Introduction

The need to reduce aggression in police-public encounters is widely perceived as a grave public concern (Longridge et al., 2023). Body-worn cameras (BWCs) have been proposed as a measure to mitigate the occurrence of the use of force by law enforcement authorities and have rapidly proliferated in police agencies worldwide (Ariel et al., 2015; Lum et al., 2019; Miller & Chillar, 2022; Petersen et al., 2021). Consequently, over the past decade, numerous studies have been conducted to explore the various effects of BWCs, as outlined in the latest systematic review by Williams et al. (2021). One central conclusion that has emerged is the importance of ensuring that individuals are made aware that an officer is wearing a camera (Choi et al., 2023; Demir et al., 2020; White et al., 2017), and even recommending that law enforcement verbally inform citizens about the presence of BWCs during encounters (Demir, 2023).

However, the efficacy of visual warnings alerting individuals of BWCs remains underexplored. Research has shown that signage advertising CCTV surveillance can constrain behaviour (Piza et al., 2019; Ratcliffe & Rosenthal, 2021). Nevertheless, we are unaware of any comparable studies investigating the utility of signage indicating the presence of BWCs in police-citizen encounters and its effect on behaviour. This knowledge gap is particularly relevant as, in tense situations, a police officer might overlook informing citizens about the camera, necessitating the visual warning as a supplementary measure to the verbal notification. A visual cue could enhance awareness and, thus, reduce the use of force; however, this is a logical rather than an evidence-informed conclusion.

The present experiment aims to address this gap by assessing the effectiveness of visual warnings regarding BWCs in reducing the incidence of use of force, compared to situations where BWCs are present without a visual warning. Importantly, we were also able to control for the effect of the audial warning; the police officers participating in this study were not instructed to proactively announce the presence of the cameras, as they often do elsewhere. Below, we report the results of this study.

Literature review

Use of force

The use of force by police has received significant attention across various academic fields. This scholarly interest underscores the considerable dedication of practitioners and decision-makers to gaining a deeper understanding of how law enforcement agencies exercise their authority and how this power can be effectively managed. The public expects that the police will employ force when circumstances deem it “necessary” and “reasonable” to achieve “legitimate objectives” (Alpert & Smith, 1994). In practice, a defining attribute of policing is the application of necessary and reasonable force in a variety of disorderly situations that may not be desired or anticipated by the public.

It remains unclear what constitutes a “police use of force”, how much force is “too much”, at what point it becomes excessive, exaggerated, or unnecessary, and who instigates the aggressive beyond what is required. Existing tracking systems for the police use of force are not entirely valid or reliable, due to the subjective nature of determining the necessary force in a given situation (Ariel et al., 2016a), and the fact that use of force is nearly entirely self-recorded by only one party: the involved officers.

What is clear is that law enforcement’s legal and ethical duty is to uphold social order, and using force is an indispensable requirement in specific situations, particularly when dealing with certain offenders (Ariel et al., 2015). A police officer can use force when there is an imminent threat to public safety or their own, or when the encounter necessitates the application of physical power (Longridge et al., 2023). In practice, they use these powers rarely — and within these extraordinary circumstances, officers responding disproportionately or unnecessarily with “force” to tense incidents with civilians is even more uncommon (Harrell & Davis, 2020; Adams et al., 1999).

Despite their rarity, these incidents have dramatic repercussions beyond the individual encounter and can harm the relationship between the public and the law. Consequently, law enforcement agencies aim to minimise the necessity of applying force, recognising that doing so can improve public perceptions of the police and, in the process, enhance the efficiency of policing efforts (Ariel et al., 2016a).

BWCs

BWCs were initially proposed as a measure to de-escalate heated situations and prevent aggressive encounters. Over the past decade, numerous communities have witnessed incidents of police shootings and deaths in police custody, as well as citizen protests and calls for improved police transparency and accountability. These events have spurred law enforcement agencies’ rapid adoption of BWCs worldwide, with the hope that these cameras would contribute to improved police conduct, transparency, and accountability, especially concerning the use of force (Lum et al., 2020).

