCs should increase the likelihood of either civilians or police being caught misbehaving on video (Ariel et al., 2017). Such evidence should, theoretically, increase the odds or certainty of punishment in criminal court or sanctions from supervisors (Ariel et al., 2017; Ariel et al., 2018). Given this, it is reasonable to believe that police officers and civilians aware of these implications should suppress their criminal or non-compliant tendencies in the presence of an active BWC.

Therefore, at face value, both theories suggest that BWCs should have a civilizing effect on police and civilian behavior. Self-awareness should alert officers and civilians of their prosocial values and, in turn, encourage prosocial behavior. Moreover, the increased certainty of punishment assisted by video evidence should discourage non-compliant behavior from civilians and police officers. Both the certainty of punishment and self-awareness provided by BWC use should play complementary roles in producing increased civility in police-civilian interactions. However, the current research suggests the effects of BWCs are not so straightforward. We argue one reason for the not-so-straightforward results may be that certain individual-level factors moderate the theorized self-awareness and deterrence mechanisms (i.e., the relationship between these mechanisms and police-civilian interactions might be dependent on one or more other unaccounted for variables). For example, the civilizing effects of self-awareness may depend on the type of attitudes people hold about the police. Those who hold antisocial attitudes towards the police may act in antisocial ways if triggered by self-awareness. And as mentioned previously, certainty-related deterrence weakens when other individual-level variables (e.g., gender, race) are controlled for. We later contend that accounting for these individual-level factors—especially the psychological ones— along with other methodological considerations are key to understanding the mixed-effects of BWCs described below.

Police Attitudes Toward BWCs

Although psychological theories have been used to hypothesize how police officers will behave while equipped with BWCs (e.g., Self-Awareness Theory), little research to date has specifically included officer thoughts and opinions about BWCs as a moderating variable outside of surveying officers about their attitudes and reporting these results. Methodologically, the majority of these studies have not moved past the basic functional analysis approach used by the behaviorists in the 20^th century (e.g., Hull, 1943). That is, they have followed a similar pattern of assigning BWCs to an experimental group and then compared this to a control group to determine a basic deterrent (or prosocial) effect on police and civilian behavior. Researchers (might) infer, or suggest in discussion, psychological processes on the basis of ‘functional analysis,’ but these methods are limited because they fail to directly assess psychological processes.

Such methodologies do not directly assess attitudes, which is problematic because decades of research suggest that, in general, attitudes can predict behavior—albeit there are some notable exceptions (see Ajzen & Fishbein, 2005; Fazio & Olson, 2014; Fazio, 1986). The importance of considering police attitudes is evident in a number of surveys of police officers. In an open-ended survey about BWC use, one police officer stated, “…Big Brother eye in the sky has just landed on our heads! There is no more honor in this world, only lawsuits…” (Ariel, 2016b, p. 752). This particular officer expressed their concern that BWCs would make their department more vulnerable to litigation and would degrade the dignity of law enforcement. In other studies, police officers have voiced their concerns about their supervisors going on “fishing expeditions” to find evidence of misconduct (Gaub et al., 2017; Gaub et al., 2020; Smykla et al., 2016; White, 2014, p. 28). Further, in one open-ended survey, Grossmith et al. (2015) found that law enforcement personnel expressed mixed opinions about BWCs. The opinions ranged from BWCs affecting job performance, “…you do take a second to think ‘well why am I potentially gonna stop [and frisk] this person’…” to BWCs not affecting job performance, “…you’re dealing with the situation that's in front of you … I don’t think it changes the decision [to stop and frisk]…” (p. 19).

Considering this, it would seem that examining the potential influence of these attitudes in BWC outcomes would be common in BWC research. However, the current BWC literature mostly consists of surveys solely examining officer perceptions of BWCs (e.g., Snyder et al. 2019; Gaub et al., 2016; Gramagila & Phillips, 2018; Huff et al., 2020; Jennings et al., 2014; Kyle & White, 2017; Wooditch et al., 2020) or experiments looking only at the effects of BWCs on behavior without including officers’ perceptions (e.g., Ariel et al., 2015; Jennings et al., 2017; Hedberg et al., 2017; Henstock & Ariel, 2017; Sutherland et al., 2017; White et al., 2017; Yokum et al., 2017). A few studies have integrated both perceptions and experiments (e.g., Headley et al., 2017; Jennings et al., 2015; Katz et al., 2014; Ready & Young, 2015); but of these few crossover studies, none could be located that have incorporated police attitudes in interaction models to see if they would moderate or impact the effect of BWC implementation on the use of force reports or civilian complaints. In addition, these studies have not included police officer attitudes as covariates to account for some of the variance in use-of-force reports, complaints, or some of the null findings mentioned previously.

