abstract. In the wake of national outrage and polarization over several high-profile police shootings of unarmed citizens, reformers have called for police officers to wear body cameras. This Note argues that, despite the seeming objectivity of the camera, video footage remains susceptible to biased interpretation by observers such as grand jurors. Reporting empirical findings based on mock jurors’ perceptions of real police footage, this Note observes that viewers’ prior attitudes toward the police color their interpretations of the events caught on tape, resulting in considerable polarization on a variety of dimensions. Further, this Note finds that video evidence does not conclusively outperform nonvideo testimony in minimizing mock jurors’ reliance on their prior attitudes. Study participants learned about an incident involving a police officer and a citizen in one of four ways. Some participants watched a video of the altercation, others read dueling accounts of the altercation written from the perspectives of the police officer and of the citizen, a third group read a single account from the perspective of a disinterested third party, and a final group read only the police officer’s version of events. Participants’ prior attitudes toward police significantly affected their judgments of the officer’s conduct in all four conditions, and the degree of bias did not differ significantly across the different types of evidence. Furthermore, people who identified strongly with the police—but not those who identified weakly—became more confident in their judgments when presented with video evidence. This Note discusses the implications of these findings for the policy debate over body-worn cameras, cautioning against the assumption that body cameras will reduce polarization and societal conflict following instances of use of deadly force by police. It concludes that we should be more skeptical of the widely held belief that video footage tells us unambiguously and definitively what happened.

author. Yale Law School, J.D. candidate; Yale University Department of Psychology, Ph.D. candidate. I am deeply indebted to Tom Tyler, who generously supervised this Note; to Avani Sood, Yael Granot, Sara Burke, Jamie Luguri, Gina Roussos, Nicholas Santascoy, and Robert Manduca for insightful comments on previous drafts; to Hannah Thai for invaluable research assistance; and to John Rafael Perez, Rebecca Lee, Michael Clemente, and the editors of the Yale Law Journal for tremendously thoughtful editing.

Introduction

In December 2014, President Obama announced the Body Worn Camera Partnership Program, a new initiative to purchase fifty thousand body cameras for use by police officers across the country.1 The proposal was a response to the fatal police shooting of Michael Brown, an unarmed African American teenager. Brown’s death in Ferguson, Missouri, at the hands of Darren Wilson, a Caucasian police officer, sparked weeks of protests decrying police misconduct and racial profiling.2 The disputed circumstances surrounding Brown’s death polarized the nation.3 A poll administered in Ferguson three months after the shooting found that 71% of Caucasian respondents believed that Wilson was seriously injured before he shot Brown, whereas only 9% of African American respondents agreed.4 A nationwide poll found that Democrats were over three times more likely than Republicans to say that Wilson was at fault and deserved punishment.5

A grand jury decision not to indict Wilson sparked further protests and further polarization. A Washington Post poll conducted after the non-indictment found that nearly 60% of Caucasian respondents approved of the grand jury’s decision not to indict Wilson, whereas fewer than 10% of African American respondents approved.6 Additionally, more than 75% of conservative Republicans approved of the decision, compared to 24% of liberal Democrats.7 Overall, 48% of the respondents approved of the decision and 45% disapproved.8

Many commentators lamented that if only the incident had been captured on camera, we could have known what happened and could have avoided the wrenching societal conflict over the shooting.9 A writer for Time magazine observed, “To many, a camera on Wilson’s uniform would have ended the uncertainty and potentially avoided the subsequent tumult that engulfed the St. Louis suburb.”10

As the country grappled with how to move forward, the months following Brown’s death brought a steady drumbeat of high-profile police killings of African American citizens. In July, Staten Island resident Eric Garner was killed by New York Police Department (NYPD) officer Daniel Pantaleo, who sought to arrest Garner for allegedly selling untaxed cigarettes.11 A video recorded by a bystander showed that Pantaleo put Garner in a chokehold, a maneuver banned by the NYPD, and ignored repeated pleas from Garner that he was unable to breathe.12 In November, twelve-year-old Tamir Rice was shot by Cleveland police officers who mistook the boy’s pellet gun for a real firearm.13 Surveillance videos captured the shooting as well as the officers’ failure to administer timely first aid to the boy, who died the following day.14 In April 2015, Walter Scott was shot eight times in the back while fleeing from officer Michael Slager of the North Charleston Police Department, who had pulled Scott over for a broken taillight.15 Slager initially claimed that he had feared for his life, but an amateur video later surfaced showing that Scott was running away when Slager fired.16

As the list of African American men and boys killed by police grows steadily longer, fueling the Black Lives Matter protest movement, advocates for reform have enthusiastically embraced the idea of putting cameras on police officers.17 Reformers plainly expect that more video footage will lead to more indictments against officers who use excessive force. Indeed, advocates calling for all state and local police to be required to wear cameras have seen fit to name their proposal the “Mike Brown Law.”18

It was perhaps natural for the White House, in the wake of Brown’s death, to turn to body cameras as a solution. In a policy realm with few areas of agreement, body cameras are widely popular. A Pew Research Center poll conducted in December 2014 found that 87% of respondents thought body cameras were a good idea.19 The support was bipartisan: 79% of Republicans, 90% of Democrats, and 88% of Independents favor the reform.20 The numbers are similarly high among African American (90%), Hispanic (89%), and Caucasian (85%) respondents.21 Notably, these figures come from a survey conducted in the days following the Staten Island grand jury’s decision not to indict Daniel Pantaleo in Eric Garner’s death, even though the episode was captured on video. Following the non-indictment, Garner’s father told reporters that the White House’s initiative was “[t]hrowing money away. Video didn’t matter here.”22 But even after such a high-profile instance of video footage failing to secure an indictment, support for the body camera reform remained high. A poll conducted five months after Pantaleo’s non-indictment found that 93% of Caucasian respondents and 93% of African American respondents favor putting video cameras on police officers.23

To those who feel that police officers too often get away with murder, body cameras promise to collect the evidence needed to hold police officers accountable. To those who feel that civil rights activists have jumped to conclusions too quickly in ambiguous cases, body cameras offer hard facts that could potentially exonerate officers falsely accused of misconduct. Indeed, despite initial resistance from police departments,24 precincts that have adopted lapel cameras have largely come to embrace them as a much-needed deterrent to frivolous lawsuits.25 Even the American Civil Liberties Union, normally an opponent of increased government surveillance, sees body cameras as a “win-win.”26

The current policy debate over body-worn cameras has highlighted numerous advantages and disadvantages of putting cameras on police officers.27 Many proponents support body cameras because they believe the police will use unnecessary force less often if they know they are being recorded.28 Indeed, promising results from a pilot program in Rialto, California found that body cameras were associated with a decrease in use of force.29 In addition, footage from body-worn cameras provides new opportunities for police training and feedback.30 Moreover, to law enforcement authorities who fear a growing crisis of legitimacy in their communities, body-worn cameras offer transparency and a way to restore public trust in the police.31 The other side of the ledger, however, contains concerns about how body cameras will affect citizens’ privacy; how the footage will be stored and maintained; and under what conditions the public will have access to the evidence.32

This Note does not attempt to provide an all-things-considered recommendation about whether body cameras amount to sound public policy. Rather, it assesses one salient argument that is frequently made in favor of body cameras: that they will reduce societal conflict and polarization by offering definitive proof of what happened.

