Technology on trial: facilitative and prejudicial effects of computer-generated animations on jurors’ legal judgments

ABSTRACT The current study explored how a computer-generated animation (CGA) illustrating a defendant’s version of events affected jurors’ judgments in a mock second-degree murder trial. We hypothesized that mock jurors who viewed a CGA illustrating the defendant’s testimony would be more likely to acquit compared to those who viewed static visual images or did not view a visual aid, and that this effect would occur regardless of whether the narrative depicted in the CGA was corroborated by pertinent testimonial evidence. In this 2 (testimony congruence: incongruent vs. congruent) x 3 (testimony modality: no-aid vs. static visual aid vs. computer-generated animation) between-subjects design, undergraduate students (N = 238) read a transcript from a fictitious trial and heard the defendant’s testimony in one of three modalities. Across congruence conditions, participants were significantly more likely to acquit the defendant when his testimony was accompanied by a CGA (OR = 5.08), compared to a static visual aid or with no-aid. Our results suggest that CGAs may have a disproportionate impact on jurors’ judgments compared to traditional forms of demonstrative evidence. Whether this impact is facilitative or prejudicial, however, depends on whether the content of the animation is congruent or incongruent with other case evidence.


Admissibility considerations for computer-generated animations
Unlike computer-generated simulations, which rely on mathematical calculations and laws of physics to scientifically model an event (e.g. exposing the mechanical failure in a plane's engine which caused it to crash; Louie et al., 2007), computer-generated animations are intended to illustrate existing case details or testimony for the purpose of clarifying facts of the case to the jury. As such, these animations are most often classified as forms of 'demonstrative evidence,' a category of evidence which also comprises more traditional static visual aids such as pictures, maps, diagrams, and models (Fiedler, 2003;Legate, 2006).
In the United States, demonstrative evidence is most often exempt from the evidentiary rules that govern the admissibility of physical evidence and, as such, does not need to meet the Frye standard of general acceptance (Frye v. United States, 1923) or to have its scientific accuracy evaluated against the more rigorous Daubert standard (Daubert v. Merrell Dow Pharmaceuticals, 1993) As long as demonstrative evidence is deemed by a judge to be relevant pursuant to Federal Rules of Evidence (FRE) 401 and 402, and a fair and accurate representation of the facts under FRE 901, it should technically be considered admissible (Federal Rules of Evidence, 2019). In addition, U.S. FRE 403 (i.e. the 'balancing test') allows judges the discretion to exclude relevant visual evidence if its probative value (i.e. its utility to prove facts at issue in a case) is substantially outweighed by the danger of unfair prejudice, confusion of the issues, or misleading the jury (Federal Rules of Evidence, 2019). Although exclusionary discretion is intended as a judicial safeguard to protect the jury from seeing inflammatory or prejudicial evidence (Paciocco & Stuesser, 2015), animations may have a greater capacity to bias or mislead jurors than traditional forms of demonstrative evidence. In the absence of empirically supported criteria to guide admissibility decisions, a judge's ability to make informed decisions about the probative and prejudicial capacity of animations may reasonably be compromised (Fiedler, 2003). Accordingly, judges may be hesitant to admit a relevant CGA, even when it technically meets legal standards of admissibility. On the other hand, a CGA may erroneously be admitted into trial based on meeting only basic evidentiary criteria (i.e. relevancy and materiality; Paciocco & Stuesser, 2015), when in its prejudicial effect should reasonably warrant its exclusion (Legate, 2006).
Evidently, the use of CGAs in trial has preceded the development of evidentiary rules that adequately dictate their use, creating a grey area where the grounds for inclusion or exclusion of these exhibits is somewhat variable and undefined (Legate, 2006). Without an adequate understanding of the persuasive capacity of animations and their effects on jurors' legal decision-making, evidentiary advisory committees are not in a position to revise existing policies or enact new ones which appropriately govern their admissibility in legal contexts. Accordingly, the intent of the current study is not to make generalized claims about the utility of animations as a whole, or to propose specific recommendations about their universal inclusion or exclusion in court. Rather, our goal is to provide empirical evidence of the cognitive effects of animations on jurors, with the hopes that this information will encourage the justice system to revisit evidentiary guidelines that are insufficient to govern novel forms of digital evidencesuch as animationsand to make amendments that align with contemporary psychological insights.
An empirical basis for the persuasive effects of computer-generated animations Although animations are commonly endorsed by lawyers with anecdotal claims such as 'a picture is worth a thousand words,' (Wiggins, 2004) and 'seeing is believing' (Bennett et al., 1999;Selbak, 1994), such claims about their purported persuasive value are also largely supported by empirical research. For example, both the dual-coding hypothesis (Paivio, 2007;Sadoski & Paivio, 2013) and the cognitive theory of multimedia learning (Mayer, 2014) rely on the assumption that auditory and visual information are processed through two distinct channels, each with a finite capacity. According to these cognitive theories, individuals organize incoming stimuli through mental representations, and have an enhanced ability to understand and retain information when it is explained verbally and with images. In addition, research has shown that vivid (defined as being emotionally interesting, concrete and image-provoking;e.g. pictures vs. words) are more memorable than pallid imagery (Clark & Paivio, 1991;Paivio, 1969;1971;1986), and research on visual attention has identified both color (e.g. Wolfe & Horowitz, 2004) and motion (for a review, see Wetzel et al., 1994) as visual features that attract attention. Within the field of psychology and law, vivid testimony is known to have powerful effects on jurors' memory and affect (Bell & Loftus, 1985), and a number of influential studies have shown that demonstrative aids such as photographs (Bright & Goodman-Delahunty, 2006Douglas et al., 1997;Salerno & Bottoms, 2009) and videos (Fishfader et al., 1996;Kassin & Garfield, 1991) have a stronger persuasive impact on jurors' judgments than verbal descriptions alone. In light of these findings, it reasonably follows that animationswhich represent an especially dynamic visual mediummay similarly enhance jurors' ability to attend to and retain case evidence, and could have a powerful influence on their legal judgments.
