Sander Van de Cruys, Kris Evers, Ruth Van der Hallen, Lien Van Eylen, Bart Boets, Lee de-Wit, Johan Wagemans
5 min
The human brain is increasingly understood as a prediction engine that constantly generates models of the world to anticipate sensory input. When incoming information deviates from these predictions, the brain computes a prediction error. In a dynamic, noisy environment, an adaptive system must decide whether to update its internal model based on these errors or to ignore them as spurious noise. This process, known as meta-learning, relies on estimating the precision (or reliability) of prediction errors. The authors argue that in ASD, this mechanism is impaired, resulting in an inflexibly high precision assigned to prediction errors regardless of context.
The authors propose the High, Inflexible Precision of Prediction Errors in Autism (HIPPEA) hypothesis. Because individuals with ASD cannot effectively discount noisy or uninformative prediction errors, they are forced to treat every minor deviation as a significant event requiring a model update. This leads to a form of overfitting: the brain creates highly specific, rigid predictions that work well in static, repetitive environments but fail to generalize to the dynamic, complex, and often ambiguous nature of real-world social interactions. This explains why individuals with ASD may excel in rote tasks but struggle with tasks requiring flexible, abstract, or context-dependent reasoning.
This core deficit has cascading effects on development. Because the system is constantly overwhelmed by unignorable prediction errors, the natural exploratory drive—which typically targets information with moderate, learnable complexity—is disrupted. Instead, individuals may withdraw into repetitive, predictable behaviors to minimize the influx of unmanageable errors. In the social domain, where cues are inherently noisy and lack one-to-one mappings, the inability to filter irrelevant details leads to sensory overload and difficulties in forming social scripts. What appears as a social deficit is, in this view, a logical consequence of a general information-processing imbalance that makes social environments uniquely taxing.
This framework moves beyond descriptive symptom clusters to provide a unified, mechanistic explanation for the heterogeneity of ASD. It reconciles the seemingly contradictory findings of superior performance in some tasks (e.g., visual search) and deficits in others (e.g., social cognition) by highlighting the role of context and task complexity. By framing ASD as a disorder of precision-weighting, the theory offers a clear path for future research to test specific neurobiological markers and suggests that interventions should focus on scaffolding the transition from simple to complex, naturalistic environments.
There have been numerous attempts to explain the enigma of autism, but existing neurocognitive theories often provide merely a refined description of 1 cluster of symptoms. Here we argue that deficits in executive functioning, theory of mind, and central coherence can all be understood as the consequence of a core deficit in the flexibility with which people with autism spectrum disorder can process violations to their expectations. More formally we argue that the human mind processes information by making and testing predictions and that the errors resulting from violations to these predictions are given a uniform, inflexibly high weight in autism spectrum disorder. The complex, fluctuating nature of regularities in the world and the stochastic and noisy biological system through which people experience it require that, in the real world, people not only learn from their errors but also need to (meta-)learn to sometimes ignore errors. Especially when situations (e.g., social) or stimuli (e.g., faces) become too complex or dynamic, people need to tolerate a certain degree of error in order to develop a more abstract level of representation. Starting from an inability to flexibly process prediction errors, a number of seemingly core deficits become logically secondary symptoms. Moreover, an insistence on sameness or the acting out of stereotyped and repetitive behaviors can be understood as attempts to provide a reassuring sense of predictive success in a world otherwise filled with error. (PsycINFO Database Record (c) 2014 APA, all rights reserved).
Sam: That's a meaningful reframe. But where does the actual evidence sit? This is a theoretical framework—what empirical traction does it have?
Alex: The authors ground it in the distinction between what they call expected and unexpected uncertainty. Expected uncertainty is just inherent task noise—the kind of variability you know is there and can account for. Unexpected uncertainty is when the predictive value of a cue changes—when the rules of the environment shift. The evidence suggests individuals with ASD handle the first type reasonably well but struggle specifically with the second.
Sam: That's a meaningful dissociation. It's not that they can't learn—it's that they struggle to update their models when the environment changes its structure.
Alex: Precisely, and that's where the meta-learning argument comes in. The framework predicts that the difficulty isn't in forming predictions, but in deciding which cues are worth tracking when the environment is ambiguous about that. In noisy contexts, you need to autonomously infer what's signal and what's noise—and that inference is what's impaired.
Sam: How well does this hold up across the heterogeneity of ASD presentations? That's always the hard question with unifying accounts.
Alex: It's the paper's most significant limitation, and the authors acknowledge it. ASD is a spectrum with substantial phenotypic variance, and a single-parameter account—even a well-specified one—is going to struggle with the tails of that distribution. The framework handles the core triad of social difficulty, sensory sensitivity, and repetitive behavior reasonably well. Where it's more strained is in explaining the variability within those domains across individuals.
Sam: And I'd imagine a careful referee would also push on the directionality question. High inflexible precision is the proposed mechanism—but how do you distinguish that from downstream effects of other processes?
Alex: That's exactly the right pressure point. The framework is largely post-hoc in its current form—it organizes existing findings coherently, but the causal architecture isn't fully pinned down. The authors gesture toward neural substrates, particularly around neuromodulatory systems that regulate precision weighting, but the mechanistic story at the implementation level is still underdetermined. What the framework does well is generate testable predictions—specifically around tasks that parametrically manipulate expected versus unexpected uncertainty—and that's where the empirical work needs to go next.
Sam: So the value right now is less as a settled account and more as a generative scaffold for experimental design.
Alex: That's a fair characterization. It's a theoretically coherent unification that brings predictive processing into contact with a large and somewhat fragmented empirical literature. Whether the single-parameter story survives contact with more targeted experiments is the open question. But as a framework for asking sharper questions, it's doing real work.
Sam: That's a useful place to land. Thanks for walking through it.
Alex: Thanks for listening to ResearchPod.