Proponents of body cameras often argue that video footage can provide unambiguous records of police-civilian encounters.33 For instance, when the New Jersey legislature approved a bill requiring local police to be filmed by in-car or body-worn cameras, the sponsor of the bill touted video footage as providing “an unbiased, accurate record [of] what transpired.”34 Video recordings of police interrogations of suspects, for their part, have been hailed as “ready and available as an objective offer of proof.”35 As Philadelphia Police Commissioner Charles Ramsey explained, “Everybody’s got their version of a story, but when it’s on tape, it’s on tape. . . . It is what it is.”36

But the assumption that video evidence will help resolve disputes over whether misconduct occurred will bear out only if fact finders reviewing the footage can agree on what it shows.37 If they cannot agree—if, for example, they conform their perceptions of facts to their expectations or preferred outcomes—then cameras may fail to deliver on the promise of definitively resolving polarizing disputes.

This Note examines the potential for fact finders to evaluate even hard video evidence in biased ways while simultaneously becoming more confident that their judgments are unbiased. It argues that, in at least some cases, psychological factors can conspire to produce biased factual findings, even among viewers who are sincerely trying to evaluate the evidence fairly and impartially. In particular, it finds that video evidence remains susceptible to significant viewer bias and simultaneously causes some fact finders—namely those who feel a strong affinity with police officers—to become more certain of their judgments and more resistant to persuasion by others who disagree. It concludes that while presenting fact finders with video footage probably does not exacerbate biased decisionmaking, we lack evidence that it constitutes an improvement over the status quo.

This Note focuses on how video evidence interacts with cognitive processes that lead well-intentioned people to form opposing views of the same situation. It does not analyze cases where video footage exposes blatant corruption or dishonesty. For instance, it does not examine cases like Walter Scott’s shooting by Michael Slager. Slager initially claimed that he discharged his weapon because Scott had taken his Taser and he felt threatened.38 But video footage captured by a pedestrian shows that at the time Slager fired, Scott was fifteen to twenty feet away and fleeing.39 And the video shows Slager picking up an object from the ground and dropping it by Scott’s body,40 leading some viewers to conclude that Slager was planting the Taser on Scott to make his account more believable.41 Commentators noted that this case “deviated greatly” from the familiar “template for controversial police shootings,” because once the video surfaced, “there was hardly the typical closing of ranks around Slager” by other police and he was quickly charged with murder.42

This Note does not examine such cases, where one party has outright lied about what happened, and video footage debunks the fabricated version of events. Rather, it looks at a far more familiar and ordinary phenomenon: different people forming contradictory interpretations of the same event.

Part I sketches a psychological account of why high-profile police shootings tend to divide the public along partisan lines, drawing on literature from cognitive and social psychology. Part II reports the results of a study conducted with over thirteen hundred participants across fifteen cities, finding that one’s prior views about the police can affect how one perceives crucial facts captured on videotape. This holds true for factual or objective judgments, such as whether a weapon was present and whether physical force was used; evaluative judgments, such as whether the police officer treated the citizen fairly; and global judgments, such as whether the police officer deserves to be sanctioned for misconduct. Part III describes the results of a follow-up study that assigned mock jurors to receive different types of evidence documenting an incident in which a police officer used force against a citizen. This study compared video evidence to nonvideo testimony in terms of how strongly mock jurors’ prior attitudes toward police colored their judgments of the incident. The results indicate that video evidence does not conclusively outperform nonvideo testimony in minimizing bias. That is, we lack evidence that fact finders reviewing video evidence rely less on their prior beliefs to decide crucial factual issues than do fact finders reviewing other types of testimony. At the same time, it appears that watching video footage leads some fact finders to become more confident in their verdicts. Part IV discusses the implications of these findings for the policy debate over body-worn cameras, arguing that we should be more skeptical of the commonsense view that body cameras can tell us unambiguously and definitively what happened in any given confrontation involving the police.

I. the psychological roots of polarization

The Sixth Amendment of the United States Constitution guarantees the accused a “right to a . . . trial, by an impartial jury.”43 The impartiality of fact finders is crucial to the proper functioning of law. Fact finders such as grand jurors are not supposed to arrive at their verdicts based on their prior views about police in general; they are supposed to ground their decisions in the evidence presented to them about the specific officers involved in the particular incident at issue. As Dan Kahan and colleagues have argued, “[L]egal decisionmakers must be neutral . . . . [T]he legitimacy of the law would . . . be compromised if legal decisionmakers, as a result of motivated cognition, unwittingly formed perceptions of facts that promoted the interests and values of groups with whom they had an affinity.”44

Yet evidence abounds that ordinary people are not “impartial, data-driven processors” and that they fall short of this standard in myriad ways.45 First, ordinary people are partisans who filter incoming information through implicit frameworks and hypotheses that they bring to the situation. A wealth of psychological research establishes that even if jurors earnestly try to evaluate evidence fairly and objectively, their background assumptions, motivations, and ideologies tend to color their perceptions of evidence.46 For example, in a classic study of biased perception, researchers asked football fans from rival schools to watch footage from a game between the two schools’ teams.47 They found that fans of each team believed the video demonstrated that the referees had unfairly favored the opposing team.48 The researchers concluded that perceptions of an event can be influenced by motivational factors such as team loyalty, even when the event is captured on video.49 More recently, Kahan and colleagues demonstrated that video footage depicting political protests is similarly susceptible to biased factual findings: the degree to which viewers perceived the protests to be obstructing, intimidating, and threatening was influenced by the putative message promoted by the demonstrators and viewers’ own cultural worldviews.50

Second, decisionmakers tend to maintain an “illusion of objectivity”51: they believe they perceive reality as it really is, and that no reasonable person could draw a contrary conclusion.52 One prominent example of the illusion of objectivity comes from the Supreme Court case Scott v. Harris,53 in which an eight-Justice majority found that that a police officer had not violated the Fourth Amendment when he used deadly force against a driver who led a high-speed car chase. The Court wrote that video footage of the chase established that “no reasonable jury” could come to the factual conclusion that the driver did not pose a deadly risk to the public, and rejected the notion that there was a genuine issue of fact.54 Kahan and colleagues, however, showed the video clip to over a thousand members of the public and found, contrary to the Court’s claim that “the videotape . . . speak[s] for itself,”55 that a substantial minority of respondents disagreed with the Court’s factual assertions.56

Third, the illusion of objectivity is supported by an additional psychological phenomenon called the “bias blind spot”: people believe that they are less susceptible to bias than others, even when they are educated about the pervasiveness of psychological biases.57 Researchers have also found that people consider their own views to be driven more by objective or rational factors and less by political ideology than the views of their adversaries.58 This “naive realism” leads people to conclude that others who disagree with them are unreasonable, incompetent, misinformed, or prejudiced.59 The result, researchers believe, is heightened hostility and distrust between partisan groups.60

Finally, and perhaps most troubling, psychological research on polarization suggests that certain types of evidence may actually exacerbate societal discord. In a classic study of biased assimilation of evidence, Charles Lord and colleagues famously demonstrated that partisans’ attitudes toward capital punishment diverged more widely following exposure to a mix of positive and negative scientific evidence about the deterrent effect of the death penalty.61 Even though participants read both information that challenged their preexisting beliefs and information that aligned with their preexisting beliefs, they selectively credited the confirmatory science. This finding represents an extreme instance of a broader psychological phenomenon called “belief perseverance”: the tendency for strongly held beliefs to endure unabated, even after contradictory evidence is introduced.62 The participants in Lord and colleagues’ study not only clung to their initial beliefs in the face of disconfirming evidence, but they also doubled down on them. The authors concluded that proffering objective evidence “will frequently fuel rather than calm the fires of debate” over contentious social issues.63

It appears, then, that exposure to certain types of evidence, such as scientific articles, may do nothing to reduce bias and instead leads people to become more certain that they are right—because science backs them up. Video footage raises the specter of the same problem. What if seeing audiovisual proof leads people to feel more certain that their interpretation is the objectively correct one? If this is true, videos could increase polarization rather than reduce it. Like scientific evidence, video footage may prove to be fertile ground for biased assimilation, polarized attitudes, bias blind spots, and naive realism.