The story model (i.e. the predominant theory of juror-decision-making), proposes that jurors actively organize various strands of evidence and then use common sense to construct a plausible narrative about the events that occurred during a crime (Pennington & Hastie, 1992). Although jurors may initially construct multiple stories to account for how a crime occurred, one version will ultimately be selected based on its coherence (i.e. the extent to which the story makes sense, is plausible, and does not contradict itself) and coverage (i.e. the extent to which the story is able to incorporate most if not all information presented at trial; Pennington & Hastie, 1992). In effect, this model predicts that imagination from the starting point of a visual animation would require less effort and attention than imagination from a blank slate (Pennington & Hastie, 1992). Indeed, research has shown that jurors are more likely to render verdicts consistent with the Prosecution or Defense's arguments (i.e. guilt or innocence), and to be more confident in these judgments, when case evidence was presented in a temporal order that made a narrative of conviction or acquittal easier to construct (Pennington & Hastie, 1992). Thus, the presence of an animation at trial may become a key point of reference for jurors, as it creates a plausible and realistic version of events for jurors, without them having to expend a great deal of cognitive energy in order to construct this narrative themselves (Norris, 2015).
With this said, assertions that CGAs invariably facilitate (i.e. enhance) legal judgmentwhether based on theoretical or anecdotal evidenceare premature; in fact, a sizable proportion of cognitive psychological research indicates that CGAs may have prejudicial effects on jurors' cognitive processes and subsequent judgments. For example, the availability heuristic is a decision bias that accounts for the way in which individuals tend to rely more on information that is easy to recall or simpler to interpret when making judgements (Tversky & Kahneman, 1973). Animations may work on a visceral level that bypasses rational faculties, essentially creating a 'pseudo-memory' of an event that may or may not have actually occurred (Selbak, 1994). Thus, jurors may overestimate the probability that the scenario depicted in an animation is likely to have occurred, simply based on the ease in which the visual imagery can be retrieved from their memories (Sanna et al., 2002).
The hindsight bias, or 'knew-it-all-along' phenomenon, describes the tendency for individuals to overestimate the predictability of a past event after being exposed to information about its outcome (Fischoff & Beyth, 1975). In a study by Roese et al. (2006), researchers compared hindsight bias among participants who were exposed to an ambiguous traffic scenario either within a computer-generated animation or a text-plusdiagram depiction. After being exposed to the scenario's outcome (i.e. a serious car accident), participants were then asked to disregard their knowledge of this outcome and attempt to predict the odds that an accident would occur, hypothetically, from the perspective of a naïve observer. Results revealed that hindsight bias was more than doubled among participants who had viewed the animation compared to those who had viewed the text-plus-diagram depiction of the scenario. In other words, participants in the animation condition were largely unable to set aside prior knowledge about the outcome of the scenario such that they reported significantly higher estimates that the scenario would result in an accident than those who viewed the diagram. Roese et al. (2006) attribute these findings to the effects of enhanced processing fluency; the visual properties of an animation made case-relevant information feel familiar to participants and provided a coherent causal explanation for the crime's outcome, thereby exacerbating hindsight bias.
Dual-systems theories of information processing (e.g. heuristic-systematic model, Chen & Chaiken, 1999;dual-process theory, Kahneman, 2011) and persuasion (e.g. elaboration likelihood model, Petty & Cacioppo, 1986) suggest that cognition and attitudinal change are governed by two distinct systemsone which is quick, intuitive and preconscious (e.g. 'heuristic'/'System 1'/'peripheral'), and the other slow, sequential and effortful (e.g. 'systematic'/'System 2'/'central'). Each of these processing routes are employed under distinct circumstances, and even minor alterations to how information is framed or presented can impact whether one processing route is engaged over the other (Gilovich et al., 2002). Research has identified the modality of presentation as a powerful determinant of the cognitive route that individuals use to process case information, and that this may negatively impact subsequent judgments. For example, recent studies have demonstrated that presentation mediums such as PowerPoint® may encourage peripheral processing, leading mock jurors to be less thorough when evaluating case evidence and compromising their ability to render informed legal judgments (Park & Feigenson, 2013;Rempel et al., 2019). These findings suggest that digital visual technologies, including CGAs, may encourage jurors to process evidence using the peripheral route, encouraging them to form inferences that extend beyond the actual evidentiary content of the case.
Finally, research has found that visual technologies tend to have the greatest impact on jurors' culpability judgments when their use is differential; in other words, when only one party uses a dynamic visual technology to present evidence, jurors tend to render judgments against the party who presented their argument with less advanced means (e.g. Dunn et al., 2006;Kassin & Dunn, 1997;Park & Feigenson, 2013). Although animations have become more affordable in recent years, their cost can easily range from $5,000 to $50,000 (Austin Visuals, 2018) and some parties (most often the defense) may have to forego the use of this evidence due to financial limitations, potentially leaving them at a disadvantage to defend their client's innocence (Hadley, 2009;Schofield, 2011).

