Maltreatment and abuse of children is a significant societal problem and with serious damaging effects on children’s behavior, psychological development and adjustment. Both Child Protective Services (CPS) and law enforcement play important roles in protecting and assisting mistreated children. In their roles as investigators, they must elicit from children detailed, coherent and reliable
accounts of their experiences. Researchers and professionals have developed best-practice guidelines for interviewing children to maximize narrative details. But despite huge investments in training programs, practitioners routinely fail to follow best-practices guidelines (Lamb, 2016). Recent studies show that interactive, computer-based learning activities can improve an investigator’s
interviewer performance (Powell et al., 2016). This project seeks to create an empirically based interview-training system using realistic avatars and to determine if such a system can effectively train CPS and law enforcement professionals to conduct interviews of consistently high quality. To achieve this goal, an interdisciplinary team of psychologists and AI experts with expertise studying and working with at-risk children will use extant AI technologies and data from past investigative interviews with maltreated children to create a real-looking child avatar capable of expressing emotion and apparently spontaneous responses. This avatar will become part of an interview-training system that will be implemented with the cooperation of the CPS and police, and studied by the project team. The an interview-training system that will be implemented with the cooperation of the CPS and police, and studied by the project team. The
training system will then be evaluated, using pre-training and post-training and long-term post-training assessments, to determine
effectiveness.