The cameras capture and store recordings of police interactions with suspects, witnesses, or victims (Ariel, 2016a; Ariel et al., 2018a, b). Although BWCs may come in various forms, they generally share two primary goals. Firstly, they aim to record the engagement and by doing so, to discourage escalations through deterrence of both parties. Secondly, they have utility throughout the criminal justice system. For example, BWCs footage can be used as incriminating or exonerating evidence (Petersen & Lu, 2023) as well as civil litigations (Powell, 2023); they can improve witnesses’ accurate recall of events (Newell & Koen, 2023), and various other applications unavailable just a decade ago (Petersen et al., 2023).

There has been widespread adoption of BWCs technology by police departments worldwide, which mirrors a growing body of evaluation research on the impacts of BWCs (e.g., Ariel et al., 2015; Jennings et al., 2015; see reviews in Lum et al., 2019; 2020). For example, it was found that BWCs lead to a significantly higher likelihood of reporting low-level street crimes (Ariel, 2016a). BWCs were linked to a reduction in citizen complaints against the police (Ariel et al., 2015; 2017) and civilian deaths (Miller & Chillar, 2022). On the other hand, results from several studies suggest that the deployment of BWCs led to an increased rate of assaults against officers who were wearing the cameras (Ariel et al., 2016b; 2018b) — the inverse of the effect found in non-policing environments, where assaults were found to be reduced (Ariel et al., 2019; Ariel et al., 2024). The authors attribute the increase in assaults in the policing context to officers’ heightened awareness that their actions are being observed, leading to potential excessive self-scrutiny, which may impair their ability to function effectively in high-stress situations. However, Boehme & Schnell (2022:1) recently proposed alternative and more convincing explanations, suggesting that the findings may be attributed to officers either reporting use-of-force incidents more frequently or feeling more confident using force due to enhanced oversight (e.g., “I know I am right for applying force”).

Given this interest in the potentially civilizing effect of BWCS in policing, one of the studied variables concerning the impact of BWCs is the use of force by officers. However, studies have not provided consistent results. On the one hand, some studies found that BWCs reduce the use of force by officers compared to those without BWCs (Ariel et al., 2015; Braga et al., 2018; Henstock & Ariel, 2017; Jennings et al., 2017). For instance, officers with BWCs had a significantly lower percentage of use of force incidents compared to officers without BWCs with a similar number of use of force incidents in the previous year (Groff et al., 2020), for several years later (Sutherland et al., 2017). On the other hand, other studies found no significant differences in the use of force by officers with or without BWCs (Ariel, 2016b; Ariel et al., 2016b; Peterson et al., 2018; Yokum et al., 2017).

Given these variations in results, a Campbell systematic review concluded that there is no discernible effect on the behaviours of most police officers or citizens (Lum et al., 2020), but a more recent meta-analysis of the evidence identified a noteworthy decline in police use of force as a result of using BWCs (− 9.6%; CI − 21.3% to + 3.8%; Williams et al., 2021, p. 11–12).

(One possible) mechanism behind BWCs

Puzzled by such variations in the results, several explanations surfaced. Some argue that the outcome variation has to do with local discretion rules and how police departments use BWCs (Ariel et al., 2016b). As the deterrence effect of BWCs is a function of discretion, weaker discretion is associated with a strong deterrence effect and less forceful police responses (Ariel et al., 2018a).

Another explanation is the awareness of the camera, which is the focus of this study. The assumption of BWCs is that when people are aware of the presence of recording devices, it sends a clear signal to everyone involved: “Beware, you are being filmed” (Ariel et al., 2018a, b). This public self-awareness can often improve conduct, as people do not want to get caught on camera for wrongdoing (Miller & Chillar, 2022).

Thus, linking BWCs to both civilian and officer behaviour revolves around two complementary theories: self-awareness and deterrence theories (Ariel, 2016a; Ariel et al., 2015; Ariel et al., 2018b). Some recent studies have dealt directly with the psychological and practical manifestation of the effect of BWCs (particularly insightful contributions can be found in Bennett et al., 2023; Patterson & White, 2021; Pezdek, 2022). Firstly, studies suggest that people change their behaviour when they are aware that they are being observed (Chartrand & Bargh, 1999; Dzieweczynski et al., 2006; Jones & Nisbett, 1971). In other words, people adhere to social norms, cooperate with rules, and modify their behaviour due to the awareness of being watched. In the context of BWCs, cameras can induce a “self-awareness effect” that motivates the police officer away from responding with excessive or unnecessary force and simultaneously alleviates the suspect’s aggressive behaviour due to the awareness of being filmed (Ariel et al., 2015; 2017; Ariel, 2016a).