Controlled-experimental research on computer-generated animations
Despite their increasing use in the courtroom, only a handful of experimental studies have evaluated the effects of CGAs on mock jurors' legal judgments. Some of these preliminary investigations suggest that animations affect mock jurors' above and beyond traditional forms of demonstrative evidence, while others have failed to find this effect. For example, graphic elements of an animation, such as the angle of presentation (Norris, 2013) or the make and model of a vehicle (Norris & Reeves, 2012) have demonstrated an effect on individuals' culpability judgments in cases involving automotive collisions. In a study by Bennett and colleagues (1999), however, an animated recreation of a motor vehicle accident did not affect participants' apportionment of guilt and damage awards to a greater extent than aerial photographs and oral testimony of the same accident. Research by Dunn and colleagues (2006) proposed familiarity as a moderator of the relationship between animations and their persuasive effects on jurors. When an animation depicted a scenario with which participants were relatively familiar (e.g. a car crash), the information provided by the animation became superfluous,counteracting the animation's persuasive effect. However, when an animation depicted a scenario that participants l were unfamiliar with (e.g. a plane crash), mock jurors were more inclined to accept the version of events presented within the animation in order to fill a gap in their own knowledge. Nonetheless, due to the small number of empirical studies in this areamany of which have methodological limitations including a lack of external validitya clear picture of the persuasive capacity of animations has not yet emerged. Kassin and Dunn (1997) were among the first to explore the impact of animations on legal decision-making. They tested the hypothesis that animations can either facilitate or prejudice jurors' understanding of case evidence and subsequent judgments, depending on the degree to which the scenario depicted in the animation corroborates other case evidence. Participants watched a trial video involving a dispute about whether a man who fell to his death from a tall building had accidentally slipped (i.e. the plaintiff's argument) or committed suicide (i.e. the defendant's argument). Participants were presented with either a pro-plaintiff ver sion of the case (in which the body was found 5-10 feet awaya distance indicating he had slipped and fallen) or a pro-defendant version of the case (in which the body was found 15-20 feet awayindicating he had intentionally jumped). Within each condition, the testimony was presented either orally or with a computer-generated animation. When participants viewed the evidence within an animation, they rendered verdicts which were more in line with the physical evidence than those who only heard the oral testimony (i.e. those in the pro-plaintiff condition rendered more pro-plaintiff judgments than those who heard only oral testimony; those in the pro-defendant condition rendered more pro-defendant judgments than those who heard only oral testimony).
In a second experiment, Kassin and Dunn (1997) presented the animation in a suggestive way that contradicted the physical evidence. For example, in the pro-defendant version, despite the victim's body being found only 5-10 feet away from the building, the victim was shown running and jumping off of the building; in the pro-plaintiff version, the victim was shown slipping and falling, despite his body being found 15-20 feet from the building. When the animation was presented suggestively, jurors were more likely to render judgments which contradicted the physical evidence. Hence, animations were found to have a stronger impact on jurors' verdicts than oral testimony alone, but whether the animation had a facilitative or prejudicial effect on their judgments depended on its congruence with other case evidence.
In sum, CGAs are being admitted into court with increasing regularity, on the grounds that they enhance jurors' understanding of contested events in a case, thereby facilitating their legal judgments (Wiggins, 2004;Norris, 2015). Although animations are generally subject to the same regulations as other demonstrative aids, psychological theory on cognition and persuasion (e.g. elaboration likelihood model, Petty & Cacioppo, 1986; dualsystems theories, Chen & Chaiken, 1999;Kahneman, 2011;story model, Pennington & Hastie, 1992) and experimental evidence (Dunn et al., 2006, Study 2;Kassin & Dunn, 1997;Roese et al., 2006) suggests that animations may have a greater influence on jurors than traditional demonstrative aids (see Bennett et al., 1999 for an exception). Therefore, in order to inform the development of evidentiary rules that reliably govern their use in the courtroom, it is vital to develop a preliminary understanding of the effects that computer-generated animations have on jurors. As such, further empirical research in this area is needed to better understand the probative and prejudicial value of animations in legal settings.

Present research overview
Beginning with Kassin and Dunn's (1997) theoretical framework regarding the potential facilitative or prejudicial effects of animations, we presented a computer-generated animation illustrating the defendant's testimony in a second-degree murder case. The current study included three substantive changes to their original work to extend the generalizability of the findings. First, CGAs have undergone a marked increase in realism and complexity in the past two decades. Accordingly, the animation used in the present study was reflective of the current standard of courtroom technology and, as such, was more sophisticated and technologically advanced than the original. Second, Kassin and Dunn (1997) noted as limitations that their trial was brief and oversimplified, and that mock jurors rendered judgments immediately after viewing the animation. As such, the current study attempted to maximize external validity by including several components that were absent from the original trial stimulus (e.g. detailed testimonies and cross-examination of all witnesses), and to create a temporal delay between the presentation of the animation and subsequent verdict decisions. Finally, rather than investigating the effects of animations within a civil trial, the current study was the first to do so within a criminal trial. The Research Ethics board at Ryerson University approved this research (REB # 2018-334).
Premised on the assumptions of dual-systems theories of information processing and persuasion, we hypothesized the following: (1) Facilitation Hypothesis. In a case where a defendant's testimony is congruent with other key testimony, accompanying the defendant's testimony with an animation would facilitate judgments, such that mock jurors would be more likely to acquit the defendant than those who viewed static visuals, or those who did not view any aid.
(2) Prejudice Hypothesis. In a case where a defendant's testimony is incongruent with other key testimony, accompanying the defendant's testimony with an animation would prejudice judgments, such that mock jurors would be more likely to acquit the defendant than those who viewed static visuals, or those who did not view any aid. 1 (3) Exploratory Analyses. In addition to these directional hypotheses, we conducted a series of exploratory analyses to determine whether modality of the defendant's testimony affected other case measuressuch as perceptions of the case strength, attorneys and defendant's testimony-and whether these valuations mediated the effect of modality on verdicts. We did not make a priori predictions about these variables.