For example, in one study conducted at Mesa, Arizona, the researchers randomly assigned police officers to wear or not wear BWCs over a period of 10-months (Ready & Young, 2015). In addition, they differentiated police officers who volunteered and were assigned to wear BWCs from police officers who did not volunteer but were mandatorily assigned to wear BWCs. For outcome variables, the officers were asked several interesting and relevant questions such as stop-and-frisk activity, perceived helpfulness of BWCs, and other measures. The results revealed that the officers wearing BWCs were less likely to stop-and-frisk civilians compared to their non-BWC counterparts. Importantly, officers who volunteered to wear BWCs were more likely to find BWCs to be helpful than non-volunteer officers and control officers (Ready & Young, 2015).

Although this difference was found, the perceptions of BWC helpfulness were neither controlled for nor included as a moderator in any of the analyses. Thus, it is unknown if perceptions of helpfulness in the volunteer officers impacted the differences in stop-and-frisk behavior between the experimental and control conditions. In fairness, the authors did control for volunteering in their analyses, which could serve as a proxy for police attitudes. Nonetheless, the analyses could have benefited from the officer's perceived helpfulness of BWCs incorporated as a moderating or control variable.

In sum, human behaviors are, at least partially, driven by attitudes. Therefore, considering officer attitudes is pivotal. Officers’ attitudes or expectations about BWCs may influence how much actual force they use in civilian encounters, which could ultimately inform their self-reported use of force. The majority of the studies in the BWC literature currently have examined only police attitudes or conducted only experiments, not typically combining the two. Furthermore, the few studies incorporating both aspects have not accounted for police attitudes in analyses assessing the effects of BWCs on officer behavior. Thus, it is possible that the effects of BWCs on civility may wax and wane or even neutralize depending on officer's attitudes towards BWCs.

Civilian Attitudes Toward the Police

While officer perspective is an important variable to study, police-civilian interactions are a two-way street. Civilians have attitudes about law enforcement that may also shape their encounters with police officers. In fact, one possible reason for the unexpected antagonistic effect of BWCs found in some studies may be civilian attitudes towards the police. Recall, Self-Awareness Theory posits that people tend to align their behavior to be consistent with their attitudes (e.g., Scheier et al., 1974). Researchers have used this theory to hypothesize that people will behave appropriately in police-civilian interactions because BWCs will make them objectively self-aware (e.g., Ariel, 2016a). This hypothesis is contingent on the assumption that people generally hold prosocial attitudes. However, it is evident that not all people hold prosocial attitudes, especially towards the police. In areas where the excessive use of force has deeply scarred communities, residents may have negative attitudes and feel anger towards the police (Kochel, 2019). Moreover, classic research on anger and aggression has shown that people who are made objectively self-aware of angry feelings or aggressive attitudes tend to act in antisocial rather than prosocial ways (Carver, 1974; Carver, 1975; Scheier, 1976).

Scheier (1976) demonstrated this antagonistic effect in an electric shock experiment. In the first part of this experiment, the participants were told to solve a difficult puzzle. While the participants were attempting to solve the puzzle, a confederate was instructed to either anger the participants (by criticizing their ability to complete the task) or not anger the participants (by remaining silent). Scheier also manipulated whether a mirror, a trigger of objective self-awareness, was present during the entirety of the experiment. After the puzzle task, the participants were assigned to be a “teacher” and the confederate was assigned to be a “learner.” The participant's objective was to teach the confederate about a topic by administering varying intensities of electric shocks every time the confederate answered a test question incorrectly. In the mirror condition, angered participants administered more intense electric shocks than non-angered participants. However, in the no-mirror condition, no such difference was found (Scheier, 1976).

The above experiment demonstrates that triggers for objective self-awareness can increase antisocial behaviors as much as they can promote prosocial behaviors. Whether antisocial or prosocial behaviors manifest may depend on an individual's attitudes. This means that BWCs may be the most effective in facilitating prosocial interactions in places where attitudes towards the police are positive. However, BWCs may be mostly ineffective, or even harmful, in areas where attitudes towards the police are negative. Taking this into consideration, negative attitudes may then explicate some of the findings on increased assaults against BWC officers (Ariel et al., 2016b). The importance of considering community attitudes is further complicated by geographical psychology.