The second theory that may explain changes in human behaviour is the criminological deterrence theory (Beccaria, 2008). Deterrence theory emphasises the need for a heightened awareness of someone not only observing our behaviour but also that they will apprehend and punish us for wrongful doing (Ariel et al., 2018b). According to this theory, supported by a substantial body of research, a higher certainty of being apprehended corresponds to a lower likelihood of engaging in socially and morally unacceptable behaviour because people are averse to the risk of getting caught (Nagin, 2013; Pratt et al., 2006; Von Hirsch et al., 1999). As BWCs document police-public interactions, utilising video recordings to provide clearer evidence of inappropriate or criminal behaviour by those involved, the camera increases the likelihood of someone being apprehended and functions as a deterrent against improper behaviour (Ariel et al., 2018a). Evidence broadly supports this deterrent and civilising effect of BWCs in heated police-public encounters (Ariel et al., 2015; Braga et al., 2018; Henstock & Ariel, 2017; Jennings et al., 2017).

Awareness of BWCs

Thus, BWCs are hypothesised to reduce the number of use-of-force incidents through deterrence and self-awareness, with a complex interplay between these two elements. Yet regardless of the specific causal mechanism — whether deterrence or public self-awareness — there is agreement that those involved in a situation must be cognisant that the police officer is wearing a camera and that the interaction is officially recorded (Ariel et al., 2015, 2020; Clare et al., 2019; Demir et al., 2020; Saulnier et al., 2020). Studies comparing officers without BWCs to those who were required to inform individuals that the interaction was being recorded found that informing citizens about the use of BWCs significantly improved their perception of various measures (Ariel et al., 2015, 2020; Clare et al., 2019; Demir et al., 2020; Saulnier et al., 2020), including increased cooperation with law enforcement authorities (Demir et al., 2020). When law enforcement informs citizens that the encounter is being recorded, it indicates transparency, leading to increased compliance from citizens and fairer behaviour from law enforcement (Demir, 2023). Accordingly, White and colleagues (2017) emphasise that most civilians (70.2%) reported a willingness to cooperate more with the police when aware of BWCs.

However, the degree of awareness that people have to the devices is still unclear (and may be a function of time). Public awareness of BWCs may occur through three primary avenues: (1) a belief or assumption that officers are required to wear them during encounters; (2) verbal notification by the officer during the encounter that a BWC is in use; (3) visual cues of the camera as perceived by the civilian (Ariel et al., 2019; Demir, 2023).

To date, most research on civilian awareness of BWCs during encounters with police has focused on the first two methods. For example, Demir’s study (2023) examined the effects of awareness and notification of BWCs on civilian perceptions across various measures within four groups: non-BWCs, unawareness of BWCs, awareness of BWCs, and notification of BWCs. The results showed that notification of BWCs use had the most significant impact on civilian perceptions overall, followed by BWC awareness, lack of BWC awareness, and non-BWCs (although there was no significant difference between BWC notification and BWC awareness). Accordingly, notifying individuals about the use of BWCs during an encounter may enhance citizens’ perceptions of law enforcement.

Another method for individuals to be aware of the camera involves recognising visual cues displayed on the camera itself. BWCs provide both visual and audio alerts when activated, even when there is no explicit notification to the public that the camera is in operation. The camera’s size, audio-visual indicators, and mounting configuration convey a message that the encounter is being recorded. However, studies have shown that the public is not always aware that officers are wearing BWCs that are recording (Seguin, 2019; White et al., 2017). This lack of awareness may be attributed to the inconspicuous nature of BWCs used in some public spaces. Many witnesses, victims, and suspects struggle to identify BWCs among the various gadgets that modern police officers wear daily (Ariel et al., 2019). The size and colour may also impact the public’s ability to detect the camera. As Timan (2016) pointed out, in the current BWC setup, the actions of aiming and recording may not be recognised by the untrained eye. Not only are the public unaware of the presence of the BWC, but they may also find it challenging to identify it as a camera. Hence, the presence of visual cues of BWCs is of great importance in raising awareness of the camera among the public, ensuring that the mechanisms influencing behaviour take effect.