Design
The study used a 2 (testimony congruence: incongruent vs. congruent) x 3 (testimony modality: no-aid vs. static visual aid vs. computer-generated animation) between-subjects design. Participants read a fictitious trial transcript which featured a police officer on trial for second-degree murder. Participants heard the defendant's oral testimony while viewing a static visual aid, a computer-generated animation or without a visual aid. To test the facilitation and prejudice hypotheses, we systematically manipulated the content of the transcripts such that key pieces of testimonial evidence either contradicted the defendant's version of events (i.e. incongruent condition) or corroborated the defendant's version of events (i.e. congruent condition). The primary variable of interest was a dichotomous verdict decision (i.e. guilty or not-guilty), however participants also completed two continuous measures, confidence in verdict and probability of guilt estimates, which are known to be more sensitive to experimental effects (Diamond, 1997).

Materials
All study materials are publicly accessible on the Open Science Framework at https://osf. io/b7p3w/.

Judicial instructions
Prior to being introduced to the case, participants were provided with a written set of preliminary jury instructions that oriented them to basic legal concepts such as the presumption of innocence, burden of proof, and the legal criteria underlying a second-degree murder charge. After participants read the case, they were provided with a second set of instructions which contained verdict criteria. These instructions reminded participants that they must acquit the defendant if the Prosecution had not met the burden of proof, or if the homicide was lawfully justified by self-defence. Participants were also provided with relevant self-defence criteria.

Trial transcripts
The trial transcript featured the indictment, opening and closing statements from each lawyer, and the testimony of eight witnesses. To manipulate congruence, the testimony of two key witnesses were altered, such that certain details that these witnesses provided either corroborated, or conflicted with, the defendant's own account of the crime. As an example, in the 'congruent' transcript, Laura Lee (the paramedic) testified that she saw a gun lying beside the victim's body; Dr. David Lawrence (the forensic pathologist) testified that an autopsy revealed that the victim was shot in his side along his ribcage. These details corroborated the defendant's testimony that the victim was in possession of a weapon, and had his arm extended in a position to shoot the defendant prior to being shot. In the 'incongruent' transcript, however, Laura Lee testified that she saw an iPhone lying beside the victim's body; Dr. David Lawrence testified that the victim was shot in his back. These details are discrepant with the defendant's 'self-defence' narrative of the crime, suggesting instead that the victim was not in possession of a weapon, and was walking away from the defendant at the time of the shooting. Aside from the testimony provided by these two witnesses, the congruent and incongruent transcripts were identical (more details on the substantive differences between the two transcripts can be found on OSF). In sum, the qualitative changes to the testimony resulted in two versions of the trial transcriptthe congruent transcript, wherein 5/8 witnesses testified for the Defense, and the incongruent transcript, wherein 5/8 witnesses testified for the Prosecution. Participants read the transcript on laptop computers within a 45-slide PowerPoint® presentation.
Defendant's Testimony. Participants listened to oral testimony from the defendant, Officer Gary Green. Within his testimony, the defendant, a white male, describes approaching a group of young black males who he and his patrol partner believe to be suspects in a local robbery. The defendant explains howshortly after approaching the groupone of the suspects takes off running, pulls out a gun and aims it at him. The narrative provided by the officer legally supports a justifiable homicide in selfdefence. Depending on condition, this testimony was accompanied with a computer-generated animation, a static visual aid, or no aid. Across all conditions, the defendant's testimony spanned a total of three minutes and six seconds.
Computer-Generated Animation. Participants listened to the testimony while viewing a computer-generated animation on a laptop. The animationoriginally created for the civil case Avery Cody Sr., et al. v. County of Los Angeles (2010), by Eyewitness Animations®depicted the defendant's account of the crime from a several vantage points. The animation was a highly realistic, 3D rendering of the scene of the crime and involved persons (as described by the defendant), and was presented in full color in time with the audio testimony.
Static Visuals. Participants listened to the testimony while viewing five static visual images representing the defendant's version of events. These still images were screenshotted from the animation and displayed to each participant digitally on a laptop in a sequence coinciding with the audio. Each visual was displayed for between 10-40 s.
No-Aid. Participants listened to the testimony with headphones while viewing a black screen.

Verdict questionnaire
The primary dependent variable was a dichotomous verdict decision. Participants were asked to indicate a verdict in the case (i.e. guilty or not-guilty) by checking the box which best represented their opinion about the defendant's guilt. For exploratory purposes, participants were also asked to estimate the probability that the defendant was guilty of second-degree murder (reported as an open-ended percentage from 0-100%), and to report how confident they were in their verdicts (on a Likert scale from 1-10).