Geographical Psychology

Replications of BWC field studies have been conducted across the US—from Florida (e.g., Jennings et al., 2015) to Arizona (e.g., Katz et al., 2015) to Colorado (e.g., Ariel, 2016a) to California (Ariel et al., 2015). One question naturally arises from this observation: Does location matter? An emerging line of research in geographical psychology would say it does. Over the years, research in geographical psychology has demonstrated the regional clustering of psychological characteristics, such as the Big-Five personality traits (Rentfrow, 2010; Rentfrow et al., 2015), prosocial/empathetic behavior (Bach et al., 2017; Levine et al., 2001), aggression (Cohen et al., 1996), and racially prejudiced attitudes (Freng et al., 2021; Chae et al. 2015). Moreover, the clustering of psychological characteristics has implications for political, economic, and health outcomes (Rentfrow & Jokela, 2016). For instance, areas with higher levels of structural racism at the state level corresponded to an increase in Black-White disparities of unarmed individuals being shot by the police (Mesic et al., 2018).

There is no direct evidence of how the clustering of psychological characteristics could influence civilian response to BWCs. Nonetheless, there is circumstantial evidence that a civilian's response to BWCs would vary based on their neighborhood of residence (Ariel, 2016a). In a field study in Denver, Colorado, the impact of BWCs on 911 calls from residents per 100 meters was examined across multiple police districts, which varied in crime counts. The findings revealed that in lower crime areas, residents exposed to BWCs (i.e., in the experimental district) reported crimes to 911 at a greater rate than residents not exposed to BWCs (i.e., in the control districts). However, for higher crime areas, the influence of BWCs on the residents was diminished. Upon a post-hoc inspection of stratas, the researcher discovered that areas with a lower crime count appeared to be middle to upper-class neighborhoods (Ariel, 2016a).

One potential explanation for this finding is the variation in perceptions of police legitimacy in areas with low SES compared to other areas. Research has shown that SES has a positive association with perceptions of police legitimacy, with high SES areas having the most positive perceptions of police (Schuck et al., 2008). Further, police legitimacy has been empirically found to impact civilian cooperation with the police. Specifically, citizens with more positive perceptions of police are more likely to report suspicious activity, call 911, and provide information about a suspect (Tankebe, 2013; White et al., 2016). Moreover, research suggests that BWC police officers are typically seen as more legitimate than non-BWC police officers (Demir et al., 2020). These two lines of research would loosely imply that lower SES areas would have lower perceptions of police legitimacy, and given that police legitimacy impacts civilian cooperation, police officers could experience less civilian cooperation in areas with lower SES. Self-awareness theory would add that BWCs would make people cognizant about their attitudes of the police's legitimacy, which would subsequently influence their thoughts and behaviors.

Overall, the evidence implicates that geography matters. People with similar psychological characteristics are inclined to live together. Thus, it is likely that individual characteristics germane to BWC research, such as police legitimacy, cluster in specific locales. Therefore, geographic clustering of certain psychological characteristics may moderate the impact of BWCs on “civilized” behavior.

Officer Perceptions of Self-Legitimacy

While it is important to understand civilian perceptions of police legitimacy when it comes to body-worn cameras, it is also important to understand their potential effects on officers’ views of their own self-legitimacy. Demir et al. (2020) found that civilians view police officers using body-worn cameras as more legitimate than police officers who do not use body-worn cameras. Research suggests that when civilians view police officers as more legitimate, officers may also be more likely to view themselves as legitimate (Tankebe & Meško, 2014). This increase in legitimacy can significantly impact an officer's approach to interactions with civilians (Trinkner et al., 2019). One of the most substantial impacts it can have is on an officer's perceptions of and willingness to engage in the use of force with civilians. Past research has shown that there is a negative association between self-legitimacy in police officers and the use of force (Bradford & Quinton, 2014; Tankebe & Meško, 2014; Trinkner et al., 2016). Bradford and Quinton (2014, as cited in Trinkner et al., 2019) stated that greater self-legitimacy may lead officers to approach difficult situations in more constructive ways and use force only as a last resort, while lower self-legitimacy may lead to quicker use of force due to a lack of belief in their own ability to maintain their authority in other ways. Therefore, it can be speculated that officers who use body-worn cameras would see an increase in their own self-legitimacy, through an increase in civilian views of police legitimacy, which then could lead to more peaceful interactions and less reliance on the use of force tactics from police officers.