The present study

Advertising the presence of CCTV surveillance cameras has been identified as a factor that constrains behaviour (Piza et al., 2019; Ratcliffe & Rosenthal, 2021). This logic stems from the understanding that covert cameras in the public domain are unlikely to deter offenders (Piza et al., 2019). It can be presumed that the presence of a visual warning on BWCs has the potential to increase awareness among civilians, leading to greater cooperation with the officer and, consequently, reducing the use of force by officers (Alpert et al., 2004). However, manipulating a visual cue has thus far not been studied in the context of BWCs.

Examining the potential of visual warnings of BWCs on the use of force by officers is crucial, as, in tense situations, a police officer might overlook informing citizens about the camera, making the visual warning a supplementary measure to the verbal notification, which has proved crucial during an encounter with a civilian (Demir, 2023; Demir et al., 2020). Importantly, to isolate the effect of a visual warning, it should be separated from the verbal warning officers equipped with BWCs usually make when engaging with suspects. To properly test the effect, conditions should involve comparing BWCs with the visual signage to BWCs without the signage while controlling for the effect of warning suspects that their behaviour is being filmed. Under these settings, the current research examines the incidence of use of force by officers through the hypothetical causal chain which we anticipate:

$$BWC+visual\;warning\rightarrow awareness\;increases\rightarrow suspect's\;behavior\;''\mathrm{cools}\;\mathrm{down}''\rightarrow officers\;do\;not\;react\;aggressively\rightarrow fewer\;instances\;of\;use\;of\;force\;by\;officers.$$

The present study attempts to provide an empirical assessment of this hypothesis.

Methods

Setting

The data for the study were obtained from a cluster-randomised controlled trial conducted by the Miami Beach Police Department (MBPD). The trial spanned a six-month intervention period lasting from January 1, 2017, to June 11, 2017 (thus being one of the early adopters of BWCs, when public awareness of these devices was limited relative to more contemporary policing). The city had a residential population of approximately 91,000, ranking as the 26th largest city in Florida. Miami Beach is geographically divided into four primary squad areas, representing the northern, central, southern, and entertainment districts. The city’s crime dynamics have historically been influenced by a high volume of nightlife, tourism, and associated drug use and distribution, resulting in more than 8000 index crimes per 100,000 population in 2019 (Petersen et al., 2021).

Unit of analysis

Thirty-six police spatiotemporal units, comprising the entire frontline officersFootnote 1of the Miami Beach Police Department, were randomly assigned to one of two groups: (1) Experimental group: officers who wore BWCs featuring yellow stickers labelled “VIDEO & AUDIO” (used for a visual warning); (2) Control group: officers who also wore BWCs but without yellow stickers (see Fig. 1).

Fig. 1
figure 1

Two types of BWCs — with and without yellow stickers

Full size image

The study included all officer shifts of the MBPD over the six-month research period. No frontline units, shifts, or events were excluded from the sample, except for incidents involving the use of force within the police station (as opposed to using force during encounters) and specialised operations. The final sample of police-public contacts included 28,393 recorded interactions during the experimental period, sourced from official police records and the agency’s computerized dispatch system.

Unlike prior trials that focused on specific spatial hotspots (Ariel, 2016a), individual officers (Yokum et al., 2017), or police shifts (Ariel et al., 2020), this study integrated time, space, and police officers (Braga et al., 2021, 2022; Petersen et al., 2021). Specifically, geographic squad areas were grouped by specific shifts and days, creating 36 spatiotemporal units consistent throughout the research period. These were randomly assigned to either the treatment or control group. Furthermore, no officers assigned to the experimental group were assigned to the control group. Those assigned to the same area had different working days or shifts, while those assigned to the same days and shifts were located in other areas.Footnote 2 For some smaller territorial units, the patrol pattern was not executed daily but decided biannually by the executive teams. The day and shift components of the experiment included separate shift assignments: day, afternoon, and midnight, and consecutive sets of days, such as Monday, Tuesday, Wednesday, Wednesday, Thursday, and Friday. The included areas covered all four primary squad regions of the city, along with several smaller patrol beats within these regions (Petersen et al., 2021). A full randomisation table is presented in Table 1.

Table 1 Miami Beach territorial policing, clusters randomisation table
Full size table

Randomisation

Cluster randomisation, as we have 36 spatiotemporal units, was employed to prevent treatment contamination and breach the stable unit treatment value assumption (see Braga et al., 2021). This precautionary measure aimed to prevent situations where officers from the experimental and control groups might respond to the same incident (Ariel et al., 2022). To minimise the potential for interaction between the treatment and control groups, all police officers operating within the same geographic area, on the same days and during the same shift, were randomly assigned.

