Mock suspects who broke observers' expectations in interviews were judged more harshly, autistic or not

Choosing suspects by how they behaved let the experiments separate two things that went together in these samples: a person's way of talking and the person's diagnosis ^[10]. Suppose the harsh ratings followed the diagnosis. Then autistic suspects should have fared worse whatever their behaviour. Suppose instead they followed the behaviour. Then the label should matter less.

Neil Brewer, an experimental psychologist and emeritus professor at Flinders ^[14], described the first idea the team tested. "First, there is the view that autistic people will be judged more harshly, purely because of any 'unusual' behavioral characteristics commonly associated with autism," he said ^[3]. The second was the double empathy hypothesis. On that view, autistic and nonautistic observers mainly struggle to read the intent of suspects who do not share their diagnostic status, and so may judge those suspects more harshly ^[4]. "However, we found no decisive evidence in line with either of these expectations," Brewer said ^[2].

The paper's title states what the data did support: expectancy violations theory ^[7]. "The important thing is the extent to which someone's behavior violates expectations about how a suspect 'should' behave," said Robyn Young, who leads the Flinders Autism Research Initiative ^[12]. She and Brewer say some autistic suspects may face particular credibility risks in investigative or legal settings, and that the same probably applies to some nonautistic suspects ^[11]. In my view this is the more useful of the two results for anyone who runs interviews, because the risk it describes follows behaviour and is not tied to one diagnosis ^[11].

The double empathy result needs a more careful reading. The phys.org write-up says the work undermines the hypothesis ^[1]. Brewer's own wording, "no decisive evidence," is narrower ^[2]. It means these experiments did not support the prediction. It does not measure autistic observers' edge and find it to be zero. The phys.org account describes "large samples" without numbers or effect sizes, and identifies the observers only as autistic or nonautistic ^[8].

The thing this doesn't tell you is whether a detective on a live case, or a juror in a real trial, would rate suspects the same way. These were mock interviews about a computer-hacking accusation, and nobody in them faced a real charge ^[8].

Alliyza Lim, a clinical psychologist and co-author ^[14], says communication behaviour may affect perceived credibility without necessarily providing evidence about whether a person actually committed an offence ^[13]. The paper appears in Law and Human Behavior ^[7].