The Role of Civilian Stress

By and large, the literature has shown that interactions with police are stressful (McLeod et al., 2020). We argue that this stress can potentially attenuate the psychological influence of BWCs on civilians. Psychological research shows that stress can impact selective attention (Chajut & Algom, 2003; Braunstein-Bercovitz, 2003). Selective attention refers to the ability of an individual to focus on task-relevant stimuli and ignore task-irrelevant stimuli. Generally, when stress is high, selective attention to task-relevant stimuli improves (Chajut & Algom, 2003; Hoskin et al., 2014). Selective attention could be relevant in various police-civilian interactions. For example, in traffic stops, officers often ask motorists to produce their license and registration. A stressed motorist should then be concentrated on finding these items and ignore other objects that are not relevant. Thus, a stressed civilian may not notice the officer is wearing a BWC because they are focused on the items specified in the officer's instructions. Moreover, even if the BWC is noticed, a civilian may temporarily forget its presence or lack the ability to process this information. Studies have demonstrated that stress may deteriorate working memory, which is the capacity to temporarily store and manipulate information (e.g., Sorg & Whitney, 1992; Klein & Boals, 2001).

For the most part, previous BWC studies have attempted to ensure that civilians initially recognize the presence of BWCs in their field experiments. Officers wearing BWCs in the experimental group are often instructed to verbally announce the use of BWCs at the outset of their contact with civilians (e.g., Ariel, 2016a). Furthermore, the appearance of BWCs has been modified to make them highly salient, such as applying bright yellow stickers (e.g., Drover & Ariel, 2015; Henstock & Ariel, 2017). Nevertheless, stress could undo initial efforts to make civilians objectively aware of BWCs. Without consistent awareness of BWCs potentially due to stress-induced selective attention and a possible decline in working memory as a result of stress, civilians may not be able to think about how being recorded by a BWC may impact the certainty of punishment or their current behavior. Also, they may not maintain an objectively self-aware state. Therefore, the assessment of the certainty of punishment and self-awareness hinges on the presumption that civilians are readily attending to and remembering a BWC is present. Since most interactions with the police are stressful, people may forget BWCs—even with a verbal warning and bright yellow stickers. For the most part, scholars have not accounted for civilian stress and awareness in their field experiments, which may explain some of the non-significant findings found in the literature (e.g., Ariel et al., 2016a; Headley et al., 2017; Yokum et al., 2017).

To account for the attenuation of the civilizing effect, future experiments should measure both civilian awareness of BWCs and stress. While it is easier to measure awareness alone, it is important to examine stress as well because it allows us to comprehend why awareness of BWCs may be disrupted. Knowing why will be important for interventions down the line: If it is found that stress impairs awareness, then real-world interventions could focus on ways to make police-civilian encounters less stressful.

Taken together, police-civilian interactions can be influenced by various psychological factors. Such factors can include police attitudes toward BWCs, civilian attitudes towards the police, geographic psychology, officer self-legitimacy, and stress. Whilst these factors may need explicit testing in the future, the main contention is that the impact of BWCs cannot be studied independently of psychological factors. Civilians and police are people. People have their own thoughts, feelings, and opinions, which color how they converse with others. These psychological differences—and the various possible combinations of these differences across diverse police-civilian encounters—may set the conditions in which BWCs are effective. For instance, it is possible that BWC may elicit high levels of civility in interactions where the officer buys into the utilization of BWCs and the civilian generally respects the police; whereas, BWCs may produce lower levels of civility in situations where officers dislike BWCs and civilians disrespect the police. Therefore, it is essential to start accounting for these moderators, or there may continue to be inconsistent results with little explanation. To address the issue at hand, forthcoming BWC studies could not only measure, but also integrate these characteristics of officers and civilians as moderators in statistical analyses. Unfortunately, though, the problems with the current BWC literature do not end with the absence of individual characteristics. Inconsistent findings in the literature could also be attributed to the methodological difficulties associated with conducting field experiments in police departments. In the next section, we discuss three of these issues in depth.

Low Base Rates

Field studies on BWCs are plagued with issues of low base rates. Despite frequent media coverage, the outcomes of interest—use-of-force reports and civilian complaints—rarely occur (Hickman et al., 2008; White et al., 2018). For example, an entire study may record only a small handful of citizen complaints (e.g., Ariel et al., 2015). Consequently, the rarity of occurrence increases the chances of Type II errors (i.e., a statistical error in which no effect is found when it exists) and power. Specifically, low-base rates make it difficult for the researchers to detect a statistically significant effect that may exist between variables. Furthermore, effect sizes at low base rates should be interpreted with caution. Ruscio (2008) has noted that standardized mean differences, like Cohen's d (Cohen, 1977), are sensitive to low base rates when variances are unequal between two populations. That is, estimates of effect size become more imprecise as the variance ratios between two populations increase. Thus, a priori power calculations based on standardized mean differences may be imprecise and/or require large effects. The problems with low-base rates are evident in civilian complaints and the use of force (White et al., 2018).

