psychedelicward
Greenlighter
- Joined
- Jun 21, 2016
- Messages
- 4
I'm interested in "Bayesian predictive processing" theories of brain function. See Andy Clark for an introduction to the predictive processing framework:
http://www.fil.ion.ucl.ac.uk/~karl/Whatever next.pdf
In particular, I'm interested in recent "drug models of psychosis" (involving ketamine, LSD, etc.) that are couched in terms of this predictive-brain framework:
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2755113/pdf/213_2009_Article_1561.pdf
I find this story about psychedelics and dissociatives pretty compelling, spanning as it does pharmacology, neuroanatomy, systems neuroscience, cognition, and phenomenology.
However, this work is framed entirely within the language of research on clinical psychosis. It needs to be expanded. We need to put together a translational story where this work on Bayesian drug models of psychosis is repurposed as a framework for a more general psychedelic neuroscience and psychedelic phenomenology. Call it Bayesian psychedelic neuroscience or Bayesian psychonautics. Of course the Bayesian-brain story is still a little hand-wavy, but can't we imagine this work as providing a level of computational and mechanistic understanding that bridges the current gap between molecules and psychedelic phenomenology?
On the Bayesian predictive processing view, the brain is an expectation machine. The whole point of the brain is to continuously generate new top-down expectations based on calculated mismatches between prior top-down expectations and incoming sensory stimulation. The effects of psychedelic and dissociative drugs, on this view, result from changing the brain's confidence in either the prior expectations, the calculated mismatches (the error signals), or both. For example, the paper above argues that given the pharmacology of ketamine (NMDA-R antagonist) and a plausible, simplified model of the computational anatomy of the brain, ketamine both turns down the confidence in prior expectations (prior beliefs) and ramps up bottom up prediction errors. The result is altered belief formation, which the authors argue underlies "delusions". A psychedelic shaman or psychonaut or psychedelic therapist might frame the drug induced change in nicer terms. Indeed, I think what needs to happen is that researchers interested in positive psychedelic science should take advantage of this Bayesian framework (and all the prior legwork that these drug-models-of-psychosis researchers have already done), and build a general Bayesian neuroscience of psychedelic drugs.
Again, how do psychedelics work on this account? Within this Bayesian framework, we're not, as Huxley thought, using psychedelics to "open the doors of perception"; rather, these drugs are being used to "take Bayes for a walk". Jacktripper put it this way: "Turn down the precision on those top-down priors. Crank up the gain on those bottum-up prediction errors. Psychedelics and dissociatives work by tweaking the mechanisms of perceptual and cognitive control. Perception is controlled hallucination. Belief is controlled delusion. Hold on to your sanity, mate!" Psychedelics and dissociatives turn the knobs on the mechanisms controlling these probability distributions:
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3667557/figure/F1/
Ultimately, we need a "computational anatomy of psychedelics" that looks something like this paper:
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3667557/
Such a story would put us on the road to an integrated account of psychedelics, starting at the levels of the pharmacology and neuroanatomy and moving all the way up through systems neuroscience, cognition, and phenomenology. Further, such a story could clarify how, exactly, psychedelics might be used (as experimental tools) for probing the mechanisms of cognition and perception more generally.
Thoughts? Anyone out there want to write an academic review paper about this?
http://www.fil.ion.ucl.ac.uk/~karl/Whatever next.pdf
In particular, I'm interested in recent "drug models of psychosis" (involving ketamine, LSD, etc.) that are couched in terms of this predictive-brain framework:
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2755113/pdf/213_2009_Article_1561.pdf
I find this story about psychedelics and dissociatives pretty compelling, spanning as it does pharmacology, neuroanatomy, systems neuroscience, cognition, and phenomenology.
However, this work is framed entirely within the language of research on clinical psychosis. It needs to be expanded. We need to put together a translational story where this work on Bayesian drug models of psychosis is repurposed as a framework for a more general psychedelic neuroscience and psychedelic phenomenology. Call it Bayesian psychedelic neuroscience or Bayesian psychonautics. Of course the Bayesian-brain story is still a little hand-wavy, but can't we imagine this work as providing a level of computational and mechanistic understanding that bridges the current gap between molecules and psychedelic phenomenology?
On the Bayesian predictive processing view, the brain is an expectation machine. The whole point of the brain is to continuously generate new top-down expectations based on calculated mismatches between prior top-down expectations and incoming sensory stimulation. The effects of psychedelic and dissociative drugs, on this view, result from changing the brain's confidence in either the prior expectations, the calculated mismatches (the error signals), or both. For example, the paper above argues that given the pharmacology of ketamine (NMDA-R antagonist) and a plausible, simplified model of the computational anatomy of the brain, ketamine both turns down the confidence in prior expectations (prior beliefs) and ramps up bottom up prediction errors. The result is altered belief formation, which the authors argue underlies "delusions". A psychedelic shaman or psychonaut or psychedelic therapist might frame the drug induced change in nicer terms. Indeed, I think what needs to happen is that researchers interested in positive psychedelic science should take advantage of this Bayesian framework (and all the prior legwork that these drug-models-of-psychosis researchers have already done), and build a general Bayesian neuroscience of psychedelic drugs.
Again, how do psychedelics work on this account? Within this Bayesian framework, we're not, as Huxley thought, using psychedelics to "open the doors of perception"; rather, these drugs are being used to "take Bayes for a walk". Jacktripper put it this way: "Turn down the precision on those top-down priors. Crank up the gain on those bottum-up prediction errors. Psychedelics and dissociatives work by tweaking the mechanisms of perceptual and cognitive control. Perception is controlled hallucination. Belief is controlled delusion. Hold on to your sanity, mate!" Psychedelics and dissociatives turn the knobs on the mechanisms controlling these probability distributions:
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3667557/figure/F1/
Ultimately, we need a "computational anatomy of psychedelics" that looks something like this paper:
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3667557/
Such a story would put us on the road to an integrated account of psychedelics, starting at the levels of the pharmacology and neuroanatomy and moving all the way up through systems neuroscience, cognition, and phenomenology. Further, such a story could clarify how, exactly, psychedelics might be used (as experimental tools) for probing the mechanisms of cognition and perception more generally.
Thoughts? Anyone out there want to write an academic review paper about this?
Last edited:
