Recent research published in natural mental health suggest that both autistic and non-autistic people use very similar strategies when learning about the preferences of others. Although this finding provides evidence that both groups rely on flexible learning methods, the unique and diverse preferences of autistic individuals may make it difficult to predict their specific likes and dislikes. These insights can help explain common social misconceptions without assuming that people with autism lack the ability to read social cues.
To successfully navigate social environments, we need to constantly be aware of what others are thinking and feeling. One of the basic components of social interaction is learning about the other person’s preferences, such as favorite foods or hobbies. Understanding someone’s preferences makes it easier to predict their behavior, assess trustworthiness, and establish meaningful connections.
Recent concepts in psychology offer new perspectives on these social disconnects. The problem of dual empathy is a theory that suggests that breakdowns in communication between autistic and non-autistic people occur because both groups experience the world differently. According to this framework, both groups struggle to understand each other, rather than social deficits that belong only to autistic people.
The current study builds on previous research to investigate how social knowledge shapes learning among different groups. “Lead author Dr. Gabriela Rosenblau published findings in 2021 using a preference learning paradigm and computational modeling. The results showed that autistic and non-autistic participants employed different strategies when inferring the preferences of their peers,” said Shannon Cahalan.
Cahalan is a postdoctoral fellow at the National Institute of Mental Health in the Section of Social and Developmental Cognitive Neuroscience. She recently completed her PhD at George Washington University and served as lead author of the current study.
“Specifically, non-autistic youth’s learning may be explained by a complex model that integrates prior knowledge about peer preferences, item relational representations, and trial-by-trial updates, while autistic youth’s own preferences influence their inferences about peers,” Cahalan told SciPost. “Although this finding provided a potential plausible mechanism underlying social challenges in ASD, there were still several gaps to explore.”
The researchers designed the study to see whether people make more accurate inferences about others with the same diagnostic background. “For example, do these findings hold true in a larger, more representative sample?” Cahalan asked. “In the first place, how do autistic self-preferences differ from non-autistic self-preferences? Do autistic adolescents actually base their reasoning on other autistic peers? How does the diversity of autistic trait expressions predict social learning performance? We sought to address these open ends in this study.”
This study was divided into three separate experiments. The first experiment mapped the personal preferences of a large group of participants. The sample included 228 non-autistic adults, 125 non-autistic adolescents, and 255 autistic adolescents.
Participants completed an online survey and rated 120 different food and activity items. They used a six-point scale featuring emoji faces to indicate how much they liked or disliked things like sushi, roller skates, and candy. This step allowed the scientists to see exactly how the self-preferences of teens with autism differed from the other two groups.
The data showed that autistic youth have much more diverse preferences compared to non-autistic adults and teens. Teenagers with autism showed a clear pattern and tended to give higher ratings to items such as candy, writing utensils, and art supplies. They also tended to give lower ratings to items such as salads, vegetables, and fitness equipment than the non-autistic group.
Autistic participants also showed signs of behavioral rigidity in their preferences. Behavioral rigidity refers to a strong need for sameness and resistance to change, and is a common characteristic of autism. Related to this, autistic teens tended to group their preferences strictly by basic categories. This means that if you have a strong preference for one type of fast food item, your ratings for all other fast food items will follow a very similar pattern.
In the second experiment, scientists tested how 191 non-autistic adults learned the preferences of teenagers. Adults were divided into two groups, with 98 adults guessing the preferences of non-autistic teens and 93 guessing the preferences of teens with autism. The adults were unaware of the diagnostic status of the teens they were evaluating.
During the task, participants guessed how much a particular teenager liked an item and received immediate feedback on their actual rating. The scientists measured prediction error, which represents the mathematical gap between participants’ guesses and the true answer. A smaller prediction error indicates higher accuracy and better social learning.
Adults had smaller prediction errors when guessing the preferences of non-autistic teens. When learning about teens with autism, the adults initially made mistakes, but steadily improved their guesses over time. The computer model showed that the adults used fine-grained learning strategies for both groups, updating their guesses based on how similar items were to each other rather than lumping everything into broad categories.