The police department determined the cluster sizes and patrol time, and the research team was responsible for generating the random allocation sequence, enrolling participants, and assigning participants to intervention. The allocation of squad shift/area/day combinations was performed using a simple random assignment generator, ensuring that each unit had an equal probability of being assigned to either the treatment or control group. This procedure resulted in 17 experimental clusters and 19 control clusters.

Intervention

The independent variable of interest is whether officers wore BWCs featuring yellow stickers labelled “VIDEO & AUDIO” to inform the public of the presence of the camera, or were in the control group (BWCs without the yellow stickers).

Crucially, the officers in Miami Beach received three levels of training. First, they were trained on using BWCs by the camera company. Second, they were trained on how to use the BWCs in police operations, with a focus on the guidelines stipulated in the Standard Operating Procedure (SOP #152).Footnote 3 The SOP explicitly does not require officers to announce that they are using BWCs during encounters, a protocol maintained in both the treatment and control conditions. Finally, the officers were given information about the experiment and its purpose.Footnote 4

Measures

Use of force

The dependent variable of interest is the use of force by police officers, classified as “yes” or “no” based on officers’ self-reporting, standard practice in most police departments. This variable includes any physical incident regardless of severity, as part of the SOPs, except for verbal violence.

Use of force level

As mentioned above, our research aimed to determine whether the presence of a visual warning of BWCs affects the occurrence of any use of force by police officers. However, use-of-force incidents encompass a spectrum of situations ranging from low-level confrontations to high-intensity encounters. Given the low incidence of force used in our study as detailed in the next chapter (0.2% compared to 99.8%) and the substantial overlap between categories of use of force, analysing levels of force could lead to statistical challenges and potentially unreliable results (small cells of data). Thus, we opted for a binary dependent variable, aligning with established studies (such as Braga et al., 2018). Nevertheless, we recognised the importance of controlling for the type of force used to manage the inherent heterogeneity in these incidents, allowing for a more comprehensive analysis considering the diverse nature of police operations. Hence, we attempted to control for the type of force officers chose to apply: physical, physical/leg sweep, k-9 bite, taser, and physical/taser.

Arrest

We incorporated the binary variable “arrest” as a covariate in the model. Arrest serves as a contextual factor (Morgan et al., 2020), and it was necessary to control for any confounding effects it might have on the relationship between the sticker and the use of force.

Statistical analysis

Univariate and then more complex models were analysed with clustered Poisson regression models (based on the spatiotemporal units) to assess group differences. Significant differences were evident between the clusters, implying heterogeneity and affirming the rationale for employing cluster analysis (Torabi et al., 2019). Group assignment (an experimental group with stickers and a control group without stickers) was set as an explanatory variable, and the dependent variable was whether the use of force occurred. In the more elaborate model, arrest, the use of force, and the encounter initiator variable (proactive/reactive policing) were included.

We chose a Poisson regression model because the dependent variable exhibits a right-skewed distribution (positively skewed distribution), and the variance and mean of our dependent variable were nearly identical (M = 0.001769 and V = 0.001766). We confirmed our choice using a goodness-of-fit test, which showed that the Poisson model is better suited for our data than other approaches.Footnote 5

Subgroup analyses

We observed the effect based on the encounter’s initiator (proactive vs. reactive policing), as this factor can produce different outcomes according to our hypothesised mechanism. When the officer initiates the encounter (e.g., stop-and-account), it may introduce an additional treatment effect compared to situations where officers respond to a received call for service. In the present trial, reactive policing refers to 911 calls, telephone calls, walk-ins, and teletypes. Forty-seven percent of the interactions (n = 13,335) were proactive, while 53% (n = 15,058) were reactive policing. We measured relative changes to evaluate differences in the use of force in this analysis. This analysis was conducted separately for the experimental group (with the yellow stickers) and the control group (without the stickers). Table 2 presents an overview of the distribution of encounters across the subgroups in both the control and experimental groups, which provides evidence of the randomisation process producing equivalent groups.

