Comparing the performance of a large language model and naive human interviewers in interviewing children about a witnessed mock-event

PURPOSE

The present study compared the performance of a Large Language Model (LLM; ChatGPT) and human interviewers in interviewing children about a mock-event they witnessed.

METHODS

Children aged 6-8 (N =  78) were randomly assigned to the LLM (n =  40) or the human interviewer condition (n =  38). In the experiment, the children were asked to watch a video filmed by the researchers that depicted behavior including elements that could be misinterpreted as abusive in other contexts, and then answer questions posed by either an LLM (presented by a human researcher) or a human interviewer.

RESULTS

Irrespective of condition, recommended (vs. not recommended) questions elicited more correct information. The LLM posed fewer questions overall, but no difference in the proportion of the questions recommended by the literature. There were no differences between the LLM and human interviewers in unique correct information elicited but questions posed by LLM (vs. humans) elicited more unique correct information per question. LLM (vs. humans) also elicited less false information overall, but there was no difference in false information elicited per question.

CONCLUSIONS

The findings show that the LLM was competent in formulating questions that adhere to best practice guidelines while human interviewers asked more questions following up on the child responses in trying to find out what the children had witnessed. The results indicate LLMs could possibly be used to support child investigative interviewers. However, substantial further investigation is warranted to ascertain the utility of LLMs in more realistic investigative interview settings.

Forensic Interview Techniques in Child Sexual Abuse Cases: A Scoping Review

Child sexual abuse (CSA) is widely recognized as a global public health problem with negative consequences for victims, their families, and society. The child's testimony is essential to the case outcome, given the frequent absence of physical or biological evidence of the abusive acts. Thus, the child forensic interview plays a decisive role in criminal investigation. The present scoping review aims to identify and describe the judicial procedures for collecting CSA victims' testimony using an evidence-based approach and a structured methodology. The review followed Preferred Reporting Items of Systematic Reviews and Meta-Analysis-Scoping Review guidelines. Studies were identified through manual reference checking and in four electronic databases: PsycARTICLES, PubMed, SCOPUS, and Web of Science. In all, 146 studies were identified according to the defined inclusion criteria, that is, empirical studies identifying judicial procedures to collect CSA victims' testimony, published in English or Portuguese. In total, 30 different forensic interview procedures to collect the child victim's testimony were found. The National Institute for Child Health and Human Development investigative interview protocol was the most frequently mentioned. Despite the variety of protocols, it was possible to conclude that they have a similar general structure. This review also identified gaps in interviewing practices with CSA victims. The scoping review corroborates the importance of forensic interviews with CSA victims, stating its implications for criminal investigation, the legal system, and the child's recovery process.

Increasing disclosures of older adult maltreatment: a review of best practices for interviewing older adult eyewitnesses and victims

The current article provides a contemporary review of the best practices for interviewing older adults about crimes they witnessed or experienced. Specifically, we provide a detailed overview of how investigators can use a range of interviewing adaptations and procedures to acquire detailed and accurate maltreatment disclosures from older adults. In addition to discussing well-established investigative interviewing practices (e.g. rapport building, free-recall, cognitive instructions and prompts), this article also summarizes more recent literature on the benefits of multidisciplinary investigations, use of assistive technology and interview accommodations, along with effective practices for working with minority and marginalized older adults. Lastly, several research, professional and policy recommendations are provided for supporting older adults during investigative interviews.

"I think they do not know how to lie:" The perceptions of legal support staff about person with intellectual disabilities/autism in Turkish legal system

BACKGROUND

Increased social integration of individuals with intellectual disabilities or autism could may raise their likelihood of interacting with justice system and legal staff.

AIMS

The present article aimed to determine the perceptions of the legal support staff about the individuals with intellectual disabilities or autism.

MATERIALS AND METHODS

In the study, conducted with the phenomenological method, the views of 15 legal support staff were obtained. The study reported what participants perceived and examined whether participants had an accurate understanding of the experiences/support needs of people with intellectual disabilities/autism. The data were analysed with content analysis.

RESULTS

Three themes were determined: (1) Experience of interview with an individual with intellectual disabilities/autism, (2) attitudes of judges, lawyers and other staff, and (3) abuse/trauma.

DISCUSSION

Findings are consistent with literature. Individuals with intellectual disabilities/autism have problems in expressing themselves, also staff do not know them, do not know their features. Studies show that staff often do not know what to do in these interviews.

CONCLUSION

Based on the participant perspectives, it was determined that individuals with intellectual disabilities/autism experienced self-expression problems, interviewees did not understand children/individuals with intellectual disabilities or autism and could exhibit negative attitudes, the children could be abused and traumatised during judicial processes, they were more prone to abuse when compared to their peers, they could be involved in certain events more frequently and could be pushed to crime.

The dynamics of two-session interviews with suspected victims of abuse who are reluctant to make allegations

Supportive forensic interviews conducted in accordance with the National Institute of Child Health and Human Development Revised Protocol (RP) help many alleged victims describe abusive experiences. When children remain reluctant to make allegations, the RP guides interviewers to (a) focus on rapport building and nonsuggestive support in a first interview, and (b) plan a second interview to allow continued rapport building before exploring for possible abuse. We explored the dynamics of such two-session RP interviews. Of 204 children who remained reluctant in an initial interview, we focused on 104 who made allegations when re-interviewed a few days later. A structural equation model revealed that interviewer support during the first session predicted children's cooperation during the rapport-building phase of the second session, which, in turn, predicted more spontaneous allegations, which were associated with the interviewers' enhanced use of open-ended questions. Together, these factors mediated the effects of support on children's free recall of forensically important information. This highlighted the importance of emphasizing rapport with reluctant children, confirming that some children may need more time to build rapport even with supportive interviewers.

Effects of Delay, Question Type, and Socioemotional Support on Episodic Memory Retrieval by Children with Autism Spectrum Disorder

Twenty-seven autistic children and 32 typically developing (TD) peers were questioned about an experienced event after a two-week delay and again after a two-month delay, using the Revised National Institute of Child Health and Human Development (NICHD) Investigative Interview Protocol. Recall prompts elicited more detailed and more accurate responses from children than recognition prompts. Autistic children recalled fewer correct narrative details than TD peers when questioned using open invitations, cued invitations, and directive questions. Nonetheless, they were as accurate as TD peers when responding to all types of prompts. The informativeness and accuracy of children's reports remained unchanged over time. Social support was beneficial when children were interviewed for the first time but not after a longer delay.

Verbal, Visual, and Intermediary Support for Child Witnesses with Autism During Investigative Interviews

Three promising investigative interview interventions were assessed in 270 children (age 6-11 years): 71 with autism spectrum disorder (ASD) and 199 who were typically developing (TD). Children received 'Verbal Labels', 'Sketch Reinstatement of Context' or 'Registered Intermediary' interviews designed to improve interview performance without decreasing accuracy. Children with ASD showed no increases in the number of correct details recalled for any of the three interview types (compared to a Best-Practice police interview), whereas TD children showed significant improvements in the Registered Intermediary and Verbal Labels interviews. Findings suggested that children with ASD can perform as well as TD children in certain types of investigative interviews, but some expected benefits (e.g., of Registered Intermediaries) were not apparent in this study.