II. study one: do prior attitudes toward police influence how fact finders perceive video evidence?

Do our general attitudes toward the police affect how we react to videos depicting police conduct? Do they influence our judgments of objective or factual matters such as who hit whom? Do they influence our opinions about more interpretive or evaluative matters, like whether the police officer made decisions fairly?64 Finally, do they influence our global judgments of whether the officers involved ultimately deserve to be punished? The answer, Study One suggests, is yes: fact finders’ prior attitudes toward police significantly influence all three types of judgments. Even when mock jurors all see the same video, their judgments about legally consequential matters show considerable polarization based on their prior attitudes toward police.

Study One examines viewer attitudes using a dataset collected from a nationally representative sample spanning fifteen American cities. Researchers at Yale University, New York University, and the American Bar Foundation conducted the original study in conjunction with the survey firm Knowledge Networks (KN).65 I obtained permission from the principal investigators to use the dataset to investigate a distinct research question: how fact finders’ general feelings toward police affect their judgments of videos depicting interactions between citizens and the police. The study participants were randomly drawn from a demographically diverse panel of forty thousand U.S. citizens maintained by KN.66 A total of 1,361 completed interviews were obtained. Fifty-one percent of the sample was female (n = 688), and the median age was 49 years (SD = 15.28). A majority of participants (72%) were Caucasian, 10% were African American, 11% were Hispanic, and 7% indicated other or multiple racial categories.

A. Method

Participants answered a series of questions about their prior experiences with local law enforcement and their opinions about the criminal justice system. In the key items of interest for this study, participants reported how strongly they identify with the police. This measure has been shown to be an important predictor of favorable dispositions toward law enforcement.67

The study asked participants to rate their level of agreement with seven statements68 presented in random order: (1) “If you talked to most of the police officers who work in your neighborhood, you’d find that they have similar views to you on many issues”; (2) “Your background is similar to that of many of the police officers who work in your neighborhood”; (3) “You can usually understand why the police who work in your neighborhood are acting as they are in a particular situation”; (4) “You generally like the police officers who work in your neighborhood”; (5) “If most of the police officers who work in your neighborhood knew you, they would respect your values”; (6) “Most of the police officers who work in your neighborhood would value what you contribute to your neighborhood”; and (7) “Most of the police officers who work in your neighborhood would approve of how you live your life.” These items were averaged into a single measure of identification with the police.69

Participants also viewed three video clips, drawn from real police footage, which each showed police officers seeking to arrest citizens.70 The three clips, which were presented to participants in random order, depicted interactions ranging in intensity.71 In Video 1 (resistant citizen), a Caucasian citizen acted rudely toward a Caucasian police officer and refused to comply with the officer’s directions.72 The police officer responded by applying force to restrain the citizen. In Video 2 (aggressive citizen), an African American citizen initiated aggressive contact against two Caucasian police officers, who tried to subdue him. The police officers responded by using force, and the clip ends with the officers trying to handcuff the citizen. In Video 3 (confused citizen), the level of force and conflict was low on both sides. A Caucasian police officer questioned a shirtless African American citizen, who was polite but disoriented and confused. The officer eventually led the acquiescent and compliant citizen away in handcuffs.73

After viewing each video, respondents answered a set of questions about what happened in the video, whether the officer treated the citizen fairly, and whether the police officer acted inappropriately. Participants were thus asked to make three different types of judgments: factual judgments, fairness judgments, and global culpability judgments.

First, participants rated their level of agreement with factual statements regarding what happened in the video, such as: (1) “The officer displayed or used a weapon”; (2) “The officer threatened to use physical force”; and (3) “The citizen complied with the police officer’s requests.” These items were reverse scored74 and combined into a single measure of objective, factual judgments favoring the police.75

The study also asked participants to rate the officer’s treatment of the citizen. A robust body of psychological research demonstrates that people care deeply about how authority figures treat them, and that judgments of “procedural justice” greatly influence perceptions of legitimacy.76 According to procedural justice theory, police officers who give citizens a chance to voice their views, who are courteous and respectful, and who make decisions fairly and give explanations for their actions are seen as wielding more legitimate authority.77 Participants were thus asked to rate their agreement with statements regarding the officer’s treatment of the citizen in the video, including: (1) “The officer allowed the citizen to express his views before making decisions”; (2) “The officer treated the citizen with respect and dignity”; and (3) “The officer made decisions about what to do in fair ways.” These statements were averaged into a single measure of fairness judgments and scored such that higher scores indicated judgments more favorable to police.78

Finally, the study asked participants to make global judgments about the overall reasonableness of the officer’s conduct. They rated their level of agreement with statements such as: (1) “The officer should be reprimanded or punished in some way” (reverse scored); (2) “It would be appropriate for the citizen to sue the police” (reverse scored); and (3) “The officer behaved appropriately toward the citizen.” These items were averaged into a single measure of global judgments favoring police.79 Appendix I (Table 1) reports the full list of outcome items and the three composite measures they constitute.

B. Study Results

I hypothesized that prior identification with police would be a significant predictor of judgments of what happened in the videos along three key dimensions: factual judgments, fairness judgments, and global judgments.80 To test these predictions, I conducted nine linear regressions predicting each of those three outcome variables for each of the three videos.81

For factual judgments, across all three videos, respondents who strongly identified with police were more likely to find facts favoring the police.82 In other words, high identifiers were less likely to agree that the officer displayed or used weapons, that the officer used insulting language, or that the citizen complied with the officer’s requests. A similar pattern was observed for fairness judgments, the second outcome variable. Across all three videos, high identifiers thought the police officer acted more fairly and respectfully than did low identifiers.83 On the third outcome variable—global judgments—the same pattern was again observed. High identifiers were more likely than low identifiers to find that the police officer acted appropriately and lawfully.84 They were less likely to believe that the police officer deserved punishment.

C. Discussion and Implications

When Eric Garner’s chokehold death was captured on camera, many wondered how a Staten Island grand jury could have failed to indict the police officer involved. Esau Garner, Eric Garner’s widow, told reporters, “There is no doubt in my mind or in the mind of all the people out there in the world that what we saw in that video cannot be disputed. How they disputed it, I don’t know.”85

The results of Study One indicate that the dispute over the Garner video is not a fluke. On the contrary: when over a thousand Americans, representative of the nation in age, race, ethnicity, gender, and education, watch video footage showing police officers arresting citizens, they bring their prior attitudes toward the police to bear on their judgments of the evidence. Crucial judgments, such as whether the officer used force, made decisions in a fair way, or ultimately violated the law, are significantly influenced by viewers’ prior level of identification with police—even though the events are captured on tape. This was true across several different videos depicting a range of interactions that varied in their intensity and use of force. Video evidence appears susceptible to considerable viewer bias.