In some BWC experiments (running a span of 12-months), the mean rate of civilian complaints and frequencies of civilian complaints are extremely low (e.g., Ariel et al., 2015; Headley et al. 2017). These low base rate studies have typically not found significant differences in civilian complaints between experimental and control groups. For instance, in Headley et al.’s (2017) study, there were 66% more civilian complaints in the control (6 complaints) than experimental (2 complaints) condition, post experiment. Intuitively speaking, this is a large percentage in reduction in civilian complaints. ² But, statistically, there is no significant difference between the experimental and control group, possibly due to the very low number of cases in each group. For Ariel et al.’s (2015) study, the effect goes in the opposite direction. There were 100% more civilian complaints in the experimental (2 complaints) than the control (1 complaint) condition—but no statistically significant difference. Again, when the percentage is being presented, it looks like a major difference. However, the raw numbers show that both groups had extremely low occurrences and no real difference can be ascertained.

On the other hand, the base rates for use-of-force cases tend to be higher than civilian complaints, which seems to help somewhat with Type II error. Ariel et al. (2015) found significantly lower use of force cases in the experimental (8 use of force reports) than the control condition (17 use of force reports). Nevertheless, the use-of-force rates are still extremely low. Consequently, Headley et al., (2017) did not find statistically significant differences between the control (9) and experimental (12) condition, potentially due to smaller case numbers or the lack of an actual relationship. One interesting observation in Ariel et al.’s (2015) study is that there was about a 50% difference in the use-of-force rate between the control and the experimental condition—a seemingly large disparity. However, their reported standardized mean difference is SMD  =  0.14 (Ariel et al., 2015, p. 523). According to Cohen's benchmarks (Cohen, 1977), the standardized mean difference is smaller than small. Again, this speaks to the potential issues of calculating effect sizes at low base rates, like the imprecision of effect size estimates at lower base rates.

Another consideration is that the overall sample sizes are small for both studies (i.e., Ariel et al., 2015; Headley et al. 2017). If evenly split between experimental and control conditions, the n per cell ranges from 25 to 27. If the frequency of occurrence in the population is rare, then it should follow that finding an effect with a low base rate would be even harder with a smaller sample. Therefore, there should be fewer instances of police officers using force in areas with fewer civilians. Based on this reasoning, the difficulties with low base rates should be compounded in studies looking at (1) smaller police departments (2) and/or smaller cities.

Although the base rate problem has been acknowledged as a potential problem in BWC studies (e.g., White et al., 2018), the consequences of low base rates have not been thoroughly considered. Low base rates can cause issues with effect size estimates and attenuate power to detect a reliable effect. These issues may be exacerbated in experiments taking place in smaller police departments and/or cities. Researchers have attempted to address these power issues through creative solutions, such as changing the unit of analysis to police shifts. However, changing the unit of analysis to police shifts comes with its own costs.

Unit of Analysis

As mentioned before, some studies suffer from a lack of power when studying smaller police departments. To address this issue, researchers have been using patrol shifts as the unit of analysis, rather than individual officers, to enhance power. Patrol shifts are a period of time when officers are on duty. Researchers have broken down 10- to 12-month stretches of on-duty police work into 6- to 10-h patrol shifts. They have used patrol shifts to dramatically increase the sample size and to increase the probability of detecting an effect of BWCs (Ariel et al., 2015; Ariel et al., 2019; Henstock & Ariel, 2017). For example, the Rialto study increased a sample of 54 police officers to 988 patrol shifts, which provided sufficient power to detect a difference in the use of force reports between experimental and control conditions (Ariel et al., 2015).

However, researchers have not sufficiently addressed potential issues with utilizing patrol shifts as a unit of analysis. First, increasing the sample size does not address the low base rate of outcomes such as use-of-force or citizen complaints. Because patrol shifts are derived from the same officers, studies are not increasing use-of-force reports or civilian complaints in actuality. Only the N is increased. In other words, the 8 use of force reports in the control and 17 use of force reports in the experimental condition found in the Rialto study (Ariel et al., 2015) will remain the same no matter how many patrol shifts are produced. The problems with low base rates remain. Moreover, even negligible differences between the experimental and control conditions can become statistically significant with a large enough N (Faber & Fonseca, 2014). Therefore, it is unclear whether studies finding a difference between use and non-use of BWC are merely detecting spurious effects.

Second, by using patrol shifts as a unit of analysis, the researcher's population of interest no longer becomes police officers, but patrol shifts. Technically speaking, researchers are inadvertently making inferences about the effect of BWCs on patrol shifts, and not the actual police officers. Patrol shifts deal with police behaviors within the frame of time. Whereas, with police officers, behaviors are examined within the frame of a person. Ideally, social scientists should be making inferences about people, the agents, and not a segment of time. Therefore, changing the unit of analysis may not be the ideal solution to power and low base rate issues.

