Transfer of Avatar Training Effects to Investigative Field Interviews of Children Conducted by Police Officers

Abstract

Previous research with students and some professional groups (psychologists) has demonstrated that repeated feedback in simulated investigative interviews with computerized child avatars improves the quality of interviews conducted with real children who have witnessed a mock event. However, it is not known whether this type of training would improve the quality of investigative interviews with actual child victims and witnesses of physical and sexual abuse. Twenty-two police investigators participated in the study. Half of them received feedback during four simulated interviews whereas the other half received no feedback during four such interviews followed by another four interviews after which they also received feedback. Transcripts of interviews both before and after the training were coded for interview quality. Receiving feedback after the simulated interviews increased the proportion of recommended questions both within the simulations and, importantly, also during interviewing with actual child victims and witnesses. This study demonstrated for the first time transfer of learning from simulated interviews to actual investigative interviews.

Keywords: child sexual abuse (CSA); interview training; investigative interviewing; serious games; training with virtual reality.

FIGURE 1

Experimental design.

FIGURE 2

Effects of feedback on proportion of recommended questions during the avatar training interviews. (A) Estimated marginal means (EMM) of the difference in initial four interviews between Group A (n = 11; received feedback) and Group B (n = 11; did not receive feedback). (B) Difference in Group B’s initial four interviews (n = 9; did not receive feedback) and subsequent four interviews (n = 9; received feedback). *Significant pairwise difference between conditions at the indicated interview. Error bars show 95% confidence intervals. Bonferroni correction applied *p < 0.05, **p < 0.01, ***p < 0.001 (2-tailed).