In the third experiment, we shifted our focus to young people with autism. A group of 83 autistic teens guessed the preferences of non-autistic teens. Another group of 119 autistic teens guessed the preferences of other autistic teens.
The authors hoped that teens with autism would be able to more accurately judge peers who belonged to their diagnostic group. Actual data showed the opposite pattern. Autistic adolescents made more accurate guesses when predicting the preferences of non-autistic teens.
Just like adults, teens with autism actively used feedback to improve their guesses over time. They also used exactly the same fine-grained learning strategies. This contradicts the old idea that teens with autism rely solely on their own likes and dislikes when judging others. The authors noted that autistic teens did indeed use general knowledge about other autistic people to form their initial expectations, but this strategy did not lead to increased accuracy.
“I was very surprised at how important ‘variability’ was to the interpretation of our results,” Cahalan said. “I came in with high expectations that the results would be clear and that young people with autism would present distinct and unique social learning strategies.”
“Rather, our insights were based on how young people with autism themselves differ from each other,” Cahalan explained. “Working on this project has fundamentally changed the way I approach research focused on autism and related neurodevelopmental disorders.”
Researchers believe that the wide range of autistic preferences may make it difficult to predict individual autistic teens. The likes and dislikes of autistic people vary so much that relying on the average profile of the group is not very helpful when judging a particular person. Both adults and autistic teens found the non-autistic profile easier to predict. Because their preferences are much less diverse and fit more standard patterns.
“Diversity is a ‘feature’ of autism, not a ‘bug,'” Cahalan points out. “So the impact of variability was a core theme of our findings. On the other hand, people with autism may have more difficulty learning in general because of variability in their preferences.”
“The preferences of an ‘average’ autistic person are unlikely to reflect the preferences of a specific autistic person, and this has far-reaching implications for social learning,” Cahalan added. “On the other hand, variation in autism traits, such as rigidity, was particularly predictive of preferred learning abilities in teens with autism.”
In addition to these broad patterns, scientists looked at how individual personality traits influence learning within the autism group. Participants who reported higher levels of overall autistic traits tended to integrate new feedback more slowly. Those who reported higher levels of behavioral rigidity tended to make more errors overall during guessing games.
Several limitations provide context for how to interpret these results. This study compared the learning styles of young people and teenagers. Because the two groups were not the same age, scientists were unable to fully separate the effects of age from the effects of autism diagnosis.
The complex nature of the guessing task also posed challenges for some participants. A small number of autistic participants were unable to complete the experiment because the rules were too strict. The authors suggest that future research should create simpler activities that target people across the autism spectrum.
The sample of girls with autism was relatively small, with only 55 female participants participating in the main task. This imbalance made it difficult to detect reliable differences between how boys and girls learn about social preferences. Future research should include more balanced groups to see how gender interacts with social learning.
The study did not reveal whether the profiles participants were rating belonged to autistic or non-autistic people. Having a clear knowledge of your partner’s diagnostic background can change your learning strategy. Future research could examine whether revealing this information changes the way people approach social guessing games.
The rating scale used in this study was also limited in scope with only six options. This limited the extent to which the prediction error could vary during the mathematical analysis. The researchers plan to use a continuous rating scale with a wider range of options in future versions of the experiment.
The researchers now plan to expand on these findings by examining brain activity. “We continue to investigate social learning at a neural level in autistic and non-autistic youth,” Cahalan said. “Importantly, we obtained extensive neuropsychological, demographic, and behavioral data on our participants, which will allow us to further investigate how individual differences predict preferred learning performance and associated functional neural activity in autistic and non-autistic youth.”
The study, “Modeling how autistic and non-autistic groups learn their own and each other’s preferences,” was authored by Shannon Cahalan, Rafael Perla, Sophia Block, Mikayla Locklin, Christoph W. Kohn, and Gabriela Rosenblau.