Table 2 Distribution of encounters within the experimental arms: proactive versus reactive encounters
Full size table

Results

Unadjusted outcomes

During the experimental period, 51 incidents of police use of force were recorded, indicating that in only 0.2% of the police-public encounters, officers applied force.Footnote 6 Out of the 51 instances of force, 29 cases occurred in the experimental group (0.22% of their encounters), resulting in a rate of 2.2 per 1000 encounters. In contrast, the control group used force in 22 encounters (0.14% of encounters), resulting in a rate of 1.4 per 1000 encounters. Calculating the percentage change between the experimental and control groups, the force deployment rate in the experimental group was 57.14% higher than in the control group.

Main effects

Univariate Poisson regression model

We first estimated the treatment effect without any coefficients except the assigned stimulus (Table 3), which mirrored the unadjusted outcomes, with an increase rather than a reduction in the use of force during treatment conditions compared to control conditions. The use of a visual warning was found to increase the odds of use of force by 155.03% relative to control conditions, although with a wide margin of error (B = 0.4384; SE = 0.2827; Incident Rate Ratio (IRR) = 1.5503; 95% CI 0.8907, 2.6982). The estimated average marginal effect of a visual warning variable of BWCs on the probability of use of force was 0.07875%, indicating a small increase in the probability of using force when the sticker was present compared to control conditions. Moreover, the estimated N increase, along with the IRR, was 22.38. However, the 95% confidence interval ranged from − 0.0002311 to 0.001806, indicating that the true effect size could be zero.

Table 3 Univariate Poisson regression main effect estimates for a visual warning of BWCs on use of force (N = 28,393): coefficients, standard error, incident rate ratio (IRR), and 95% confidence intervals (CI)
Full size table

Clustered Poisson regression model

Similar results were found under the clustered design but with greater precision of the estimates. We found a statistically significant treatment effect of a visual warning of BWCs on the use of police use of force — but in the opposite direction to the predicted effect. Treatment officers who wore BWCs featuring yellow stickers labelled “VIDEO & AUDIO” were involved in more incidents of force than the control group without these stickers. As detailed in Table 4, on average, using a visual warning is associated with 146.82% higher odds of use of force by officers compared to cases without a visual warning (B = 0.9035; SE = 0.3483; p < 0.01).

Table 4 Clustered Poisson regression effect estimates for a visual warning of BWCs on use of force (N = 28,393; 36 clusters in total)): coefficients, standard error, incident rate ratio (IRR), and 95% confidence intervals (CI)
Full size table

Subgroup analyses

Overall, in proactive policing engagements, we found a 204.75% higher overall rate of use of force by a police officer, an outcome consistent in both experimental groups (B = 1.1143; SE = 0.3658; p < 0.002). However, when examining the use of force within proactive policing interactions, the sticker led to more frequent use of force (3 cases of used force per 1000 cases). In contrast, in reactive policing, the use of force was less likely (1 case of used force per 1000 cases). In the treatment group, proactive policing was associated with 200% more incidents of force use than reactive policing. In the control group, a similar trend was observed, albeit weaker compared to the treatment group: In proactive policing, there was a relative increase in the use of force (2 cases of used force per 1000 cases) compared to reactive policing (less than 1 case of used force per 1000 cases). Thus, proactive policing was associated with a 150% higher use of force than reactive policing when the officer did not wear a sticker. Overall, as shown in Fig. 2, officers with the sticker were more likely to use force than officers without the sticker, both when comparing proactive policing between the groups and when comparing reactive policing between the groups.

Fig. 2
figure 2

Rates of use of force per 1000 police-public encounters: comparing proactive policing to reactive policing (N = 28,393)

Full size image

Discussion

Several studies have examined the effects of awareness and verbal notification of BWCs across various measures (Ariel et al., 2020; Clare et al., 2019; Demir, 2023; Demir et al., 2020). Some concluded that law enforcement should verbally inform citizens during encounters about the use of BWCs to enhance the cameras’ effectiveness (Demir, 2023). However, no research into the use of visual warnings that alert individuals about the presence of BWCs has been conducted until now. Amidst the tension of a police-citizen encounter, it is plausible that officers might overlook informing the suspect about the camera, therefore, visual cues to increase awareness could be necessary for the BWC effect to materialise. A clearly visible sticker on the camera that indicates to the citizen that a BWC is present can supplement verbal notification, promoting civilian awareness of the camera’s presence.

