A key behavioral characteristic of autistic individuals is difficulty with mentalization—the ability to understand and respond to the mental states of others. While these difficulties have important social and clinical consequences, the neurocognitive mechanisms underlying them remain poorly understood. In this presentation, I will argue that one reason for this lack of progress is the limitation of standard experimental approaches used to study mentalization. These methods typically focus on the static application of fixed mentalization strategies and fail to capture a crucial feature of real-life social interactions: the fact that other people’s behaviors and mental states are constantly changing over time.
To address this, I will present a novel behavioral paradigm and computational model designed to assess adaptive mentalization — the ability to dynamically estimate an interaction partner’s level of behavioral sophistication and to flexibly adjust one’s own strategy in response. Using this approach in combination with fMRI, we identified a multivariate neural fingerprint that can predict, even in new individuals, how well they can update their mental models in response to changes in others’ behavior.
We have now applied this paradigm in a preregistered study of autistic individuals. Our findings show that higher levels of autistic traits are associated with both reduced capacity for adaptive mentalization and a weaker expression of the same neural fingerprint. These results demonstrate that it is possible to measure a specific, dynamic deficit in mentalization that is linked to autistic traits—both behaviorally and neurally. We hope this work contributes to a more nuanced theoretical understanding of both mentalization and autism, and may ultimately support improved methods for its assessment and diagnosis.