I spent the summer after my sophomore year of high school bedridden with pneumonia, watching every episode of the crime show White Collar. I now associate sickness with procedural dramas — those whodunnit TV series that introduce a crime then solve it in the same episode. Watching these shows, I undermine my own standards for what makes “good” media. Instead, plot twists irk me. I crave predictability, the swaddling certainty that evidence leads logically to a perpetrator.
For most of us, television is the most vivid reference for what happens in the courtroom. Jurors raised on crime shows expect damning visual or forensic evidence, a phenomenon called “The CSI Effect.” Beyond Hollywood productions, however, the perfect illustrative evidence often does not exist.
It can be generated, though. Increasingly, attorneys turn to computers to create the missing visual evidence that will help them win their cases. The results are legal animations: computer-generated videos that illustrate an expert’s testimony or a witness’s description of an event.
These animations become visual aids to help jurors understand and remember the facts of a case. When used as tools to communicate testimonies, animations improve comprehension and decrease court time.
Still, something else — something more insidious — is happening when we view an animation. As computer-generated exhibits, animations lend a technical authority to the specific, often seductive, narratives they depict. Even though animations reproduce testimony, the experience is now mediated through a computer. The computer has become the witness.
On a large screen beside the jury box, the animation begins: Jennifer stands in her living room, realistically rendered with wire-rimmed glasses and animal-print pajamas. Her husband, Michael Serge, enters the room, gun drawn. A bright blue line extends from the gun through Jennifer’s torso, showing the trajectory of Serge’s first shot. Serge then fires into a wall before aiming again at his wife, now cowering on the ground. Another line, this time fluorescent green, depicts a final shot through Jennifer’s body. The screen turns black. An eerie silence grips the courtroom.
This 2001 murder case marked the first high-visibility use of animations as evidence, with members of the local legal community attending the trial to spectate. “They watched my client execute his wife,” Serge’s defense attorney later said. After less than two hours of deliberation, the jury found Serge guilty of first-degree murder.
Within the courtroom, animations are demonstrative evidence, meaning they are meant to aid the jury in understanding other evidence or testimony, but they are not evidence in their own right. To create a crime scene animation, a developer first gathers data from sources such as police reports, witness interviews, depositions, photographs, and expert investigations. She then decides the composition of each frame and renders it on the computer.
In an evidentiary sense, animations exist as a foil to simulations, the more scientific, more stringently-defined category of computer-generated exhibits. Simulations must use verified data and approved software to perform analysis. A car crash simulation, for instance, might input raw data about a car’s weight and speed to recreate the collision. Once admitted as evidence, a simulation is treated as substantive (or “real”) evidence and may be used to establish the facts of a case.
Animations, by contrast, are inherently subjective as reproductions of witness testimony. In Hinkle v. City of Clarksburg, a case involving a police standoff where a man was shot to death, the defense presented an animation showing the victim raising a shotgun towards the officers, though additional evidence of the victim’s body location, the timing of gunshots, and another officer’s testimony all conflicted with this animation. The court allowed the animation because it was “designed merely to illustrate [the defendant’s] version of the shooting.”
Animations point to two shades of truth that may conflict: truth as what factually occurred and “truth” as perceived by a witness. This softer, more subjective form of “truth” captures one version of the story, often a perspective deliberately chosen for persuasion. Animations therefore enable different sides of a case to construct coherent, convincing narratives in support of their arguments.
Animations point to two shades of truth that may conflict: truth as what factually occurred and “truth” as perceived by a witness.
Developers aiming to produce these narratives illustrate a tension between the fact of evidence-based animations and the fiction of film and entertainment. During cross examinations, developers argue that animations are not “movies,” with all the artificiality and melodrama that this distinction suggests. Instead, they emphasize animations as grounded in scrupulously-collected data, making them as legitimate and “objective” as other forms of evidence.
Still, developers reference Hollywood films to lend credibility to their animation techniques. Michael Schumaker, who worked with George Zimmerman’s defense team, bragged about using the same technology as the movie Avatar: For his animation, actors in motion-capture suits reenacted the events of Trayvon Martin’s death. Jason Fries, creator of an animation for the Van Dyke trial, compared his process of measuring the crime scene with lasers and drones to the techniques in Toy Story and Monsters, Inc.
It shouldn’t be surprising that animations are successful for many of the same reasons as films. Animations, like films, often make emotional appeals by focusing on the experience of those involved. In Salvator v. ValueJet, a case involving a plane crash, an animation allowed jurors to experience the “terror of the three and a half minutes it took [the defendant’s] plane to fall from the sky.” To quote Brian Stonehill, a media studies scholar, legal animation “works on a visceral level that quite easily bypasses skeptical, rational faculties.”
Here is the wizardry of animations: their ability to tell a story, to frame it as objective, and to do it all without the viewer even realizing.
