Effective social interactions are essential for human well-being. Individuals with Autism Spectrum Disorders experience disproportionate difficulties with social interactions, causing isolation for patients, suffering for families, and extensive societal costs. However, in spite of decades of effort, we do not yet know what aspects of brain functioning explain these disproportionate deficits. We propose a novel hypothesis: that deficits in social interaction reflect reduced neural flexibility of social information processing. We have designed a task that separately measures neural responses to dynamic emotional facial expressions that are driven by the external stimulus (e.g. the appearance of the face) versus those that are determined by the observer’s internal goals (e.g. to pay attention to the person’s emotion, while ignoring her age). In a pilot study, we found that in control adults, patterns of neural responses in many regions of cortex reflected the participants’ internal goals, rather than the external stimulus. We hypothesize that in individuals with Autism, the influence of flexible internal goals will be reduced, while the influence of the external stimulus will be preserved or even enhanced. Reduced neural flexibility of social processing may help to explain the impairments experienced by individuals with ASD, in real world social interactions.