Deficits in uncertainty processing in autism? Testing a new cognitive theory for autism based on high, inflexible precision of prediction errors.

The importance of prediction or expectation in the functioning of the mind is appreciated at least since the birth of psychology as a separate discipline. We continuously predict the world around us, because we learned the statistical regularities that govern it. It is often only when predictions go awry —when the sensory input does not match with the predictions we implicitly formed— that we become conscious of this incessant predictive activity of our brains. In the last decennia, a computational model called predictive coding emerged that attempts to formalize this matching process, hence explaining perceptual inference and learning. We recently applied predictive coding to autism spectrum disorders, a psychiatric syndrome characterized by difficulties in social, cognitive and affective processing. We argue that the weight (or ‘precision’) attributed to prediction errors in autism is unduly high and inflexible, rather than determined by context uncertainty. This may explain both their peculiar behavior in cognitive and perceptual tasks, and their affective-motivational troubles. The current research project attempts to systematically test the explanatory mechanisms of this new theory, using learning tasks, perceptual tasks and psychophysiological measures (e.g., pupil size).