# Avatars and humans may not elicit the same accent-related biases in mock courtroom research

## TL;DR

This study shows that using avatars in simulated courtrooms may not produce the same accent-related biases as using human actors, suggesting avatars are not yet a reliable substitute.

## Contribution

The study introduces a novel comparison of human and avatar eyewitnesses in mock courtroom settings to assess accent-related biases.

## Key findings

- Avatar witnesses were rated more favorably than human witnesses.
- No main effects of accent or country of origin were found for avatars.
- The Lebanese non-standard accented human witness was rated more poorly than others.

## Abstract

Conducting research to better understand the role of extralegal factors in courtroom decision-making requires either labor intensive methods, such as simulating a trial, or approaches that are not ecologically valid, such as using short written case vignettes. If avatars could be used in simulated courtrooms, experiments could more easily manipulate extralegal variables for study without requiring significant resourcing, for example hiring actors and having access to a courtroom. The current study used previously developed stimulus materials of a human eyewitness in a courtroom and created a comparable avatar eyewitness and virtual courtroom to assess ratings of the human and avatar.

Participants (N = 703) saw one of 12 videos depicting an eyewitness on the stand at a criminal trial recounting a burglary. The design was a 2 × 2 × 3, mode of presentation (human or avatar), accent (General American English or non-standard) and country of origin (Germany, Mexico or Lebanon). Three actors voiced each human and avatar pair using General American English and one of the non-standard accents (German, Mexican or Lebanese) so that variation in ratings could be attributed to presentation mode, accent and country of origin.

An analysis of covariance revealed that the avatar witnesses were rated more favorably than the humans and there were no main effects of accent nor country of origin, contrary to previous research using the human video stimuli. A three-way interaction showed the Lebanese human non-standard accented witness was rated more poorly than her standard-accented counterpart, her avatar counterpart, and the Mexican and German human non-standard accented witnesses.

Findings reveal that avatar witnesses cannot yet reliably replace their human counterparts. Discussion as to what can be done in future to further investigate how to create courtroom stimulus materials is presented along with possible explanations as to the reasons for different findings in this research than previous studies.