This last part, the unconscious convincing, is essential. Jurors are often unaware about the effects of animations on their judgements. In a study by Meghan Dunn, Peter Salovey, and Neal Feigenson, mock jurors who were influenced by animations later self-reported that these displays were not important to their verdict.
Given the differences between court cases and the importance of context for juror decisions, research on persuasion is challenging. Existing studies show conflicting results: while multiple psychology studies observed changes among participants who viewed animations (more damages awarded, more acquittals), another study found no effect.
More conclusive studies focus on what kinds of animations persuade, and researchers agree that perspective can drastically influence juror opinions. In a 1995 study, participants saw either an animation of a plane crash from the cockpit view, an audiotape and written transcript of the cockpit voice recorder, or a reading of the transcript of the cockpit voice recorder. Mock juries who viewed the animation instead of the voice recording provided significantly lower ratings of responsibility for the flight crew. Researchers suggest that this difference may be due to empathy from viewing the simulation from the pilot’s perspective. In 2014, another study led by different researchers replicated this finding, with participants who viewed an animated sequence of a car collision reporting significant differences in their judgments of culpability based on the perspective of the animation (overhead, facing, or internal).
This is grounded in the Actor-Observer Effect, which describes the tendency to attribute the behavior of others to internal causes and attribute the behavior of ourselves to external, situational causes. A study on videotaped confessions found that participants rated confessions as genuine when viewed from the suspect’s point of view and coerced when viewed from the detective’s point of view.
The effects of animations are magnified in situations for which jurors lack direct experience. In the same study by Dunn, Salovey, and Feigenson, animations of a car crash did not affect mock juror decisions. For a plane crash, however, viewing an animation changed jurors’ judgments of responsibility. Researchers suggested familiarity with the subject matter as an influence for this discrepancy.
Most unnerving, though, is the fact that animations can mislead. In a 1997 study, mock jurors determined whether a man’s fall from a building was an accident or suicide. When jurors only saw the fall, judgements were consistent with physical evidence. However, when animations depicted the man running and jumping from the building in addition to the fall, jurors ruled the death a suicide, even though this conflicted with physical evidence. Alarmingly, the animated (demonstrative) evidence led jurors to disregard the physical (substantive) evidence.
The persuasive power of animations may be wielded unequally if only one side can afford this evidence. Today, animation firms quote prices starting at 1,000 dollars — a significant investment for a non-essential visual aid. The final cost may be much greater depending on the complexity of the display. Moreover, the costs of animations are felt by both sides. The opposing side may be pressured to develop their own animations or settle due to the inequality in evidence.
Often, specialized software is necessary to create animations, and the developer’s credentials may be examined when the animation is presented as evidence. This means specialized firms are often hired to create animations. Even if a litigant had the technical skills to create an animation, it might not be accepted by the court. Animations therefore remain out of reach for those with limited budgets or time, such as people represented by public defenders.
Unsurprisingly, this technology recreates and exacerbates already-existing inequalities. Still, legal animations create an entirely different form of inequality in the courtroom: though Americans tolerate a legal system that allows litigants to pay for better attorneys, animations introduce opportunities for people to pay for better evidence.
Animations therefore remain out of reach for those with limited budgets or time, such as people represented by public defenders.
Animations prove to be a slippery, often-inaccessible form of narrative construction. As seen by their use in police brutality trials, the reliance on this technology can lead to perverse consequences.
Following a low-speed reckless driving case in Brunswick, Georgia, police officers shot Caroline Small, a white woman, eight times. For the subsequent case, the Glynn County Police Department created an animation of Small running into officers with her car. The animation explained their decision to shoot, but the scenario was impossible: Small’s car was stuck in a ditch, with another officer’s SUV blocking her path. A Georgia Bureau of Intelligence (GBI) agent called the animation a “grossly inaccurate” picture of the crime scene. Still, after viewing the animation, the grand jury failed to indict the officers. The animation allowed the police department to manufacture their own truth, effectively covering up murder.
The perspective-altering ability of animations benefits officers during the legal process. In an era of ubiquitous cameras, police brutality is often captured on video by dash cameras, bystanders, or security footage. Yet, when these cases go to trial, the defense frequently uses animations to illustrate the situation from the perspective of the police officer.
This is illustrated in the murder trial for Chicago police officer Jason Van Dyke. Van Dyke shot Laquan McDonald, a Black man, 16 times in 2018. In one view in the animation, the camera appears to hover over Van Dyke’s shoulder while his gun aims at McDonald — a scene resembling a third-person shooter video game.
The prosecution noted multiple inaccuracies in the defense’s animation: McDonald is shown moving from 39 feet to 13 feet away from Van Dyke in four seconds (a speed the prosecution found unfeasible) and is depicted in all-black clothing (he actually wore jeans). The defense also argued that a dark red line highlighting McDonald makes him appear taller than in real life.
