| Abstract [eng] |
Allegations of child sexual abuse often present procedural challenges, primarily stemming from a lack of supporting evidence. In these instances, the child interview emerges as a pivotal element in the investigation, underscoring its significant importance. It is strongly recommended that interviewers adhere to best practices to facilitate the child in providing pertinent details related to the case. Employing open-ended questions is advised to reduce the risk of introducing suggestive influences and contaminating the child’s narrative. Despite these recommendations, research indicates a tendency among interviewers to deviate from best practices, resorting to leading and confirming questions. Literature analysis suggests a connection between confirmation bias and the tendency to ask closed questions, with emotions playing a role in this deviation. Particularly in cases of child sexual abuse, emotions come into play, and their psychophysiological correlates serve as indicators of the cognitive demands of the task and are related to confirmation bias. This thesis aimed to explore the relatively unexplored realm of interviewer emotions and their correlation with confirmation bias in investigative interviews related to child sexual abuse. To achieve this objective, a series of pre-planned experimental studies were conducted, with slight modifications based on prior experiment outcomes. A contemporary approach was adopted, involving the objective measurement of facial emotions through a facial recognition engine, the assessment of psychophysiological responses using wireless portable sensors, and the manipulation of stimuli through virtual child avatars and scenarios depicting alleged sexual abuse. The findings from these studies unveiled a preliminary understanding of the interplay between emotions, psychophysiological indicators, and confirmation bias in the assessment and execution of child sexual abuse interviews. This knowledge contributes to a better understanding of how training for such interviews can be improved. |
