Thanks to Sarah H. and the people at her house for help understanding this paper]
The predictive coding theory argues that the brain uses Bayesian calculations to make sense of the noisy and complex world around it. It relies heavily on priors (assumptions about what the world must be like given what it already knows) to construct models of the world, sampling only enough sense-data to double-check its models and update them when they fail. This has been a fruitful way to look at topics from depression to autism to sensory deprivation. Now, in Relaxed Beliefs Under Psychedelics And The Anarchic Brain: Toward A Unified Model Of The Brain Action Of Psychedelics, Karl Friston and Robin Carhart-Harris try to use predictive coding to explain the effects of psychedelic drugs. Then they use their theory to argue that psychedelic therapy may be helpful for “most, if not all” mental illnesses.
Priors are unconscious assumptions about reality that the brain uses to construct models. They can range all the way from basic truths like “solid objects don’t randomly disappear”, to useful rules-of-thumb like “most get-rich-quick schemes are scams”, to emotional hangups like “I am a failure”, to unfair stereotypes like “Italians are lazy”. Without any priors, the world would fail to make sense at all, turning into an endless succession of special cases without any common lessons. But if priors become too strong, a person can become closed-minded and stubborn, refusing to admit evidence that contradicts their views.
F&CH argue that psychedelics “relax” priors, giving them less power to shape experience. Part of their argument is neuropharmacologic: most psychedelics are known to work through the 5-HT2A receptor. These receptors are most common in the cortex, the default mode network, and other areas at the “top” of a brain hierarchy going from low-level sensations to high-level cognitions. The 5-HT2A receptors seem to strengthen or activate these high-level areas in some way. So:
