Avatar technologies are becoming increasingly common across many professional fields, including healthcare and education. However, their application as investigative interviewers for young people (children and adolescents) is comparatively under-researched.

This research investigated the effect of using human-operated avatars on the quality of young people’s forensic disclosure, episodic recall, and susceptibility to false information.

Using a mock investigative interview paradigm; Study 1 compared human interviewers with human-appearing avatars, also manipulating whether avatars were of the same or opposite sex as participants. Study 2 replicated the methodology of the previous study, comparing human interviewers with non-anthropomorphic avatars (characterized by a faceless, disembodied shape). Participants in both studies were between 11 and 17 years old.

Both studies suggest that avatar interviewers, regardless of their visual appearance, elicit comparable levels of disclosure and recall accuracy as human interviewers face-to-face. The limitations, stakeholder views, and practical implications will be discussed.

Recent technological advancements have unlocked the capability for synthesizing highly-realistic, but fake-faces which are indistinguishable from real-human faces. Such technologies offer promising alternatives for lineups by enabling the use of synthetic-filler faces. To investigate this, we tailored a pre-trained generative model to produce passport-like synthetic-faces of different races, which were evaluated alongside real-human faces and synthetically-reconstructed real-faces in two visual-perception experiments. In experiment 1, people sequentially classified faces as real or fake. Whereas experiment 2 adopted an eyewitness identification-lineup paradigm to examine whether a real-human face perceptually ‘popped-out’ when presented simultaneously alongside synthetic-filler faces. Results showed consistent response biases for assuming faces are real. Independently viewing faces found people were superior at discriminating real-faces but no better than guessing for fake-faces. Seeing faces in a group emphasized these perceptual differences, creating a ‘pop-out’ effect for real-faces. Race-effects were also detected with synthetic-faces for certain racial-groups being less convincing than others.

Online investigative interviews (tele-forensic interviews) with children have become more common. However, there is still limited information about how distractions in the online environment during an interview, such as the interviewer’s background and a child’s own video feed on their computer screen, may impact a child’s memory. The current study aimed to add to this growing literature by recruiting 100 children between 5 – 12 years old from a summer camp to take part in a tele-forensic interview about science activities where the teacher makes a transgression. We used a 2 (interviewer background: simple, child-friendly) x 2 (child’s video feed: seen, hidden) between-subject design and examined rates of accuracy, completeness, disclosure of the transgression and age differences. Preliminary results suggest that distractions such as interviewer background and video feed visibility may not impact children's memory as expected. Further results and implications will be discussed.

Accurate witness identification is a cornerstone of police inquiries and national security investigations. Yet the technology used to display lineups has not fundamentally changed over the past century. Eyewitness mistakes can have dire consequences, having been implicated in about 70% of wrongful convictions. Previous work has demonstrated that providing witnesses with a 3D interactive lineup - allowing the witness to rotate the faces 90° to the left/right - increases that ability to discriminate between innocent and guilty suspects compared to a typical photo lineup or video parade. The current study (N=750) employed a 4 (amount of rotation) x 2 (target present/absent) design to investigate whether the improvements associated with interactive lineups would persist without the full 180° of rotation. Participants demonstrated superior discrimination accuracy in all interactive conditions compared to the static condition. This persisted even when participants encoded frontally posed images of the suspect (as opposed to profile images).