![[PsyPost]](https://sp-ao.shortpixel.ai/client/to_webp,q_glossy,ret_img,w_750,h_375/https://www.psypost.org/wp-content/uploads/2025/03/brain-11-750x375.jpg)
A new study published in Nature Mental Health suggests that adults who report high levels of autistic traits through online surveys may not reflect the same social behaviors or clinical profiles as those who have been formally diagnosed with autism spectrum disorder. Although these two groups may appear similar on paper, the study found meaningful differences in their social interactions, mental health symptoms, and the way they perceive and influence others during social tasks. The findings raise questions about the widespread reliance on self-report surveys in online autism research and suggest that diagnostic evaluations by trained professionals remain essential.
The researchers, led by a team at the Icahn School of Medicine at Mount Sinai, were motivated by a growing trend in psychiatric research: the shift toward large-scale, web-based studies. Online platforms like Prolific allow researchers to gather data from thousands of people quickly and inexpensively. While this approach has clear advantages in terms of scale and accessibility, it often relies solely on participants’ own descriptions of their traits and mental health symptoms.
“By allowing people to participate in research remotely and anonymously, online studies ease many of the burdens associated with in-person involvement and allow researchers to recruit large samples quickly. However, as online studies become increasingly popular in psychiatric research, it is essential to evaluate their limitations,” said study authors Sarah M. Banker, who is now a postdoctoral fellow NYU Langone, and Xiaosi Gu, an associate professor and director of the Center for Computational Psychiatry at the Icahn School of Medicine at Mount Sinai.
“Notably, anonymous online research typically relies on self-report to measure symptoms and identify participants with specific psychiatric and neurodevelopmental conditions. Without the added context of clinical evaluation, self-reported symptoms may be limited in their ability to capture the experiences and diagnoses of people who have difficulties describing their emotions and struggles – including many autistic individuals. Our study sought to evaluate whether autistic participants identified through online self-reported symptoms show similar profiles to those identified through in-person clinical evaluation.”
The researchers compared three groups of adults. One group included 56 individuals who were diagnosed with autism after undergoing in-person clinical evaluations. The second group consisted of 56 people recruited online who reported high levels of autistic traits using a standard survey. A third group, also recruited online, reported low levels of autistic traits and served as a comparison. All participants were matched in age and gender to allow for fair comparisons.
The study involved a mix of questionnaires and interactive tasks. All participants completed the Broad Autism Phenotype Questionnaire, which measures characteristics often associated with autism, including social difficulties and rigid behavior. The in-person group also underwent clinical interviews and standardized assessments conducted by trained professionals.
Additionally, all three groups participated in two computer-based tasks designed to simulate social interactions. One task measured the participant’s ability to influence the behavior of virtual characters to gain rewards, while the other tracked their social decision-making over time, such as whether they acted warmly or distantly toward new characters.
The results showed that although the high-trait online group and the in-person autism group reported similar levels of autistic traits, their profiles differed in key ways. The online group with high autistic traits reported more symptoms of social anxiety and avoidant personality disorder than the clinically diagnosed group. In fact, their psychiatric profiles resembled those of people who are socially anxious but not necessarily autistic. This suggests that some people who score high on autism trait questionnaires in the general population may actually be describing a different type of social struggle—one rooted more in anxiety than in autism.
One of the most striking findings came from the comparison between self-report and clinical assessment. Among participants with a confirmed autism diagnosis, there was no meaningful relationship between how they rated their own symptoms and how clinicians rated them. This held true for both core domains of autism: repetitive behaviors and social communication. The two types of measures seemed to capture different aspects of the condition—one reflecting internal experience and the other focusing on observable behavior.
“We were surprised to find that, for adults with a confirmed autism spectrum disorder diagnosis (and without co-occurring intellectual disability), higher clinician-rated symptoms did not correspond to higher self-reported symptoms,” Banker and Gu told PsyPost. “This finding suggests that the two measures capture different aspects of autism: while self-reported assessments may capture internal experiences, clinician-rated assessments may capture external presentations.”
Differences also emerged in how participants behaved during the social interaction tasks. In the first task, which tested whether participants could influence virtual characters by accepting or rejecting monetary offers, the in-person autism group showed reduced ability to recognize and use social influence. They rejected fewer high offers and did not adjust their behavior in ways that would have led to better future outcomes. They also reported less awareness of their ability to influence the game, suggesting a diminished perception of social controllability.
In contrast, both the high-trait and low-trait online groups performed similarly on this task and showed more awareness of their influence. This points to a key distinction: while individuals in both online groups may share internal feelings of social difficulty, their external behavior in social contexts does not mirror the social differences seen in people with a clinical autism diagnosis.
In the second task, which involved choosing how to interact with different characters during a simulated narrative, the in-person autism group again showed unique patterns. They were less likely to choose friendly or affiliative responses, even when their self-reported feelings toward the characters were similar to those of the online high-trait group. This gap between internal attitude and external behavior was specific to the clinically diagnosed group, suggesting that their social behavior was more affected by their condition than by their feelings or intentions alone.
“Despite similarly high levels of self-reported autistic traits, adults recruited online (without self-identification of an autism spectrum disorder diagnosis) and those recruited in-person (with an autism spectrum disorder diagnosis confirmed through clinical evaluation) displayed different psychiatric profiles and tendencies during simulated social interactions, suggesting they may represent different groups,” Banker and Gu explained.
“Our results suggest that, without additional information (such as diagnostic information collected by trained professionals), self-reported symptom surveys may not be sufficient to identify or draw representative conclusions about certain diagnoses, including autism spectrum disorder. This work emphasizes the important distinction between diagnoses and traits and provides a caution against the potential for generating assumptions about one group based on the findings from research conducted in another.”
This matters not just for research, but also for clinical care. If online studies are used to shape understanding of autism or to develop interventions, it is important that the participants truly reflect the population the research aims to help. Otherwise, there is a risk of creating treatments or policies that are not effective—or even relevant—for those who are clinically diagnosed with autism.
“This finding is also interesting when considering the emerging use of online mental health platforms, where clinician evaluation is not always available and symptom severity (and improvement) only relies solely on self-report,” the researchers noted. “Our finding suggests that such approach might need to be revisited and compared against human clinician evaluations/ratings.”
But the researchers were also careful to note that self-reports still have an important role to play. They are valuable tools for understanding how individuals experience their own challenges. In fact, self-reports can offer insights that clinicians might miss, particularly when it comes to internal states like anxiety or depression.
“Although our findings highlight discrepancies between self-reported and clinician-rated symptoms, they do not diminish the value of self-report surveys in autism research,” Banker and Gu said. “In fact, self-reports are crucial for understanding individuals’ own experiences, internal distress, and overall wellbeing. They also empower individuals with lived experience to shape narratives about their condition and help to challenge inaccurate assumptions about the reasons behind autistic behaviors. Rather than discounting personal perspectives, our results emphasize caution for researchers in relying solely on self-reported symptom measures when defining or generalizing about diagnostic groups.”
The study, “Phenotypic divergence between individuals with self-reported autistic traits and clinically ascertained autism,” was authored by Sarah M. Banker, Miles Harrington, Matthew Schafer, Soojung Na, Matthew Heflin, Sarah Barkley, Jadyn Trayvick, Arabella W. Peters, Abigaël A. Thinakaran, Daniela Schiller, Jennifer H. Foss-Feig, and Xiaosi Gu.