Evidence strength scale
For exploratory purposes, participants completed the Evidence Strength Scale, adapted from Koehler et al. (2016) which previously reported good internal consistency (α = .92). This scale measured participants' perceptions of attorney efficacy (e.g. 'How prepared/credible/competent/persuasive was the prosecutor/defense attorney?'), the defendant's testimony (e.g. 'How convincing was the testimony from the defendant, Gary Green?'), and the strength of the Prosecution and Defense's cases (e.g. 'How strong was the Prosecution/ Defense's case?') on Likert scales from 1-7. Within this questionnaire, participants also answered four multiple-choice attention check questions (e.g. 'Which of the following individuals did not testify in the court case?') to ensure participants were paying attention and reading the transcript thoroughly. Participants passed the attention check by correctly answering at least 3 out of 4 of the questions.

Demographics questionnaire
Participants completed a brief demographics questionnaire, which included questions about their age, gender identity and ethnic/cultural background.

Procedure
After providing written and verbal consent, participants were presented with preliminary jury instructions which outlined their responsibilities as mock jurors and oriented them to general legal terminology. On a laptop computer, each participant read through the trial transcript which was presented in a series of PowerPoint® slides. Participants were able to navigate through the transcript slides at their own pace; however, they could not go back and view slides that had already passed. The transcript began with the indictment, opening statements from each attorney, and the examination and cross-examination of the Prosecution's witnesses. Next, participants listened to the defendant's oral testimony which was embedded within the PowerPoint® as an audio file. This audio recording was presented without an additional aid (i.e. participants viewed a black screen), or was accompanied by static visuals or a computer-generated animation which played on screen as participants listened. Following the defendant's testimony, participants read the examination and cross-examination of the remaining Defense witnesses, as well as closing statements from each attorney. Participants were then presented with a folder containing verdict criteria, and a hard copy of the verdict questionnaire in which they were directed to record their responses. Participants placed their confidential verdict questionnaire in a file folder and returned it to the PI when it was completed. Participants then completed the adapted Evidence Strength Scale (Koehler et al., 2016) and demographics questionnaire on Qualtrics® before being debriefed. Participants were given one hour to complete the study.

Verdict analyses
Data was collected from 240 participants, but two participants failed the attention check and were excluded from the analysis. Preliminary analyses revealed nonsignificant effects of participant age, gender identity and ethnic/cultural background. All reported analyses are collapsed across those factors. Across conditions, 161 (67.6%) reported that the defendant in the case was guilty, compared to 77 (32.4%) who reported that the defendant was not-guilty (see Table 1 for frequencies across congruence and modality condition). Two post-hoc power analyses were conducted using G*Power3 (Faul et al., 2007) for each of our predictors. These two-tailed analyses revealed that our study was adequately powered (0.99) given the established medium (OR = 5.08) and large (OR = 10.49) effect sizes for modality and congruence, respectively.

Verdict
To test for main effects of modality and congruence, we conducted a binary logistic regression with congruence and modality as independent predictors of the dichotomous verdict variable (where 'guilty' was coded as 0 and 'not-guilty' as 1). A Helmert contrast was used to compare the mean probability of participants rendering a not-guilty verdict between 1) computer-generated animation vs. static visual aid and no-aid conditions, and 2) static visual aid vs. no-aid conditions. Results of this analysis revealed a significant main effect of modality on verdict, participants who viewed a computergenerated animation (51.3%) were 5.08 times (B = 1.63, SE = .36) more likely to report that the defendant was not-guilty compared to those who viewed a static visual aid or no-aid (22.8%), Wald's X 2 = 20.91, df = 1, p < .001, OR = 5.08 95% CI [2.53, 10.2]. Moreover, the odds of participants rendering a not-guilty verdict did not significantly differ between the static visual aid (25%) and no-aid (20.5%) conditions, Wald's X 2 = .44, df = 1, p = .581, OR = 1.32, 95% CI [.58, 2.99] (see Figure 1).
Results of this analysis also revealed a significant main effect of congruence: participants in the congruent condition (54.4%) were 10.49 (B = 2.35, SE = .37) times more likely to report that the defendant was not-guilty compared to participants in the incongruent condition (12.3%), Wald's X 2 = 40.65, df = 1, OR = 10.49, 95% CI [5.09, 21.6] (see Table 1 for verdict distribution across conditions).

Exploratory analyses
For exploratory purposes, we analyzed the effects of modality on perceptions of the case strength, attorney efficacy, and the defendant's convincingness. Linear regression analyses revealed that the interaction between modality and congruence was not significant across any of the subsequent exploratory analyses: perceptions of the defense (p = .506) and prosecution's (p = .951) case strength, perceptions of the prosecutor (p = .940) or the defense attorney (p = .252), and defendant's testimony (p = .580). Therefore, all exploratory results represent the relationship between modality and each index of evidence strength with the interaction term removed from the linear regression models while controlling for congruence. Statistical results of these analyses can be found in Table 2.
These analyses revealed null effects for three of five exploratory variables. While evidence modality did not significantly affect perceptions of the Defense's case strength (p = .149), the prosecutor (p = .207) or the defense attorney (p = .128), modality did have a significant effect on participants' perceptions of the Prosecution's case strength (p < .05) and the defendant's testimony (p < .05). A follow-up analysis of simple main effects using planned contrasts revealed that, while those in the no-aid condition perceived the Prosecution's case was significantly stronger than those in the animation condition (p < .05), this difference was not significant between participants in the no-aid condition and those in the static visual condition (p = .586). In addition, those in the no-aid condition perceived the defendant's testimony as significantly less convincing than those in the computer-generated animation condition (p < .05); however, this difference was not significant between participants in the no-aid condition an d those in the static visual condition, p = .398.