The current research aimed to address this gap and assess the effectiveness of a visual warning of BWCs (via stickers) in reducing the use of force by officers compared to BWCs without a visual warning. We were able to isolate the visual cue effect from the audial message, as officers in Miami Beach were, perhaps uniquely, not required to inform members of the public that a BWC was present.

We found a significant difference in the use of force, but in the contrary direction to our hypothesis: Officers in the experimental group, equipped with BWCs with yellow stickers for visual warning, were significantly more likely to report the use of force than officers without a visual warning. The causal mechanism behind this finding is puzzling. While a visual warning should theoretically increase awareness, foster compliance, and reduce aggression by all involved through deterrence and self-awareness, our results indicate the opposite. This prompts the question: Why did the visual warning of BWCs exacerbate the use of force?

One possible explanation involves the integration of two theories: reactance theory (Brehm, 1966) and procedural justice theory (Tyler, 2006). First, when people perceive a threat to their freedom or autonomy, they may respond by engaging in certain behaviours to regain their perceived independence. This reactive response can manifest in a hostile and aggressive manner (Steindl et al., 2015). The awareness of being observed, which may be facilitated by BWCs, may be perceived as a threat to their autonomy, potentially reducing self-control and amplifying reactive behaviour (Esmark et al., 2017).

Second, procedural justice theory highlights the significance of police interactions in shaping the public’s perceptions of the police and their legitimacy. This framework focuses on the process by which police officers enforce laws and interact with citizens rather than solely on the outcomes of these interactions. The absence of procedural justice can significantly deteriorate the relationship between police and suspects, potentially leading to extreme responses, such as citizens attacking officers (Tyler, 2006). When individuals perceive their treatment by police as unfair or unjust, it often generates frustration and anger (Sunshine & Tyler, 2003). This perception of unfairness undermines police legitimacy, leading to non-compliance or even active resistance against law enforcement authorities (Langley et al., 2021; Tankebe, 2013). Human emotional responses play a critical role in this dynamic; feelings of disrespect, humiliation, or threat can trigger aggressive reactions, particularly in high-stress encounters with police (Jonathan-Zamir & Weisburd, 2013). Moreover, in communities where distrust of the police is prevalent, there may be a collective mindset that encourages hostility towards officers, sometimes escalating to physical aggression (Weitzer & Tuch, 2005).

In this experiment, the stickers may have heightened the perceived sense of monitoring and posed a threat to citizens’ autonomy. Crucially, since the officers did not announce or explain to members of the public why the BWCs were in use during the contact, the devices may have intensified the threat to their autonomy. Ariel et al. (2016a) suggested that the transparency offered by BWCs could enhance perceptions of fairness and legitimacy in police-public encounters, implicitly indicating that failing to inform individuals about being recorded could undermine these perceptions. Individuals in police-public encounters may find the lack of a simple verbal statement, such as “you are being filmed”, disconcerting. They may feel that the effort required to notice the stickers and understand their significance places an additional burden on them, potentially leading to annoyance and resistance towards the officer. Therefore, instead of having a civilising effect, awareness of the yellow stickers triggered an adverse reaction manifested in reactive behaviour, including resistance or aggression. According to the causal mechanism we depicted earlier, the yellow stickers seem to have triggered an escalation of antagonistic behaviour from the public, leading officers to use force in response. As a result, officers used more force in efforts to control the belligerent situation.

This explanation is consistent with studies that have shown that the actions and resistance of suspects during police-public encounters escalate police use of force (e.g., Alpert et al., 2004). Therefore, the self-awareness of being observed was expected to lead to a calmer demeanour in suspects. However, the intentional failure of police to inform people that BWCs are recording them represents a breach of procedural fairness — a lack of transparency, fairness, and respect in interactions between law enforcement and the public (Ariel et al., 2015) — which resulted in more aggressive encounters and more use of force.

Admittedly, these postulations are speculative without direct evidence, and the mechanisms involved remain within a “black box”. Future studies should examine these complexities more carefully. Further research is required to delve into whether audible warnings of BWCs are more effective than visual warnings, as suggested by studies across various fields (Bella & Silvestri, 2017).