These results are especially striking because the pattern was observed even among factual judgments involving minimal room for interpretation, such as whether the officer displayed a weapon. In addition, and less unexpectedly, prior identification shaped more evaluative judgments about how respectfully the officer acted and whether he should be punished for his conduct.

The results of Study One are bolstered by recent research using other stimuli, including videos that have been doctored to remove the audio and videos depicting altercations between novel groups (such as members of Green and Blue teams, denoted by t-shirt colors).86 Yael Granot and colleagues used eye-tracking technology to examine the effect of group identification on perceptions of video evidence among university students and community members.87 Their results support the findings reported here and add an important moderating variable: where viewers look when they watch the videos. Granot and colleagues found that viewers who spontaneously tended to focus on the police officer in the footage showed the same polarization effect described in Study One.88 That is, these participants’ judgments of the officer diverged according to the strength of their identification with police. But among viewers who fixated less frequently on the police officer, no such polarization effect was observed. Among these participants, judgments of the officer did not differ depending on whether the viewer identified strongly or weakly with police. This same pattern was replicated when the researchers instructed viewers on where to direct their attention as they watched the video.89 Viewers told to watch the officer displayed divergence based on identification, whereas those who watched the citizen showed no such polarization.

Taken together, the findings from these studies indicate that, in many cases, viewers rely significantly on their prior beliefs to determine what happened in the videos and what the legal consequences for the parties should be. Insofar as the Body Worn Camera Partnership Program and similar reforms seek to reduce wrenching societal discord over high-profile cases, these results raise doubts about whether body cameras will succeed in achieving these aims. In fact, these findings suggest that even if a grand jury is given video footage of an incident, the jurors may nonetheless come to divergent conclusions about the justifiability of the officer’s actions, based on their preexisting views toward police. For fact finders who doubt that the police share their views or values, it’s easier to arrive at the opinion that the officer violated the law. By contrast, for fact finders who feel that police officers, in general, share their background and values, it is harder to see misconduct in the footage.90

III. study two: is video evidence less susceptible than nonvideo evidence to the influence of prior attitudes?

The results of Study One establish that even video evidence is susceptible to biased interpretation based on prior views toward police, contrary to the commonsense belief that videos provide unambiguous proof. But to assess whether body cameras might nonetheless constitute an improvement over the status quo, we must ask how video evidence compares to the available alternatives. Study Two thus investigates whether video footage reduces the influence of prior identification relative to other types of nonvideo testimony describing the same event. It also examines whether reviewing video evidence increases fact finders’ confidence in their verdicts, relative to reviewing other types of evidence.

In the summer of 2009, a forty-two-year-old Caucasian motorist named Donel Adam Stogner was choked to death by Deputy Chris Sturdivant of the Livingston, Louisiana police department.91 Following the incident, the Sheriff’s Office released a seven-minute dashboard camera video of the encounter, arguing that the footage showed that the deputy had followed procedure. In the video, Stogner can be viewed hiding something in his clenched fist while the deputy tries to arrest him. As the two men scuffle, Stogner tries to swallow the object. The deputy hits Stogner in the back of the head with his fist and forearm, wraps his arm around Stogner’s neck, and wrestles him to the ground. The two men engage in a struggle for several minutes.92

Study Two uses the Stogner case to address the following question: does the effect of prior identification change depending on whether people learn about an event through watching a video versus reading a written description? Participants were assigned to one of four conditions93: (1) video footage; (2) dueling accounts; (3) neutral perspective; and (4) single police account. The stimuli in all four conditions were based on the events depicted in the Stogner video, allowing us to compare four different ways of learning of the same event. The four conditions are described below.

A. Method

1. Stimuli for Four Conditions

a. Video Footage

Participants in the video-footage condition were shown a forty-nine-second segment of the dashboard camera video.94 The clip depicts the initial stop, the heated dispute between the two men, and the first twenty seconds of the physical struggle. Participants in this condition see the deputy slam Stogner’s head against the car and put him in a chokehold, but the clip ends while Stogner is still alive and struggling against the arrest. No participants were told that the citizen ultimately dies from the encounter.

 b. Dueling Accounts

In the dueling-accounts condition, participants were given two first-person written descriptions of the events captured in the forty-nine-second video clip of the Stogner incident, one from the perspective of the citizen and one from the perspective of the police officer. Participants in this condition did not see the video that inspired the written accounts, nor were they told that such footage exists.

Testimony from the citizen:

The cop pulled me over and asked for my license. Even though I had no idea why he pulled me over in the middle of the night by the side of the road, I took out my wallet and showed him my ID. Then he started to accuse me of having something in my hand. I said I didn’t have anything in my hand, I swear to you. Then he attacked me out of nowhere. He twisted my arm behind my back and shoved me up against his patrol car. I was twisting and screaming in pain. I was completely terrified. Then as my chest was pressed up against the hood of his car, face down, he punched me in the back of the head, slamming my head full force against the hood of the car. I almost lost consciousness. He slammed me so hard I thought he had cracked my skull. Then even though I was completely disoriented at that point and completely powerless to resist, he put me in a chokehold. He hooked his left elbow around my neck and pulled up, cutting off my blood flow even more. I thought he was trying to strangle me. He wrestled both arms around my neck and jumped up, choking and choking me.

Testimony from the officer:

I pulled the suspect over and asked him for ID. He pulled out his wallet to show me his license. That’s when I saw his hand, his right hand, was clenched in a fist. I asked him what he had in his hand. He said “nothing,” but I could see that he was definitely hiding something. I instructed him to open his hand. He refused, saying “I swear to you I have nothing.” But I knew he did, because it was clenched in a fist and he was acting suspicious. That’s when I decided to arrest him. I wrestled his right arm behind his body and pushed him up against the hood of my car so that I could give him a pat down to check for weapons. I again instructed him to open his hand. He again refused to do so. He jerked his hand out of my grasp and transferred the object to his left hand. It looked like he put it in his mouth, like he was trying to swallow it so I wouldn’t see it. I used my left elbow to strike the back of his neck, trying to knock the object loose from his mouth. I tried to wrestle him to the ground by putting both of my arms around his neck and chest area. He was a big guy, so I needed to jump up and try to pull him down.

The purpose of the dueling-accounts condition was to see how participants responded when presented with two versions of the same event, each generated by a self-serving party. This research design was meant to simulate a trial in which both sides tell their story and the jury, having heard both accounts, decides what happened.95

c. Neutral Perspective

As a foil for video footage, dueling accounts represent only one potential comparator. This study also asks how video evidence compares to a single neutral account of the event. For instance, if an uninvolved passerby had been present during the incident, she might be called to testify about what she observed. Thus, a third condition exposed participants to neutral testimony written from the perspective of a disinterested third-party observer.

To generate the stimuli for this condition, I asked a group of four naive participants, who were blind to the research purpose and hypotheses, to watch the video clip and write a description of what happened.96 Of the four essays they generated, two were too long to be suitable for an online survey interface. That left two essays that were comparable in length to the narratives in the dueling-accounts condition. These two narratives became the stimuli for the neutral-perspective condition. Participants in this condition were randomly assigned to read one of these two narratives,97 which were presented as “eyewitness testimony.”