Other omissions in the animation heightened the impression that Van Dyke was in danger. The animation does not show Van Dyke’s bulletproof vest or the other officer trailing McDonald. Further, the animation only shows Van Dyke shooting McDonald five times, ending when McDonald hits the street without showing the remaining (11) shots fired. The animator testified that he only included the bullets he could track, an omission that fails to communicate the absurd number of shots made by Van Dyke. In this sense, the animation is not incorrect but incomplete. Yet, the omission works in their favor: At the end of the animation, if these additional 11 shots were shown, would jurors find Van Dyke’s actions justified?
Look at the controversial cases of the past few decades, and many feature animations on the side of the defense. We know the names of the victims: Rodney King, Tamir Rice, Laquan McDonald, Jacob Blake.
These animations attempt to show events before witnesses’ cameras began filming, creating a narrative that cannot be disproven with other video evidence. In cases of police violence, the defense may have special abilities to develop the animations: unlike the opposing side, they have access to resources and the ability to defend the perspective. In cases where the police officer fatally shot the victim, an opposing animation from the point-of-view of the victim is impossible. No witness testimony can accompany an animation from this perspective, so it would likely fail to be admitted as evidence. Instead, the defendant holds the power to construct a narrative.
Legal animations reflect contemporary animation technology, including the uncanny realism of humans. While most animations use generic, crash-dummy figures to represent those involved, police brutality animations often create life-like models with people’s specific attributes. For many of these cases, this is a reminder that the perpetrator is white and the victim was Black.
The realism of animations suggests that creators aim to represent not just the crime but the specific people involved. But should this be the case?
Often, the realism and perspective choice advances a troubling portrayal of Black victims, routinely humanizing the perpetrator while dehumanizing the victim. This is seen in the video game-like portrayal of Van Dyke shooting McDonald. Though not an example of police violence, the Zimmerman case also exemplifies the unequal portrayal of a white perpetrator and a Black victim. In the animation, Martin and Zimmerman appear as semi-realistic figures. At times, the viewer can see a side profile of Zimmerman’s face, while Martin’s face is always obscured by a black hoodie. Only Zimmerman is made available for empathy.
The realism of animations suggests that creators aim to represent not just the crime but the specific people involved. But should this be the case? All animations are strategic, and in these cases, their purpose is to make the victim seem more dangerous, therefore justifying the officer’s use of force. In this light, the videos become stomach turning. The figures are not just animations but effigies, used to discredit someone dead in favor of their killer.
What could fix the worst of computer-generated evidence?
The most direct solution is to prevent faulty animations from entering the courtroom. This involves clarifying and tightening the standards for legal animations. Currently, animations must meet the foundational requirements for any demonstrative evidence (authenticity, relevance, fairness), but federal admissibility rules cite no specific considerations for evidence created by a computer. This leads to inconsistencies in how judges rule on admissibility of animations, with judges themselves seeking greater guidance.
Given their increased persuasive effects compared to other forms of demonstrative evidence — like charts, drawings, or diagrams — animations merit special guidelines. These should include consideration of the realism of the display or the inequality if only one side presents animations. Additional guidelines could also codify language for attorneys to reference when opposing the admission of evidence.
Beyond policy-centered approaches, hope may be found in other applications of this technology. The visual investigations group SITU Research creates computer-generated evidence for civil rights and racial justice cases. Using a hybrid of animation and video evidence, SITU has worked with various civil liberties groups to reconstruct police repressions of peaceful protests in Charlotte, Portland, and The Bronx.
This kind of analysis is only possible with animation technology. Much of SITU’s work involves sifting through video evidence to determine when and where police actions occurred, then synchronizing evidence from multiple perspectives to show the full scope of police movements. These animations help to untangle what happened and elevate the voices of those who experienced the event, rather than relying on those in positions of power (who otherwise shape the narrative). In this light, a total ban of computer-generated evidence becomes not only unreasonable but potentially harmful, preventing a coherent narrative that can lead to better accountability. This also highlights the potential for pro-bono work by those with the correct technology and sufficient skills to make animations.
These measures achieve important work to address the current situation of legal animations. Still, more is needed. As animations converge to more and more realistic imagery, they require greater study and scrutiny, specifically in the context of race. This must come from academics across disciplines (law, psychology, media studies, ethnic studies), along with journalists and other practitioners.
It’s important to note that these solutions do not address the technology itself, but rather how it is used. On a technical level, work to improve the animations will improve the power of the tool, regardless who wields it; in some ways, this only increases the disparity between those with and without access to animations.
Still, technologists can exercise a wide latitude of control over their creations, including choices about perspectives and degree of realism. The “fairer” animation may be one that is less polished, less coherent. These technical choices won’t solve everything, but can be a starting point. Building a better world can mean building less, or knowing when to stop.
Christina Tuttle is a software engineer based in New York City.
Archana Ahlawat is a researcher and software engineer at the Princeton Center for Information Technology Policy.