Rather, the best practice would be to recruit more officers at the individual level. Albeit a difficult thing to do, recruiting more individual officers would lead to a general increase in officer contact with civilians. In turn, general contact should increase the likelihood of encounters with unruly civilians, and thus, increase opportunities for the use of force. Also, it could be expected that an increase in officer contact with civilians increase the likelihood of complaints. This would be a more organic and tenable way of increasing power to detect a true effect in experiments with low base rates—in contrast to artificially generating power for the sole purpose of finding a significant p-value. Hinkle et al. (2013) advocate for policing researchers to conduct studies at multiple locations to increase sample size, which would improve the long-term probability of detecting a true effect (Howell, 2013). Moreover, the civilizing effect is an inherently human phenomenon, pertaining to the interaction between officers and civilians. As such, studies of the civilizing effect should be making conclusions about human beings. Examining individuals, as the unit of analysis, is the most valid and direct way of making generalized conclusions about the human population. On the other hand, when the unit of analysis is fragmented into pieces of time, it is harder to make generalized inferences about human population because the entity is no longer human. Further, keeping individuals as the unit of analysis enables social scientists to make “apples-to-apples” comparisons. That is, studies with humans as the unit of analysis could be compared to other studies with humans as the unit of analysis—a practice that would cultivate future comprehensive reviews and meta-analyses to be conducted. On the contrary, time-shift studies are not on the same metric and are, therefore, incompatible with a majority of other BWC studies based in human units. Thus, individual-level recruitment should be the gold standard.

Contamination

During a patrol, police officers are mostly alone, but sometimes, they are paired with a partner. This pairing can lead officers assigned to BWCs (the experimental group) to come in contact with officers not assigned to BWCs (the control group). As a consequence, the control officers are exposed to the experimental manipulation (i.e., the presence of a BWC) and are no longer a neutral base for comparison; the effects of the manipulation in the experimental condition cannot be disentangled from spillover effects in the control condition. Additionally, in police shift studies, a police officer can be randomly assigned to both the BWC and non-BWC shifts—another spillover effect built into a so-called experimental design. Overall, the contamination of control units with experimental units violates the basic assumption of experiments—known as the stable unit treatment value assumption (SUTVA; Sampson, 2010). There are two assumptions of SUTVA: (1) an experimental manipulation has the same effect on an individual no matter how they are exposed to the experimental manipulation and (2) the effect of an experimental manipulation on an individual is independent of the effect of an experimental manipulation on others (Schwartz et al., 2012). When either assumption is broken, cause-and-effect inferences become unreliable (Schwartz et al., 2012). Moreover, violations of SUTVA can produce undue “noise” in the data, which may attenuate the ability to detect an effect. Similar to the low base rate issue, the violation of SUTVA presents a tough obstacle for BWC researchers who work laboriously to design methodologically rigorous field experiments involving randomized control trials. The difficulties with SUTVA violations are illustrated in Ariel et al.’s (2017) study.

Ariel et al.’s (2017) study was essentially a replication of the Rialto study (Ariel et al., 2015) but mainly focusing on civilian complaints. The researchers found a 93% overall drop in civilian complaints from baseline to after the experimental manipulation; however, they did not find a significant difference between the experimental and control shifts. The researchers acknowledged that these findings could have been a consequence of violating SUTVA. Specifically, rule two of SUTVA was violated. Given police shifts were the focus of analysis and not individuals, police officers were randomly assigned to both the experimental and control group. In addition, BWC officers were allowed to make contact with non-BWC officers in the field—further complicating the distinction between the experimental and control condition. The dual exposure of officers to experimental and control conditions may have led to the lack of differences in civilian complaints between subjects but a significant overall reduction in civilian complaints.

Ultimately, contamination can lead to an inability to make accurate causal statements about BWCs. One suggested remedy for contamination in literature has been to randomly assign BWCs to certain squad areas within a precinct rather than assignment by shift (e.g., Hedberg et al., 2017; Morrow et al., 2016; as cited in Lawrence & Peterson, 2020). In this type of design, every officer within one squad would be assigned a BWC while the other squad would not be assigned a BWCs. These designs are intended to reduce the potential for officers in separate conditions to arrive at the same scene. However, there is still room for contamination. With the officers all operating within the same precinct, there will inevitably be contact between officers across conditions before, during, and after shifts (Morrow et al., 2016). While this could be an effective method of reducing contamination, it can never fully be accounted for. This is due to researchers’ inability to curb the contact of control and experimental officers without disrupting the essential operations of a police department. That is, it would be both dangerous and unethical to have officers respond to risky calls alone, with no backup, for the sake of a field experiment. Therefore, Lawrence and Peterson (2020) suggest including a measure of contamination in analyses to allow researchers to attempt to account for its effect on outcomes of interest (i.e., officer behavior). While this is a statistical remedy for contamination, future researchers must account for this issue and attempt to minimize its impact.