Eyewitness Account A:

It’s dark outside, and a lone officer approaches a suspect. The suspect is outside of the vehicle, but does not seem to be hostile towards the officer. They have ashort talk, as the suspect is showing the officer something in his hand. As the suspect is handling something in his hands, the officer says something, but what he says is too low to understand. As the suspect lowers his hand, the officer asks him what he has in his hand. The suspect’s right hand is in a closed fist. The suspect claims to have nothing in his hands, but the officer doesn’t believe him, and asks him repeatedly to show what is in his hand. As the officer forcibly grabs the suspect’s hand to open it, the suspect still claims to have nothing. The officer then proceeds to bend the man’s arm behind his back, in an effort to open his hand. Slamming the suspect against the car, the officer yells again for the man to open his hands. The man pretends to comply as the officer lets go of his arm, the suspect shoves whatever it is into his mouth. As this happens, the officer hits the man on the back of the head so hard it knocks the man’s hat off. He then puts the suspect into a choke hold.98

Eyewitness Account B:

The police officer met the driver of the truck at the back of his truck between it and the officer’s car. One of the men said, “How are you doing?” The man approaches the officer closely. The man is doing something with an object in his hands. The officer reaches toward the man’s belt around the right hip, asking the man to keep his hand up. The man responds with something indecipherable. The cop, holding a flashlight, says, “give me your hand.” The man says, “I promise you.” The cop moves to his left close to the man and takes the man by the wrist. The cop begins looking for something. The man says “I don’t have nothing in my hand, I swear.” The cop calls in a code on his radio. The cop pulls the man’s arm behind him. The man says, “hey look, what are you doing” as the cop handles the man. The cop drives him forward, against the front of the police car. The cop tells the man “open your hand.” The man says “alright, hey.” The cop tells him again to open his hand. The man is saying “no no no, please listen to me” then something indecipherable. The cop again says again for him to open his hand and begins to struggle with the man. The cop continues to struggle with the man and hit his arm across the back of the man’s head. The cop then jumps on the man, putting his arms around or across the man’s throat.

The purpose of the neutral-perspective condition was to simulate how a bystander might testify about witnessing the Stogner incident. Readers will notice that these two accounts differ from one another in numerous ways. These differences should not concern us, however, because we are interested not in comparing the two accounts with one another, but in comparing them with video evidence. Nonetheless, I analyzed the two accounts separately to confirm that they did not differ in how participants judged them or in how strongly identification influenced judgments of each account. The results indicate that participants gave similar ratings to both eyewitness accounts, and that participants’ reliance on their prior attitudes toward police largely did not differ between the two accounts.99 I thus analyzed the two accounts together, collapsing them into a single category called the “neutral perspective.” For readers interested in knowing how each eyewitness account performed separately, Appendix II reports the disaggregated data.

d. Single Police Account

Finally, participants in the single police account read only the paragraph describing the police officer’s testimony from the dueling-accounts condition. The purpose of this fourth condition was to explore how fact finders respond when one side is unavailable to tell his version of the events. In the real Stogner case, after all, Stogner did not live to recount what happened from his perspective. If the event had not been captured on video, a jury would have had access only to the deputy’s telling. Thus this fourth condition was included to simulate the imbalance in testimony that a jury often hears when a police-citizen encounter turns deadly. When fact finders are given only the police officer’s version of the story, does it change the degree to which they rely on prior identification in forming their judgments?

2. Study Procedure

Participants were recruited from Amazon’s Mechanical Turk, an online labor pool.100 A total of 246 U.S. adults completed the study, reflecting a 96% completion rate. About half (n = 121) were women. The median age was thirty-two years (SD = 12.07). Eighty-one percent of participants were Caucasian, six percent were African American, and seven percent were Hispanic.

Participants first answered questions about their level of identification with the police.101These seven survey items were averaged into a single measure of identification with the police.102 Next, participants were exposed to one of the four stimuli: the video clip itself, the dueling accounts based on the clip, one of the neutral descriptions based on the clip, or only the police officer’s side of the story.

Following presentation of the stimuli, participants answered an expanded series of questions about what they observed, presented in random order. Case law establishes that a number of factors are relevant to a jury’s determination of whether a police officer’s use of force is reasonable, including the severity of the citizen’s crime, the extent of the citizen’s injury, and whether the plaintiff was actively resisting arrest.103 Thus a handful of new questions were added to the set from Study One to assess how participants would respond to questions a real jury might be asked.104 For instance, participants were asked to report how strongly they believed Stogner was being arrested for a severe crime, and whether they thought he resisted arrest.

Overall, five different types of judgments were measured: factual judgments, subjective judgments, fairness judgments, global judgments, and certitude. Participants were asked to rate their level of agreement with several objective, factual statements such as, “The officer hit the citizen,” and “The citizen tried to hide something from the officer.” These items were averaged into a single composite measure of factual judgments favoring the police officer.105

Participants were also asked to rate their level of agreement with a set of subjective or inferential statements regarding particular elements of the altercation, such as “There is a high likelihood that the citizen was armed at the time” and “The officer attempted to defuse the situation.”106 These items were inspired by case law specifying factors that determine whether use of force in particular instances was reasonable. These subjective items were averaged into a single composite measure of subjective impressions favoring the police officer.107

Participants were additionally asked to rate their agreement with statements regarding police use of procedural justice, such as “The officer made decisions about what to do in fair ways.” These items, which were used in Study One, were averaged into a composite measure of fairness judgments favoring police.108

Finally, participants were asked to rate their level of agreement with several global statements regarding the police officer’s overall culpability. These items were averaged into a single composite measure of global judgments favoring the police officer.109 The full list of questions is reproduced in Appendix I (Table 2).

In addition to the objective, subjective, fairness, and global judgments, I asked participants how confident they were that the police officer had acted reasonably or unreasonably, and how difficult it would be to persuade them of the opposite conclusion. These two new measures—confidence and openness to persuasion—were averaged into a single measure of certitude.110 As described earlier, researchers believe that the illusion of objectivity leads people to become overly confident that their perceptions of reality are objectively correct and to conclude that anyone who disagrees must be incompetent or unreasonable.111 This sense of infallibility exacerbates polarization and heightens distrust between groups.112 Moreover, certitude is a legally consequential psychological posture. When judges determine that no genuine dispute as to any material fact exists, they may decide to grant summary judgment, essentially taking the decision away from a jury. Thus these questions measuring certitude were included to examine whether various kinds of evidence are differentially likely to prompt the illusion of objectivity.

B. Study Results

Does the influence of identification differ depending on whether the stimulus is a video recording of the event versus another type of testimony? To answer this question, I conducted four linear regressions predicting factual judgments favoring police, subjective judgments favoring police, fairness judgments favoring police, and global judgments favoring police. In each model, I included identification with police, condition (dummy-coded),113 and the interaction between identification and each dummy-coded condition as predictors.114 As we will see in the next section, the four different types of evidence all gave rise to judgments that are significantly—and similarly—biased by prior attitudes toward police.

1. Factual Judgments

First, does the type of stimulus affect how strongly identification with police colors objective, factual judgments such as, “The citizen hit the officer,” and “The officer hit the citizen”?

Figure 1.

the effect of identification on factual judgments

Figure 1 depicts the effect of identification on factual judgments favoring police, modeled at one standard deviation (SD) above and one SD below the mean identification score, when participants are given video evidence, dueling accounts, a neutral perspective, or a single police account.115 The overall interaction between identification and condition was not statistically significant. Indeed, identification remained a significant predictor of factual judgments in all four conditions.

