The brain: Different expert systems propose strategies for action, keeping track of the precision of the predictions within each system, exerting control over many different expert systems simultaneously to produce sophisticated behavior

Why and how the brain weights contributions from a mixture of experts. John P. O’Doherty et al. Neuroscience & Biobehavioral Reviews, January 11 2021. https://doi.org/10.1016/j.neubiorev.2020.10.022

Rolf Degen's take: https://twitter.com/DegenRolf/status/1348683591333670914

Highlights

• The brain can be thought of as a “Mixture of Experts” in which different expert systems propose strategies for action.

• This is accomplished by keeping track of the precision of the predictions within each system, and by allocating control over behavior in a manner that depends on the relative reliability of those predictions.

• This reliability-based control mechanism is domain general, exerting control over many different expert systems simultaneously in order to produce sophisticated behavior.

Abstract: It has long been suggested that human behavior reflects the contributions of multiple systems that cooperate or compete for behavioral control. Here we propose that the brain acts as a “Mixture of Experts” in which different expert systems propose strategies for action. It will be argued that the brain determines which experts should control behavior at any one moment in time by keeping track of the reliability of the predictions within each system, and by allocating control over behavior in a manner that depends on the relative reliabilities across experts. fMRI and neurostimulation studies suggest a specific contribution of the anterior prefrontal cortex in this process. Further, such a mechanism also takes into consideration the complexity of the expert, favoring simpler over more cognitively complex experts. Results from the study of different expert systems in both experiential and social learning domains hint at the possibility that this reliability-based control mechanism is domain general, exerting control over many different expert systems simultaneously in order to produce sophisticated behavior.

Keywords: cognitive controlPrefrontal cortexbasal gangliaTheoretical neuroscienceDecision-making

Conclusion

Here we outline a framework for conceptualizing the contribution of multiple systems to behavioral control in the human brain. We suggest that the brain utilizes a framework loosely analogous to the mixture of experts in machine learning, in which a prefrontal-based manager (which we hypothesize specifically involves the ventrolateral prefrontal cortex), reads out the reliability of the predictions by each of the constituent experts, and uses these predictions to allocate control over behavior to the experts in a manner that is proportional to the relative precision or uncertainties in their predictions. We suggest that this reliability-based arbitration process between experts is both necessary and sufficient for the efficient allocation of control between systems, as this approach takes into account not only the accuracy and hence the average expected value of the actions nominated by each expert, but also implicitly takes into account the cognitive costs and cognitive constraints. The interaction between systems that makes up the experts is we suggest, better conceived of as one of polling the advice from different systems that each have different relevant expertise that can and should be respected owing to differences in the nature of the information that is being processed, and in the algorithmic transformations that are performed on that information. These experts should be listened to as a collective, because they provide the right mixture of opinions needed to act in the world effectively.