Altogether, these methodological issues may negatively impact the results of current BWC studies. Low base rates make effect size estimates unstable, compromising the integrity of statistical inferences. They also weaken the ability to detect an effect, potentially requiring larger samples. Low base rate problems cannot be resolved by replacing the unit of analysis with police shifts. Police shifts artificially increase sample size without actually increasing the incidence of police use of force. With a large enough sample size, even spurious effects can become significant. In addition to low base rates, both experimental and control conditions can be contaminated. Therefore, it becomes hard to ascertain the unique effects of a BWC manipulation, affecting the ability to make accurate conclusions. Despite the limitations of the current methodology, researchers have attempted to address many of these problems and conduct as rigorous experiments as possible. However, rigorous field experiments, alone, are not enough. We advocate for using various mixed methodologies in addition to field experiments to measure the effect of BWCs on officer and civilian behaviors.

Laboratory Experiments

Presently, most of the BWC literature is plagued with internal validity concerns due to the methodological limitations described above (e.g., contamination and low base rates). However, laboratory experiments, the close cousin of field experiments, could provide support in this area. To be clear, we do not claim that lab experiments are a panacea for the existing problems in BWC research. Rather, we see the promise in utilizing laboratory experiments because they provide opportunity for strong internal validity. Researchers who conduct lab experiments can determine when and where participants are exposed to the experimental condition and in what manner. This increased control would allow researchers to separate the experimental and control conditions, thereby eliminating contamination. Moreover, isolated lab settings allow researchers to be more justified in making cause and effect statements. This is due to the ability to control which variables vary and which variables remain constant (Schönpflug, 2001). For example, a researcher would be more justified in claiming a civilizing effect due to the presence of a BWC when given the ability to rule out or hold constant other irrelevant variables (e.g., bystander intervention).

However, as noted, lab experiments are not an all-encompassing solution. Like field experiments, lab experiments have their own shortcomings. In exchange for interval validity, lab experiments sacrifice some degree of ecological validity. That is, lab experiments cannot exactly simulate the same risk, urgency, intensity, and stress as a true police-civilian encounter (Baldwin et al., 2019). In addition, simply being part of a lab experiment can trigger self-awareness in police officers, influencing behavior, which may present unwanted variability in data. Nevertheless, there are procedures that can followed to make lab experiments as ecologically sound as possible and to minimize the impact of lab-induced self-awareness. We will describe these procedures with virtual reality. Virtual reality is certainly not the only laboratory method, but it has been used to study police behaviors, which may be applicable to the study of BWC effects.

Quantitative Field Surveys

Few published BWC studies exist that have distributed quantitative surveys in the field (after a police encounter) to directly measure perceptions of the encounters, with most prior research focusing on general public perceptions of BWCs (see Crow et al., 2017 for examples). However, there have been some attempts to contact citizens typically by telephone to measure perceptions of their recent encounters with police (see White et al., 2017). Studies like this could aid researchers in observing psychological moderators in the real-world. During a field experiment (where BWCs are randomly assigned to officers), researchers could ask officers to distribute surveys directly after an encounter with a civilian. Through likert-based measures, these surveys could assess multiple moderators and dependent variables of interest. For example, the Stress Response Inventory (Koh et al., 2001) could be used to assess a civilian's present levels of stress. The Attitudes towards Police Legitimacy Scale (Reynolds et al., 2018) could appraise the extent to which civilians believe that police officers are legitimate authorities. Finally, officers’ perceptions of self-legitimacy can be measured by the items contained in Nix and Wolfe’s (2017) article on negative police publicity. All of these could then be incorporated as possible moderators in analysis.

Additionally, civilians could be asked to rate the degree to which an officer treats them in procedurally just ways (e.g., Hazen & Brank, 2021; Sahin et al., 2017), and officers can rate whether a civilian was respectful (e.g., Hazen & Brank, 2021; Pickett & Ryon, 2017). This way, there is more of a gradient to the operationalization of civility rather than a binary yes-or-no response to use of force and civilian complaints. Evaluating civility in a continuous manner may have a couple of benefits. One, it may provide better insights into the perceptions of milder expressions of incivility (e.g., verbal disrespect), which could have the potential to escalate into more severe displays of incivility (Roithmayr, 2016). Two, verbal and less extreme forms of disrespect may be more common than outright physical encounters (Pizio, 2014). Therefore, researchers may acquire greater power by collecting data on more minor yet common forms of incivility rather than collecting data on severe but rare occurrences, such as the use of force and civilian complaints. While it may be difficult to survey citizens directly after encounters, it has been utilized in previous studies successfully. For example, Demir and Kule (2020) asked citizens to participate in a survey about their experiences after a traffic stop where they were randomly assigned to officers who were or were not wearing BWCs, with a 73% response rate. If there are concerns about participation, especially after more tense exchanges, the data could be collected shortly after within a few days as used in White et al., 2017.

