Decoding another person’s mindset can be difficult. For autistic individuals, it can be even more challenging1,2. Yet this skill is essential to how people connect, communicate and build relationships. Clues to someone’s inner world can be found in their preferences, since what people think about often depends on what they are interested in. And compared to private thoughts and emotions, these inclinations are usually more openly shared and easier to observe.
Because preferences are a natural reflection of how someone sees and experiences their environment, they are also a useful way to probe why autistic and nonautistic people so often struggle to understand one another. The ‘double empathy problem’ theorizes that this difficulty is mutual and does not arise from shortcomings on either side, but rather from both groups developing different ways of experiencing and interpreting the social world over time3. To test this directly, Gabriela Rosenblau of George Washington University, a SFARI Fellows-to-Faculty awardee and SFARI Investigator, led a new study comparing how autistic and nonautistic individuals learn about each other’s preferences.
The researchers first asked all three groups (nonautistic adults, nonautistic adolescents and autistic adolescents) to rate how much they liked a series of foods and activities, from fruits and desserts to music, sports and arts and crafts. Autistic adolescents showed far greater variability in their preferences than the nonautistic groups. Where nonautistic adolescents tended to cluster around similar ratings for the same items, autistic adolescents had a wider distribution of responses.
They then gave nonautistic adults and autistic adolescents a learning task using the preference profiles gathered earlier. Participants were shown a profile of an adolescent and asked to guess, item by item, how much that person liked each food or activity. After every guess, they received the actual rating as feedback. The task was designed to capture social learning in action, piecing together a picture of someone else’s tastes through trial and error. Each participant completed it twice, once for a nonautistic adolescent and once for an autistic adolescent. From there, computational modeling revealed the strategies underlying each group’s learning.
The results were surprising. Both nonautistic adults and autistic adolescents predicted nonautistic adolescents’ preferences more accurately than those of autistic youth. The expected in-group advantage suggested by the double empathy hypothesis, where each group would do better with people most similar to themselves, did not emerge. The researchers point to the variability observed earlier as an explanation. Because the preferences of autistic youths varied so widely from person to person, knowing the group average said little about any specific person within it. The difficulty, in other words, was not a failure of social understanding, but rather that autistic adolescents’ preferences were harder to anticipate.
Despite their differences, autistic and nonautistic participants approached the learning task in remarkably similar ways, making precise, item-level updates rather than broad generalizations. Learning that someone loves sushi, for instance, might shift a prediction about their taste for other Asian foods, but not necessarily for candy. An earlier, smaller study from the same lab had suggested that autistic adolescents relied mainly on their own preferences when making inferences about others4. But the current, larger sample showed that autistic adolescents are more capable and flexible social learners than previously thought. Not all autistic adolescents performed the same, however. Those reporting greater behavioral rigidity or higher overall autistic traits tended to learn more slowly and make more errors, a reminder that social learning ability is far from uniform.
Together, these findings reframe how we think about social understanding between autistic and nonautistic people. The difficulty that each group has in reading the other runs deeper than a simple skills gap. Much of it stems from the diversity of preferences within the autistic population, which makes personal tastes harder to anticipate. Yet the study also found that both neurotypes rely on the same sophisticated learning strategy, suggesting that autistic adolescents are more adept at learning about others than previously appreciated. The computational framework used here makes it possible to move past broad group comparisons and reveal both commonalities and meaningful differences. Given the marked variability observed across the autism spectrum, this degree of precision may ultimately shed new light on the everyday challenges of social reciprocity and help inform support better tailored to individual needs.
