Flinders study highlights credibility risks for some autistic suspects

Flinders University highlighted research on October 9 in a report published by Phys.org showing how unexpected communication in mock police interviews can influence judgments of suspects. The findings identify particular credibility risks for some autistic interviewees and warn against treating the way someone answers questions as evidence of guilt.

The development is the university’s public report on experiments published online in Law and Human Behavior on September 21. The research examined observers’ impressions of interviewees; it did not measure wrongful convictions or announce a change in police interviewing procedures.

How the mock police interviews were assessed

The study, by Neil Brewer, Carmen A. Lucas, Katie Logos, Robyn L. Young and Alliyza Lim, used two experiments involving autistic and nonautistic observers. The first included 212 autistic and 215 nonautistic observers; the second included 125 in each group. Those figures describe the people evaluating interviews, not the number of suspects.

Observers watched online mock police interviews with autistic or nonautistic chatroom participants suspected of computer hacking. Interviewees were selected because they displayed relatively high or low levels of difficulty interpreting figurative language in the first experiment, or speech hesitation in the second.

Specific risks for autistic interviewees

The study abstract reports harsher evaluations of autistic male suspects than nonautistic male suspects. It also reports harsher evaluations when autistic suspects displayed more pronounced speech hesitation. These specific findings identify vulnerabilities within the broader pattern of responses to communication behavior.

Across the experiments, the observer’s own autism diagnostic status did not affect evaluations of suspects. Departures from expected behavior did, regardless of whether the suspect was autistic. The authors interpreted this pattern as support for expectancy violations theory: judgments were associated with whether behavior matched what an observer expected.

“The important thing is the extent to which someone's behavior violates expectations about how a suspect ‘should’ behave,” said Young, who leads the Flinders Autism Research Initiative, in the university report.

The researchers also tested a prediction associated with the double empathy hypothesis: that judgments would differ when observers and suspects did not share the same autism diagnostic status. Their experiments did not support that prediction. The findings do not establish that the hypothesis fails across other social settings.

Brewer said the researchers found no decisive evidence for their broad expectations of harsher judgments based on autistic behavioral characteristics or differing observer and suspect diagnostic status. The more specific findings in the published abstract nevertheless identify autistic men and autistic interviewees with pronounced hesitation as vulnerable to negative evaluations.

Earlier research on giving an interview account

A separate study by Ralph Bagnall and colleagues at Bath and Vrije Universiteit Amsterdam, published in Frontiers in Psychology on March 22, 2023, examined how people produced accounts during mock suspect interviews. Its virtual-burglary experiment involved 32 autistic adults and 33 nonautistic adults matched for age and IQ.

Innocent participants were instructed to tell the truth, while guilty participants were instructed to lie in a subsequent interview. Innocent autistic participants provided fewer details supporting their innocence: an average of 5.71 out of seven possible details, compared with 6.73 among innocent nonautistic participants.

The earlier researchers found no significant diagnostic-group differences in overall investigation-relevant information or consistency with evidence. Autistic participants nevertheless reported greater difficulty with questioning, more anxiety and less support. That study concerned giving an account, rather than observers’ judgments, so it was not a replication of the newer experiments.

Bagnall’s team cautioned that its above-average-IQ sample did not represent the whole autism spectrum. Participants had neither an appropriate adult nor a legal adviser, and a virtual task might not reproduce the stakes of a real investigation. The researchers called for further testing of supportive interviewing approaches.

What the findings mean for police and courts

In the October report, Flinders co-author and clinical psychologist Alliyza Lim said communication behavior may affect perceived credibility without necessarily providing evidence that someone committed an offense. The university’s warning applies to assumptions drawn from communication styles and interpretations of responses, intentions and motivations.

The report says some nonautistic suspects probably face credibility risks too. Its practical implications remain a research warning: neither the report nor the study abstract establishes that changing interview procedures improves real-world verdicts. No implementation timetable was announced.