Follow-up mediation analyses
Premised on the exploratory linear regression analyses, which demonstrated that evidence modality was a significant predictor of the prosecution's case strength and the defendant's convincingness, we conducted additional mediation analyses to test the possibility that the effects of modality on participants' verdicts were mediated by either of these variables.

Discussion
The current study found support for Kassin and Dunn's (1997) facilitation and prejudice hypotheses and contributes to the extant experimental research on computer-generated animations in the courtroom. As predicted, mock jurors in our study were substantially more persuaded by a defendant's testimony when it was presented alongside a computer-generated animation compared to a static visual aid or without an aid. This effect occurred across congruence conditions, such that when case evidence corroborated a defendant's testimony, the animation brought judgments in line with the facts of the case; when case evidence conflicted with a defendant's testimony, however, the  animation had a biasing effect on mock jurors, such that their judgments diverged from the verdict that would be expected had they adequately considered other pertinent case facts (i.e. a guilty verdict). Thus, the premise of our paper is that it is not the defendant's testimony itself that appealed to mock jurors, but rather the inclusion of a dynamic visual animation that impacted their judgments above and beyond traditional means of testimonial presentation.
Aside from serving as a successful conceptual replication of Kassin and Dunn's (1997) original study, our results extend the previous findings in several ways. First and foremost, the observed facilitative and prejudicial effects in the current study closely mirror those of Kassin and Dunn (1997), despite the current animation being substantially more realistic and technologically-advanced than the original. Indeed, some scholars have suggested that it is the mere presence of an animationrather than the quality or realism of the animationthat is especially important to jurors' cognitive processing of its content (e.g. Bennett et al., 1999). Second, because we situated our animation within a criminal case (complete with legal terminology, instructions, and standards of proof unique to a criminal trial) we can conclude that results initially observed by Kassin and Dunn (1997) extend beyond the context of civil law. Finally, the addition of certain exploratory measures allowed us to glean valuable supplementary data about why jurors chose a particular verdict, as is delineated below.
According to Kassin and Wrightsman (1983), judgments of guilt arise when jurors' estimates of an individual's probability of guilt exceed their criterion for reasonable doubt. Intuitively, we observed a relationship between participants' probability of guilt estimates and verdicts, such that participants who held strong convictions about the defendant's guilt were less likely to acquit the defendant. Accordingly, individuals in the incongruent conditionwho generally believed the defendant was culpablewere also more confident in their verdicts than those in the congruent condition, who were more likely to believe in his innocence. In terms of modality, those who viewed the animation were less convinced of the defendant's guilt than those in the no-aid or static visual aid condition and were also less confident in their verdicts. These results fit well with suppositions about the persuasive capacity of animations: individuals appear to have been compelled by the narrative depicted within the CGA to the extent that it diminished their belief in the defendant's guilt, and ultimately swayed their verdicts toward acquittal.
Notably, our study is the first to implicate two specific trial components as mediators of the relationship between computer-generated animations and their impact on jurors' verdicts. Specifically, the use of a CGA was found to increase the convincingness of the defendant's testimony, which, rather than bolstering the Defense's case, undermined the narrative that was endorsed by the Prosecution. These findings are in line with theories of persuasion, such as the elaboration likelihood model (Petty & Cacioppo, 1986) and related research which has demonstrated that individuals tend to draw inferences about the credibility of especially dynamic or vivid information that extend beyond the actual content of this information (e.g. Cheng & Lo, 2012;Cooper et al., 1996). We speculate that the animation's null effect on the Defense's case may reflect that participants distinguished between discrete pieces of evidence in the Defense's case, such that they found the animation particularly convincing, but this did not impact their impressions of the Defense's case strength as a whole. The increased acquittal rate in the animation condition may indicate that, although participants weren't overwhelmingly convinced by the Defense's case, the animation itself was robust enough so as to prompt reasonable doubt in the Prosecution's case. Future investigations may wish to further expand on this line of inquiry, and supplement these findings with qualitative data, such as open-ended responses from participants, which may allow researchers to better understand how perceptions of animated evidence translate to the observed influence on legal judgments. In general, the results of this study fit well with psychological theory regarding jurors' decision-making processes as well as existing cognitive research on persuasion and learning.
The use of a CGA to deliver the defendant's testimony differentially affected jurors' judgments when compared to the other forms of demonstrative evidence that were presented. In line with the story model (Pennington & Hastie, 1986), it could be that mock jurors in our study had to expend substantially less effort to construct a plausible narrative about how the crime occurred, as this narrative was constructed for them within the animation. Given that mock jurors in our study would have reasonably had limited personal knowledge about the series of events that occurred in this case, the narrative depicted in the animation appears to have become a key reference for jurors which they heavily relied when during their decision-making process.
The results of this study also align with research on the effects of visual multimedia on individuals' information processing (e.g. dual-coding theory, Paivio, 2007). Regardless of the veracity of the events depicted in the animation, the findings in our study support that animations are superior to static visual aids in terms of their ability to encourage intentional (i.e. learning that occurs deliberately; Hulstijn, 2003) and incidental learning (i.e. learning that occurs unintentionally; Kelly, 2012). For example, participants in our study evidently acquired substantive case information that informed their verdicts, but may have also picked up incidental details about the crime itself or parties involved in the crime that werealso integrated into their judgments. Given the degree to which the features of the animation used in the current study (i.e. a highly realistic computergenerated animation with simultaneous first-person, oral narration) corresponded with the principles of multimedia learning (see Mayer &Moreno, 2002 andMayer, 1999), it reasonably follows that the animation that was employed in this study would have a robust impact on mock jurors' verdicts, and that its influence would presumably extend to jurors in the real world as well.