Figure 1 displays how someone high in identification (one SD above the average) would differ from someone low in identification (one SD below the average), in each of the four conditions.116 As Figure 1 illustrates, in all four conditions, higher identification with the police is significantly associated with objective judgments favoring the police.117 That is, across all four stimulus types, high identifiers were significantly more likely than low identifiers to agree with factual statements such as, “The citizen tried to hide something from the officer” or “The officer hit the citizen” (reverse scored).

The linear regression indicates that the overall interaction between identification and stimulus type did not quite meet the threshold for statistical significance, although it is close.118 This means that the four ways of presenting the evidence gave rise to similarly strong relationships between identification and factual judgments about what the citizen and officer did. In other words, video evidence was not significantly more effective than testimony from various perspectives at counteracting the influence of mock jurors’ bias in favor or against police officers—at least for objective, factual judgments.

While we lack evidence of a systematic relationship between the type of stimulus and the strength of participants’ reliance on their prior attitudes toward police, we can nonetheless drill down further to compare individual stimulus types to one another. For instance, does video evidence reduce the effect of identification when you compare it to just the dueling-accounts condition? What about when you compare it to just the neutral-perspective condition? The results reveal that video evidence does not differ significantly from the dueling-accounts condition or the neutral-perspective condition when we examine these pairwise comparisons.119

However, the comparison between the video-footage condition and the single police account condition revealed a significant difference: the effect of identification was stronger in the single police account condition.120 People who heard only the police officer’s version of events relied more heavily on their prior identification than did people who watched the video clip of the incident.

Overall, these results paint a somewhat murky picture. The interaction between prior identification and condition did not quite meet the threshold for statistical significance. This makes it difficult to conclude that the four conditions perform differently from one another in terms of how strongly identification predicts objective judgments. Only one of the more specific, pairwise comparisons showed a difference: fact finders given video footage rely less on their prior attitudes toward police when the comparator is the single police account. This finding indicates that video evidence is only a significant improvement when the alternative is a one-sided narrative.

2. Subjective Judgments

We next turn to the question of whether the type of stimulus affects the influence that identification has on subjective or inferential judgments—agreement with statements such as, “The citizen was being arrested for a severe crime,” and “There was a high likelihood that the citizen was armed at the time.”

Figure 2.

the effect of identification on subjective judgments

Figure 2 depicts the effect of identification on subjective judgments favoring police, modeled at one SD above and one SD below the mean identification score, by condition. The effect of identification did not differ significantly across the four conditions, and identification remained a significant predictor of subjective judgments in all four conditions.

Figure 2 shows the comparison between the four stimulus types in terms of how strongly prior identification affects subjective judgments. In all four conditions, participants’ level of identification with police maintains a significant positive association with subjective interpretations favorable to police.121 For instance, high identifiers were more likely than low identifiers to agree with statements such as, “The citizen posed an immediate threat to the safety of the officer or the safety of others” and “The officer tried to limit the amount of force he used.”

The regression predicting subjective judgments reveals that the overall interaction between identification and condition is not significant.122 The different ways of presenting the evidence made no difference in the relationship between identification and judgments about subjective matters such as whether the police officer tried to defuse the situation or limit the amount of force he used.

As before, even though we found no evidence that evidence type makes a significant difference overall, I conducted individual comparisons between the video-footage condition and the other three conditions. These comparisons again reveal that video evidence did not reduce the effect of prior identification when compared to the dueling accounts123 or neutral perspective124 conditions. Video evidence does reduce the effect of prior identification when compared to the single police account condition, but only to a marginally significant degree.125 We thus lack strong evidence that fact finders given video testimony relied less on their prior attitudes toward police when deciding subjective and inferential factors such as whether the citizen posed a threat or was likely armed at the time of the incident. These judgments are important factors in determining whether an officer’s use of force was reasonable.126 These study results thus indicate that consequential matters will remain susceptible to biased interpretation, even when jurors have access to hard proof in the form of video footage.

3. Fairness Judgments

We next turn to the question of whether the type of stimulus affects the influence identification has on perceptions that the police acted fairly—agreement with statements such as, “The officer treated the citizen with respect and dignity” and “The officer allowed the citizen to express his views before making decisions.”

Figure 3.

the effect of identification on fairness judgments

Figure 3 depicts the effect of identification on fairness judgments, modeled at one SD above and one SD below the mean identification score, by condition. The effect of identification did not differ significantly across the four conditions, and identification remained a significant predictor of fairness judgments in all four conditions.

Figure 3 shows the comparison between the four stimulus types in terms of how prior identification affects perceptions of procedural justice—that is, the sense that the police acted fairly. In all four conditions, participants’ level of identification with police maintains a significant positive association with procedural justice judgments favoring the police.127 For instance, high identifiers are more likely than low identifiers to agree with statements such as, “The officer made decisions about what to do in fair ways.”

The regression predicting fairness judgments reveals that the overall interaction between identification and condition is not significant.128 The different ways of presenting the evidence made no difference in the relationship between identification and judgments about whether the officer acted fairly and respectfully.

As before, I conducted individual comparisons between the video-footage condition and the other three conditions. These comparisons showed that video evidence did not reduce the effect of prior identification when compared to the dueling-accounts condition,129 the neutral-perspective condition,130 or the single police account condition.131 In sum, we lack evidence that people reviewing video footage relied less on their prior attitudes toward police when deciding whether the officer treated the citizen fairly.

4. Global Judgments

Finally, we turn to the question of whether the type of stimulus affects the influence of identification on global judgments of police wrongdoing—agreement with statements such as, “The officer’s use of force was reasonable here” and “The officer should be reprimanded or punished in some way” (reverse scored).

Figure 4.

the effect of identification on global judgments

Figure 4 depicts the effect of identification on global judgments, modeled at one SD above and one SD below the mean identification score, by condition. The overall interaction between identification and condition was not statistically significant. Identification remained a significant predictor of global judgments in all four conditions.

Figure 4 depicts the comparison between the four stimulus types in terms of how strongly participants rely on their prior attitudes when making global judgments about the officer’s culpability. Prior identification has a significant effect in all four conditions132: low identifiers were more likely to agree with statements such as, “The officer’s use of force was unnecessary here” “The officer’s use of force was excessively violent here” and “The officer’s use of force was reasonable here” (reverse scored).

The regression results reveal that the overall interaction between identification and stimulus type is not statistically significant.133 The four methods of presenting evidence give rise to similarly strong relationships between identification and global judgments favoring the police officer. That is, when fact finders are deciding global matters such as whether an officer deserves punishment, they rely on their prior attitudes toward police, and the amount to which they draw on these prior attitudes is about the same regardless of whether they reviewed video evidence or other types of evidence.

The pairwise comparisons revealed that the video-footage condition significantly reduced the influence of prior identification relative to the neutral-perspective condition134 and marginally reduced135 influence of prior identification relative to the single police account condition.136 But the video-footage condition does not differ from the dueling-accounts condition,137 and the dueling-accounts condition in turn does not differ from the other two conditions,138 suggesting that the overall picture is inconclusive. We ultimately lack evidence that fact finders who are given different types of testimony rely any more or less on their prior attitudes toward police when making global judgments such as whether the police officer acted appropriately.

5. Certitude

A key question throughout has been how different types of evidence affect fact finders’ confidence in their interpretations. As described earlier, psychological phenomena such as the illusion of objectivity, the bias blind spot, and naïve realism generally lead people to be unduly confident that their judgments reflect reality, while other people’s judgments are hopelessly biased. We thus turn to the question of whether different types of evidence prompt different levels of certainty among fact finders reporting high and low identification with police.