Overall, quantitative field surveys conducted directly after the encounter, or shortly after, allow researchers to get a better sense of which psychological moderators play a role in the real-world encounters between police officers and civilians. Everyday scenarios, like traffic stops, would be a good way of piloting the logistics of this methodology. By capturing the less extreme interactions, a more continuous measure of civility in a quantitative survey may help to increase power and provide better insights into lesser forms of incivility, which could act as the antecedents to uses of force and civilian complaints.

Qualitative Approaches

Although a quantitative survey can reveal a whole host of information about people's psychological states and tendencies, it often lacks descriptive explanations for why people feel a certain way. Most current studies on officer attitudes regarding BWCs use likert-type items to measure officers’ beliefs (e.g., “I believe that my agency should equip all officers with body-worn cameras”; Kyle & White, 2017). Such items may inform us of general attitudes towards BWCs—but they do not tap into the rationale behind officers’ opinions. However, when quantitative studies do indeed explore the officers’ rationale, they often include the researchers imposing their own questions instead of allowing officers to provide their own reasons (e.g., Gaub et al., 2016; Gramagila & Phillips, 2018; Ready & Young, 2015). Understanding personal rationales could be important for future intervention efforts: If an officer's attitude towards BWCs is found to be key moderator in quantitative data analyses, then interventions should be focused on addressing the reasons why officers may oppose BWCs.

The most common method of collecting qualitative data has been the interview. Studies have used variations of in-person interviews (e.g., structured, semi-structured, one-on-one, focus groups) to probe officer thoughts about BWCs (e.g., Fallik et al., 2020; Goetschel & Peha, 2017; Grossmith et al., 2015; Pelfrey & Keener, 2016; 2018). Other studies have attached few open-ended questions to larger quantitative experiments (e.g., Ariel, 2016b).

However, irrespective of the method, qualitative data can elaborate on the psychological moderators identified in quantitative-based methodologies. They can explain why officers choose to oppose or support the implementation of BWCs in their departments. However, the utility of qualitative approaches is not limited to probing police perceptions. Open-ended questions and interviews could be administered to assess why civilians believe that the police are (il)legitimate or why interactions with officers induce stress. In brief, qualitative data will enable researchers, or commanding officers, to influence the real-world factors which may amplify or attenuate the civilizing effect of BWCs.

Taken together, these additional methodologies—lab experiments using VR, quantitative field surveys, and qualitative approaches—in conjunction with field experiments can provide a more holistic picture of the actual impact of BWCs on civility. These methodologies are certainly not replacements, but a supporting cast for the lead role of field experiments. Boasting high levels of ecological validity, field experiments are invaluable as they demonstrate whether the effects are substantively viable, which is relevant information for policy makers and police departments. However, like any other methodology, field experiments have its limitations. Thus, we have argued that the civilizing effect has been inconsistent in field experiments due to a lack of control and internal validity. Hence, the findings of field experiments, alone, may not be enough to convince the scientific and lay community of its reliability (Lum et al., 2019).

A Note on Practical Significance

Thus far, we have attempted to reconcile and provide methodological recommendations for future research on the civilizing effect. However, one further implication we have not addressed is whether the presence of a civilizing effect in research has practical applications (i.e., is there a large enough effect to make a difference in the “real world”?). This is a difficult question to answer given the evidence presented in this review: inconsistencies in detecting the effect, methodological issues, and unaccounted variables. In addition, variations in unit of analyses and other statistical incompatibilities make it difficult to conduct the meta-analyses necessary to ascertain a reliable overall effect size and p-value estimate (Lum et al., 2020). However, if future research clarifies and establishes the civilizing effect through strengthened and varied research methods, then we believe that any reduction—even if it is small—in use of force incidents and civilian complaints can be beneficial. Countless other rare events (e.g., certain types of victimization, serial killers, school shootings) still command significant and warranted attention in social science research primarily based on prediction and prevention implications alone. The possibility of having even one less child, parent, sibling, or friend injured or die in a use of force encounter is practically important and scientifically worth pursuing. Further, any technology that could help improve police-civilian relations and prevent disputes from escalating into potentially lethal violence is a good investment. Simply put, an effect is an effect.