Implications
Aside from its theoretical value, the present study produced findings that have applied utility within the justice system. First and foremost, the current study provides further information about the potential probative and prejudicial value of computer-generated animations; namely, our findings support the conclusion that animations are a potent form of evidence that can both facilitate and prejudice participants' legal decisionmaking, depending on the whether the scenario they depict accurately represents the facts of the case or not. Although the animation in our study did facilitate jurors' judgments when the scenario it depicted aligned with other pertinent evidence, when this scenario misrepresented case facts, the animation prejudiced jurors and led them to render judgments which favored the Defense's self-defence theory of the crimeeven when the remainder of case evidence did not support this theory.
As with other demonstrative aids, the purpose of computer-generated animations is to illustrate testimony in a way that is an accurate representation of the facts of the case. To the extent that a mistaken eyewitness identification may prompt an entire case to be built around the wrong defendant, computer-generated animations may similarly jeopardize the administration of justice by perpetuating a false narrative of the events that occurred. In the current study, it could be the case that mock jurors interpreted the version of events depicted in the animation as factual, rather than as a theory of events derived from interviews with involved parties or witnesses to a crime. Because research on the prejudicial effects of computer-generated evidence is in its relative infancy, however, it remains to be determined exactly how animations exert their powerful persuasive effects. With this said, we might expect that these preliminary investigations into the prejudicial impact of animations may follow a similar trajectory to that of other, more well-established contributors to wrongful conviction; that is, that they may well eventuate into a robust body of literature which not only provides answers to empirical questions, but also engenders substantive changes to trial procedure (e.g. such is the case with Telfaire instructions, which now precede the admission of eyewitness evidence into trial; United States v. Telfaire, 1972).
Given that Federal Rule of Evidence 403 dictates that evidence must be excluded from trial if its probative value is substantially outweighed by the danger of unfair prejudice, the admission of potentially misleading or inflammatory animations may violate a defendant's right to a fair trial. Thus, these findings may serve as a reference for judges in their appraisal of this evidence, and as an educational tool for legal professionals who seek to develop this novel form of evidence in accordance with federal rules of evidence. Although our findings undoubtedly provide a valuable starting point for this assessment, it would be difficult, if not impossible, to identify 'universal' element(s) of an animation that can be deemed as prejudicial across all circumstances. The intent of this studyand this line of research in generalis not to isolate specific graphical elements (e.g. angles, colors or details) of an animation and mandate their inclusion or exclusion into court, but rather to provide empirical evidence of the robust effects of animations on jurors, and to encourage judges to be mindful of this when determining their admissibility. In the short term, lawmakers may wish to consider requiring attorneys to establish the relevancy, materiality, and authenticity of an animation prior to it being admitted into the courtroom. In the long term, policymakers should certainly consider holding computer-generated animations to a higher standard of evidentiary regulationone that is more akin to the criteria which governs substantive evidence.