Figure 5.

the effect of identification on certitude in final verdict

Figure 5 depicts the effect of identification on certitude in final verdict, modeled at one SD above and one SD below the mean identification score, by condition. The interaction between identification and certitude was statistically significant. Participants who identified strongly with police (+1 SD) expressed greater certitude when they reviewed video evidence than when they reviewed other types of evidence. By contrast, participants low in identification (-1 SD) showed no such increase in certitude.

As Figure 5 shows, some fact finders—namely, those who identify strongly with police—became more certain in their ultimate judgments when they reviewed video evidence than when they reviewed other types of evidence. That is, they became more certain that the police officer had acted reasonably or unreasonably, and they expressed less openness to the possibility of being persuaded to change their minds.139 Interestingly, this phenomenon was not observed among people low in identification with police: among this group, fact finders expressed no more certitude when they watched the video than when they received nonvideo testimony.140

C. Discussion and Implications

It appears that video evidence is not completely successful in eliminating the influence of identification with the police; our prior attitudes color our interpretation of events. Identification with police is a statistically significant predictor of objective, factual judgments, subjective judgments, fairness judgments, and global judgments across video and nonvideo evidence. Thus, even when we all watch the same footage, we come to different conclusions about important matters such as what happened, who started it, and what the legal consequences should be. These results are consistent with the findings of Study One. They indicate that proponents of body cameras may be unduly optimistic that video footage can decisively resolve ambiguous cases.

Videos, then, are susceptible to biased interpretation, despite their reputation for providing irrefutable, objective proof. But are they better than the available alternatives? This study examined three alternatives: (1) dueling accounts from the perspectives of the two adverse parties involved, the police officer and the citizen; (2) neutral perspectives written from the point of view of disinterested third parties; and (3) a single police account, meant to simulate the situation in which the citizen has died in the encounter and only the police officer is left to recount what happened.

The fact that the overall interaction between identification and condition is not statistically significant for all four types of judgments suggests that the different ways of presenting testimony do not make a significant difference in minimizing fact finders’ reliance on their prior attitudes. We can probably say that video evidence is not worse than other types of testimony, but whether it is superior to the alternatives remains an open question. Arguably, the data reported here suggest that videos might be slightly better at reducing viewers’ reliance on prior identification when it comes to objective, factual judgments and when the alternative is a lopsided trial offering only the police officer’s account—but the effect sizes are not large enough to rule out the possibility that these findings are due to random chance. Ultimately, we lack strong evidence that videos improve upon the information that jurors already use.

Participants who saw the encounter with their own eyes were not significantly less likely to draw on their prior identification with police when making decisions—but they were more certain of their opinions if they had a pre-existing tendency to identify with the police. When we compare the responses of participants given video and nonvideo testimony, we find that those who saw the videos and already identified with the police were more likely to express certitude in their judgment that the officer had acted reasonably or unreasonably. This finding should give pause to advocates who hope that body cameras will make it easier to indict and convict police officers for excessive force. These results suggest that video evidence fails to reduce polarization significantly while simultaneously prompting fact finders who most strongly identify with police to become more unshakable in their judgments.141

IV. general discussion

A. Why Is Video Evidence Inconclusive?

How can it be that we can all see the same footage yet draw different conclusions about the actors involved based on how we feel, in general, toward the police? I do not claim that identification with police can literally and directly affect what we see.142 Indeed, we lack evidence that participants’ identification with police altered their basic sensory experience of vision in any way. What seems more likely is that participants’ identification with police, along with identification’s concomitant beliefs, desires, emotions, and motivational states, affected cognitive factors such as what they paid attention to; what aspects of the video they credited and discredited; and what details stood out in their memories as they answered questions about what they saw.143 These cognitive processes—attention, biased assimilation, and selective memory—could likely account for the Note’s findings without our needing to postulate that vision itself was altered.

The design of this study does not allow us to pinpoint the psychological mechanisms that account for participants’ polarized factual findings. But we can draw upon general psychological principles to sketch an account of how cognitive processes might have conspired to bias participants’ views of video evidence without their awareness.

Imagine a participant who doubts that the police share his views or values, and who therefore scores low on identification with police. When he watches the video footage, he might eye the police officer warily, watching with trepidation to see whether the officer is going to rough up the citizen. By focusing on the officer, he might miss some context that would have complicated the situation, such as aggressive movements by the citizen. Or he might see the citizen’s aggressive movements, but easily come up with a plausible justification for them (for example, “I would be scared too if I were in his situation, so I can understand why he would behave that way”). Simultaneously, he might react more critically to similarly aggressive behavior by the police officer (for example, “That was hasty; why didn’t he try harder to calm the man down before trying to arrest him?”). As Lord and colleagues note, one of the hallmarks of biased reasoning is taking evidence that disconfirms our preexisting beliefs and subjecting it to hypercritical scrutiny, while taking evidence that confirms our preexisting beliefs and readily accepting it at face value.144 Finally, the low-identifying participant might encode the video in his memory in a biased fashion. Consequently, when he later formulates answers to the questions about what he saw in the video, he might draw disproportionately upon the details that stand out most saliently in his memory—those that favor the citizen and disfavor the police officer.145 Of course, for a participant who identifies strongly with the police, one would expect that all of these cognitive processes would operate in the opposite direction. The end result would be considerable polarization within the group despite the fact that everyone saw the same footage.

B. Directions for Further Research

The above sketch is one hypothesized, speculative account of how cognitive processes might operate to produce polarized factual findings. Future empirical research is needed to identify the precise cognitive processes by which our level of identification with police affects our perceptions of video evidence. Relatedly, further research is needed to identify what interventions might interrupt the cognitive mechanisms responsible for polarized factual findings, and whether they could feasibly be integrated into the legal system.

For instance, if biased recall is to blame, perhaps asking fact finders to rewatch the video over and over as they answer questions would help counteract biases in memory. On the other hand, existing psychological research gives reason to worry that repeatedly watching a video may simply give viewers more chances to find ammunition to argue for their favored interpretation.146 This line of research finds that people tend to seek out evidence that would confirm their initial hypothesis and largely neglect to search for evidence that could disconfirm their hypothesis. Research on this well-documented phenomenon, called “confirmation bias,” suggests that letting people rewatch the stimulus would not serve to reduce polarization unless viewers were explicitly instructed to look for disconfirming evidence.

If, by contrast, polarization stems from an earlier point, such as from divergences in where viewers focus their attention when they first watch the stimulus, then we might settle on a different prescription. Perhaps we would want to direct viewers’ attention so that they all attend to the same details in the footage. On the other hand, Granot and colleagues found that attention divides: in their study, asking everyone to focus on the police officer led to greater polarization in decisions about whether to sanction the officer.147 This finding suggests that a simple instruction directing everyone to focus on the same thing will likely be insufficient to eliminate polarization.

As we contemplate this Note’s findings, it is important to acknowledge recent examples demonstrating that body-worn cameras can, in some instances, help secure indictments against police officers. For example, in July 2015, University of Cincinnati police officer Ray Tensing shot motorist Samuel DuBose in the head after pulling him over for driving without a front license plate.148 Tensing, who is Caucasian, claimed that he shot Dubose, who was African American, in self-defense after being dragged by DuBose’s car.149 Video footage captured by Tensing’s body camera, however, appeared to contradict this account, and a grand jury quickly indicted Tensing on charges for murder and voluntary manslaughter.150 The Hamilton County prosecutor told reporters that he probably would have believed Tensing’s version of events had it not been for the video footage.151 This case provides a clear example of how video footage can alter viewers’ initial conclusions.