Under the bonnet of the effect of yellow stickers in proactive or reactive encounters

Beyond the impact of a visual warning on the use of force by police officers, we also examined the influence of the encounter initiator (proactive or reactive policing) on the use of force, separately for the experimental group (with yellow stickers) and control group (without the stickers). Our findings indicate that in proactive policing, officers are more likely to use force, whereas in reactive policing, the use of force is less likely. This outcome was observed among both research groups. This finding can be explained by the nature of proactive policing, where law enforcement actively seeks out situations that may be problematic (Weisburd & Majimundar, 2018), often engaging with the public while searching for signs of “criminal intent” and “anticipated future criminal behaviour”. Consequently, the experience of being stopped and the “feeling of suspicion” is associated with a decreased willingness to cooperate with law enforcement (Tyler et al., 2015). Therefore, officers involved in proactive policing may find themselves in situations where individuals are less willing to cooperate, leading to an increased tendency to use force as a defensive or precautionary measure.

Our findings align with those of Terrill and his colleagues (2023), who found that officers were four times more likely to use escalated force in encounters during proactive policing compared to reactive policing. They argue that in proactive policing, officers feel a heightened necessity to take control of the situation due to a lack of information, as they take responsibility for intervention rather than responding to public calls. Hence, given the frequent encouragement of officers to adopt proactive policing in combating crime (Lum et al., 2020), it is advisable to consider research findings and previous studies (such as Terrill et al., 2023) that highlight the increased use of force as a potential cost associated with proactive policing, within the framework of BWCs.

Limitations and future research

We examined the use of force based on officers’ self-reports, and we cannot corroborate what occurred during the engagements. There may have been undocumented incidents of use of force in either one of the experimental groups or levels of force applied that went beyond or below the recording provided by the officers involved.

Moreover, we lack unmediated information about which party instigated the use of force, evidently an essential aspect of understanding the dynamics of such incidents (Engel et al., 2000). In addition, although the officers were not instructed to verbally inform the public that the camera was recording them during the encounter, we lack information on whether they adhered to this rule. Therefore, it is not clear whether the stickers in the control group experienced nil audial warning conditions (i.e., spillover) and, by implication, affected civilians’ awareness of the BWCs. Generally, we are uncertain if the officers consistently followed the policy regarding the operation of the BWCs and may have, on some occasions, notified civilians of the BWCs. However, this potential treatment contamination creates more stringent experimental conditions for a statistically significant (backfiring) difference between treatment and control arms to be detected. Thus, the reduced intergroup variance caused by the spillover effect has likely diluted rather than exacerbated the magnitude of the effect size, with a reduced likelihood for a Type I error. Still, future research should analyse BWC footage to understand the effect of the BWCs being equipped with visual warning signage. This analysis could incorporate systematic social observations, preferably by analysing the BWC footage to understand these mechanisms empirically rather than relying solely on logical deductions.

Thirdly, MBPD possesses unique characteristics, and the officers and incidents involved in our study may not be fully representative of those in other jurisdictions. The distinct nature of Miami Beach, marked by a vibrant night-time economy and substantial population fluctuations during the holidays, sets it apart from typical police jurisdictions.

Fourthly, the directive given to all officers not to verbally inform the public that they were wearing BWCs is specific to this context. While this directive offered an optimal setting to assess the effect of a visual stimulus independently of a verbal warning, it distinguishes Miami Beach from other locations (Petersen et al., 2021).

Finally, it is essential to acknowledge that the yellow sticker could alter officers’ psychological landscape, potentially affecting their behaviour during interactions with the public. This effect might manifest in various ways, such as increased adherence to protocols, more cautious behaviour, or a heightened sense of security in exercising authority due to the perceived protection offered by video evidence (Ariel et al., 2018a). Future research should investigate these potential behavioural changes among officers, considering both the immediate and long-term impacts of such visible recording indicators. Furthermore, examining yellow stickers’ dual impact on civilian and officer behaviour could provide valuable insights into the complex interplay between transparency, accountability, and public trust. This can help develop more effective policies that maximize the benefits of BWCs while addressing unintended consequences.

Conclusions

It is evident that a visual indicator of recording is a necessary element to signal surveillance and potential accountability; however, this alone is not sufficient to induce the hypothesised civilising effect of BWCs in police-public engagements. As a standalone stimulus, a visual warning can have unintended consequences, potentially escalating the use of force by officers, possibly due to the lack of procedural fairness in the encounter. Instead, it points to the complexity of human behaviour under surveillance, suggesting that for BWCs to have the civilising effect envisioned, verbal notifications may be more effective than visual cues alone. Further research is needed, preferably with a larger sample size and more diverse populations.