Limitations and future directions
Although we made every attempt to replicate the experience of real jurors and the circumstances of a real trial, this study was conducted in a controlled, laboratory setting with a sample of undergraduate students. Despite our sample of participants representing a relatively diverse range of ethnic backgrounds, our sample consisted exclusively of undergraduate students from a Canadian university, and therefore may not be entirely representative of the general population. In addition, mock jurors worked individually for the duration of the study and did not have the opportunity to deliberate with a jury of their peers before rendering their verdicts. Thus, it is difficult to conclude whether the results of this controlled experiment with undergraduate participantswherein certain components of a real trial were noticeably absentwould generalize to real trial settings with jury-eligible citizens. With this said, the extent to which our convenience sample and lack of deliberation truly impairs the verisimilitude of this research is debateable; research investigating such concerns has shown minimal differences between student and nonstudent samples across various decision indices (Bornstein et al., 2017), and deliberation has shown to be negligible at nullifying jurors' initial impressions of a defendant's guilt (Kalven & Zeisel, 1966;MacCoun & Kerr, 1988).
While our findings represent a valuable addition to the extant research on computergenerated animations in legal settings, our streamlined trial paradigm and materials may have limited the ecological validity of our findings in certain ways. In a real trial setting, for example, we would expect a judge to deliver a 'cautionary' or 'limiting' instruction prior to the admission of potentially prejudicial evidence into trial (Aronson & McMurtrie, 2007;Hadley, 2009). These instructions, which direct jurors to disregard inadmissible evidence (Cush & Delahunty, 2006), specify the purposes for which the evidence may be used (Lieberman, 2009) or to educate the jury on factors which may affect the reliability of the evidence (e.g.., United States v. Telfaire, 1972) were notably absent from our study. While the exclusion of such an instruction was necessary to maintain experimental control, future research should investigate whether a limiting instruction, which explicates to jurors that a particular animation is intended only as an illustrative aid and not as substantive evidence, serves as an effective remedial strategy to reduce jurors' overreliance on the animations.
Moreover, in the current study, an animation was presented by only one legal party (i.e. the Defense) rather than by both attorneys. In a real trial setting, however, it would be reasonable to expect that the Prosecution may choose to admit a comparable animation depicting their own theory of events. Research exploring other digital presentation technologies (i.e. PowerPoint ®) has shown jurors are most persuaded by digital technologies when their use is differential (Park & Feigenson, 2013). As such, it is possible that the observed results overestimate the magnitude of the effect of animations on mock jurors' perceptions of evidence and verdicts. It would therefore be advantageous to explore whether and/or how jurors' judgments are affected when animations are used by both legal parties, rather than by only one party andin additionto determine whether the order of presentation impacts the saliency of an animation in jurors' Moreover, it is not entirely clear why participants in our study felt that the defendant's testimony was especially convincing in the animation condition, yet this did not influence their overall valuations of the Defense's case strength. Previous research has demonstrated a discrepancy between the true influence of animations on mock jurors' judgments, and what mock jurors perceive the influence of an animation is. In a study by Dunn and colleagues (2006), researchers observed a negative correlation between selfreported persuasion and actual persuasion. When participants were asked about the degree to which an animation impacted their verdicts, they reported that a defendant's use of an animation made their case stronger, despite this animation not influencing their verdicts. In contrast, participants did not believe that an animation created by the plaintiff made the plaintiff's case stronger, yet this animation did have an effect on their verdicts. In the current study, participants may have been similarly unaware of the degree to which the animation bolstered the Defense's case, such that they felt convinced by the animation but not feel that the animation affected their perception of the Defense's case as a whole, or that it overpowered their judgments. Without more information, we can only speculate on why this is the case; however, this would certainly be an interesting question to explore in future research.
Finally, it is crucial to acknowledge the socially charged nature of the trial materials that were used in the current study (i.e. a mock trial featuring a white male police officer and a black victim). This study was conducted in late 2018 and the particular animation and case stimulus were selected well in advance of the tragic events of early 2020 that drew widespread attention to Black Lives Matter (BLM) and related social-justice movements. Since this time, flagrant instances of violence toward black individuals at the hands of white police officers (such as the highly publicized murders of George Floyd and Breonna Taylor) have rightfully led to an insurgence of public attention toward racial injustice. Although racial stigma was not the primary focus of the current study at the time it was conducted, we feel that this is a crucial area of research that warrants further exploration. As animations allow for the visual depiction of an individual's visible features, a defendant or victim's race may become increasingly salient within an animation compared to other modes of presentation. Future research should explore how demographic characteristics such as race, sex and age are represented and perceived within animations, and use these findings to inform recommendations pertaining to how animations can be developed in a way that minimizes the biases and stereotypes associated with various demographic factors. For example, animations could be presented monochromatically, rather than in full color, and avatars could be made to look more generic without identifiable physical features and skin tones.

Conclusion
Computer-generated animations are increasingly being admitted into court with minimal regulation yet, to date, there has been limited empirical research which has evaluated their effects on jurors. The results of this study indicate that CGAs have both facilitative and prejudicial potential that extends beyond that of existing forms of demonstrative evidence. It is therefore crucial that this novel form of evidence is properly validated by the appropriate rules of evidence before being admitted into trial. In the short term, our findings regarding the probative and prejudicial value of computer-generated animations provide valuable preliminary information about their persuasive capacity which may assist judges with appraisals of their admissibility. Looking forward, future research should further deconstruct the persuasive capacity of animations observed in the current study, from both a basic and applied psychological perspective. For example, researchers may wish to determine the cognitive and affective mechanisms underlying the persuasive capacity of animations, and to manipulate graphical components of an animation at a more granular level (e.g. POV, speed, repetition, color, level of detail, avatar representation). Together, these converging lines of research should lead us toward a more holistic understanding of animations, which may be used to inform lawmakers for the purposes of generating concrete evidentiary guidelines that govern CGAs thereby promoting fairness and equitability during trial. Note 1. Operationally, support for the facilitation and prejudice hypotheses was predicated on three statistical predictions. First, we expected a significant main effect of modality, such that participants in the animation condition would be more likely than those in the static visuals or no-aid condition to render not-guilty verdicts. Second, we expected a significant main effect of congruence, such that participants in the congruent condition would be more likely to render not-guilty verdicts than those in the incongruent condition. Third, we expected a non-significant interaction between these variables. If these three conditions were met, with the computer-generated animation increasing the likelihood of participants' acquitting the defendant across congruence conditions (i.e. a non-significant interaction between modality and congruence), the study will have provided evidence that the animation facilitated participants' judgments in the congruent condition (wherein an acquittal is supported by other key case testimony) as well as prejudicing participants' decisions in the incongruent condition (wherein an acquittal is discrepant with other key case testimony).

Data Availability statement
The data that support the findings of this study are available from the corresponding author [author initials], upon reasonable request.

Disclosure statement
No potential conflict of interest was reported by the author(s).