So why was the Tensing video persuasive, at least at the indictment stage, while other videos were not? Additional research is needed to determine the boundary conditions of the findings reported here. Specifically, we need to know the situations in which video evidence performs no better than nonvideo evidence and the situations in which video evidence successfully serves to reduce polarization. The motivated reasoning literature suggests that, in general, stimuli that are more ambiguous provide more room for bias to infuse viewers’ interpretations.152 This literature would suggest that the more ambiguous the video, the more susceptible it would be to biased interpretation. This general principle is helpful, but we must keep in mind that “objective” evidence can be difficult to identify and define. Recall that many people viewed the footage documenting Eric Garner’s death as damning evidence against Officer Pantaleo, but the Staten Island grand jurors who viewed the same footage found no probable cause to bring criminal charges.153

It also bears mentioning that this Note did not examine the effect of combining video footage with other types of evidence. It is possible that reviewing video evidence alongside nonvideo evidence might affect the degree to which fact finders bring their biases to bear on their judgments, or the extent to which videos inspire certitude and resistance to persuasion among viewers. Future research should probe the effects of presenting fact finders with multiple sources of information, perhaps varying the order in which jurors are presented with first-person testimony, third-person testimony, and body camera footage.

Finally, this Note has focused on the cognitive roots of polarization—the psychological processes that lead well-intentioned people to form opposing views of the same situation. As I have argued, video evidence is not demonstrably less susceptible to these biases than is nonvideo testimony. But the Note has not addressed the effects of introducing video footage when the problem at issue is something noncognitive, something more like corruption or outright dishonesty. As described earlier, video evidence may turn out to be decisive in cases where one party has blatantly lied about what happened—not because they perceived reality differently from others, but because it was in their interest to lie. Psychological phenomena such as the illusion of objectivity can leave viewers with the distinct impression that people who disagree with them are corrupt or downright crazy, even when people on all sides are honestly reporting reality as they see it. But if one side really is corrupt—if they are reporting not the truth as they see it, but outright falsehoods instead—then we might expect video footage to more thoroughly and conclusively debunk the manufactured version of events. Again, further research is needed to test these hypotheses empirically.

C. Policy Implications

The empirical results reported here indicate that video evidence remains susceptible to bias, and while videos probably do not exacerbate bias compared to other types of testimony, we lack evidence that they are an improvement. Does this mean that body cameras are a bad idea, contrary to the views held by over ninety percent of Americans? What is the harm of having them, one might ask, if they probably don’t make things worse and people from across the political spectrum are clamoring for them?

One risk is that ocular proof may lead people to become more convinced that they are right. If the justification for body cameras is that they will handily resolve disputes by providing objective, decisive evidence, then we should worry that the public is overly optimistic about videos’ effect on polarization. The results reported here paint the opposite picture: we lack convincing evidence that videos reduce bias in our judgments, and we have evidence that they increase certitude that our judgments are correct. Video evidence, then, may provide a recipe for increased entrenchment and polarization.154

Second, civil rights advocates embracing body cameras hope that more video footage will hold accountable police officers who use excessive force. This Note’s findings raise the possibility that, on the contrary, indictments and convictions of police officers may become harder to obtain in some cases. Even with video evidence, people tend to bring their prior attitudes toward the police to bear on their judgments, and people who strongly identify with police appear especially likely to become resolute in their stances when they feel they have video proof backing them up. It’s possible, then, that the Mike Brown Law could actually make it harder to get a majority of grand jurors to vote to indict police officers like Darren Wilson.

Third, this research has focused on the potential for videos to be susceptible to viewers’ biases based on their attitudes toward police. But even beyond this type of viewer bias, we might worry that videos are susceptible to other sources of bias. For instance, at what point during an altercation does the camera start recording, especially if the police officer controls the on/off switch?155 Or what if the camera, because of its superior microphone or low-light technology, captures things that the police officer himself couldn’t see or hear—might juries implicitly penalize the officer for failing to notice these elements? Cameras are often thought of as neutral observers, but they can introduce their own distortions.

The angle of the camera can also be powerful in framing the encounter. The position of a body-worncamera, in particular, is worth considering. A camera that is affixed to a police officer’s uniform faces the citizen, not the police officer. Granot and colleagues’ work suggests that footage focusing mostly on what the citizen does might reduce biased factual findings: the researchers observed polarization based on identification only among viewers who looked at the police officer.156 On the other hand, we might worry that body-worn cameras will produce reams of footage, all from the perspective of the police officer. Previous psychological research indicates that camera footage that foregrounds the citizen while placing the police officer out of view tends to leave viewers with an overall impression favorable to police.157 Indeed, civil rights proponents imagine that body-worn cameras will surveil the police,158 but body cameras are importantly different from the civilian-captured footage that has galvanized the Black Lives Matter movement. Body cameras face outward, taping the public. They do not face the police.159

In short, we have numerous reasons to worry that body cameras are susceptible to different types of bias. But ultimately, whether the body camera proposal amounts to sound public policy depends on the justifications for the reform. The public’s sweeping support for body cameras is likely driven by several factors. As described earlier, people may support the reform because they believe body cameras will deter the unnecessary use of force, enhance transparency and public trust in the police, and provide new opportunities for police training, among other reasons. There are numerous public policy arguments in favor of body cameras, many of which are unrelated to their anticipated effect on bias. And, of course, there are many public policy arguments against body cameras that are independent of their effect on bias: concerns about privacy, cost, and unresolved questions about who will have access to the footage.160

It is beyond the scope of this study to provide an all-things-considered conclusion about the wisdom of the body camera proposal. The results suggest, however, that we should remove from the policy calculus one powerful assumption: that videos will provide unambiguous evidence and thereby reduce polarization and societal conflict.

Conclusion

Commentators on all sides of the debate over police shootings have assumed that video evidence is an improvement over nonvideo evidence in terms of how decisively it can resolve contentious deadly force cases. Common sense suggests that ocular proof is objective. But the study findings reported here suggest that video evidence remains susceptible to biased interpretation. In deciding factual matters about what happened—such as whether a weapon was present, whether physical force was used, whether the citizen complied with the officer’s requests—fact finders reviewing video footage brought their prior attitudes toward the police to bear on their judgments. In deciding more subjective matters—such as whether the citizen posed a threat, was likely armed at the time, or was resisting arrest—viewers again relied on their level of identification with police. When judging whether the police treated the citizen fairly—a key determinant of whether police are seen as wielding legitimate authority—viewers again drew upon their prior attitudes. Finally, when it came to making the ultimate judgment about whether the individual officer should be held accountable for excessive force, viewers’ judgments were influenced by their general identification with police. Overall, we lack evidence that fact finders reviewing video footage show less susceptibility to bias than fact finders reviewing other forms of testimony. At the same time, evidence suggests that access to video footage causes some fact finders—namely, those who already strongly identify with police—to become more adamant that their interpretations of the facts are objectively correct.

Police-worn body cameras are broadly popular among the public, and they will likely proliferate in the future. In the wake of public outrage and polarization over police use of force, body cameras seem like a promising solution. But before we embrace the idea of having police officers videotape everything they see, we should demand empirical evidence that cameras represent an improvement over the status quo. We simply do not have evidence to support this assertion, and the data reported here give reason to doubt it.

For the Appendix to this Note, please see the PDF version.