I like the idea of linking the molecular level with the phenotypic level maybe via an algorithmic explanation. For example, these two new studies find separately (and by coincidence) that there are 4 genetically-informed dimensions of bipolar disorder and also autism (ASD). Maybe a similar study could be done to try to find subclasses of schizophrenia?
This is awesome Charles. Thanks for sharing. Will read and get back to you. I’m always amazed at how much of the literature you know! Always grateful for your input.
This strikes me as a very sensible and well -argued perspective. In an era when far too many articles (written by neuroscientists who ought to know better) are still lazily taking about “dopamine hits” your piece stands out a genuine contribution our thinking about psychological dysfunction. Nice work!
Its quiet an insightful take on how to explain to patients why they are feeling what they are feeling. Rather than going with medical terms, names and pathways, patients will be able to better understand the underlying mechanism of how their brain reacts.
Though I'm not an expert, I think this opens up possibilities of non-pharmacological methods in retraining your brain, based on the respective algorithmic explanation. I believe our brain is in a constant learning mode. Each action gives an input which gives a feedback and this loop continues, thereby adjusting its responses from physical pain to emotional turmoil.
This kind of self awareness within patients kind of create a momentary lag in usual response where they analyse and goes, " This is because of the following reason".
Also it might improve the initiative taken by the patients to receive the treatment or to better themselves.
Not going to lie, you did lose me a bit with the math 😅 but I still get where you’re coming from and find it a really interesting mode of explanation. Thanks for sharing and prompting me to think about this differently.
This is an excellent piece and pretty much covers why I was originally interested in pursuing neuroscience/psychiatry/neurology. In grad school, I read a lot of classical ethology on my own to see “mechanistic” explanations of behavior without any actual mechanisms— I now see those are like the algorithmic frameworks you outline. In the end, I think David Marr nailed the brain with its three levels of computation, and the molecular level you detail here would be at Marr’s implementation scale.
Thanks. Yes, we've written about Marr's levels academically. The more I think about it, the more I'm not actually sure that it always works as cleanly as we would like. I think it's partly because we don't have as crisp of an understanding of the neural code that the line between algorithm, computation and implementation can be a bit blurry-- but your point generally is well taken and it's helpful to at least attempt to draw distinctions.
It might be worth taking some notice of the discussion in philosophy of science (starting in the 1970s) of 'levels of explanation' and various models of 'why-questions' in science. (Why did the automobile accident occur? For some purposes we say "the driver of car A was inattentive" and for other purposes 'the momentum of car A was thus-and-such while that of car B was...".)
I'm thinking especially of Bas van Fraassen's model of why-questions and Peter Lipton's account of 'contrastive explanation', both of which try to capture something about the interest-relative nature of questions (and their appropriate answers), among other things.
~ your “model” response of how the brain works is incorrect; two systems, yes - but the brain doesn’t fire on predictions and rewards the way LLM’s do.
Perhaps a redefinition of mental illness - what’s considered healthy, what isn’t? Intrinsic vs extrinsic motivations and internal reward systems based on unpredictability ~ spontaneity, for example, the big ideas that so often find their way into our consciousness that we feel need exploration and explanation, to what end?
Misinformation might actually be the biggest threat to psychiatric conditions - especially if it’s being used to inform what’s needed in the field [it might be better if you were open about how this content was created…and why…especially if input is being asked for]
Assumed intelligence without omniscient refence puts your work at risk.
I don’t believe you’re asking for an explanation of what you consider illness; if anything ~ you’re asking for an algorithm that explains what it means to be human…
I like the idea of linking the molecular level with the phenotypic level maybe via an algorithmic explanation. For example, these two new studies find separately (and by coincidence) that there are 4 genetically-informed dimensions of bipolar disorder and also autism (ASD). Maybe a similar study could be done to try to find subclasses of schizophrenia?
https://pubmed.ncbi.nlm.nih.gov/40666370/
https://www.nature.com/articles/s41588-025-02224-z
This is awesome Charles. Thanks for sharing. Will read and get back to you. I’m always amazed at how much of the literature you know! Always grateful for your input.
This strikes me as a very sensible and well -argued perspective. In an era when far too many articles (written by neuroscientists who ought to know better) are still lazily taking about “dopamine hits” your piece stands out a genuine contribution our thinking about psychological dysfunction. Nice work!
Thank you! It’s very kind of you to say this. I am working on a follow up and look forward to hearing more of your thoughts.
Its quiet an insightful take on how to explain to patients why they are feeling what they are feeling. Rather than going with medical terms, names and pathways, patients will be able to better understand the underlying mechanism of how their brain reacts.
Though I'm not an expert, I think this opens up possibilities of non-pharmacological methods in retraining your brain, based on the respective algorithmic explanation. I believe our brain is in a constant learning mode. Each action gives an input which gives a feedback and this loop continues, thereby adjusting its responses from physical pain to emotional turmoil.
This kind of self awareness within patients kind of create a momentary lag in usual response where they analyse and goes, " This is because of the following reason".
Also it might improve the initiative taken by the patients to receive the treatment or to better themselves.
Exactly right! Medications and therapy are working on the same substrate. So good models should be able to help us optimize each and combine them
Not going to lie, you did lose me a bit with the math 😅 but I still get where you’re coming from and find it a really interesting mode of explanation. Thanks for sharing and prompting me to think about this differently.
Thank you! The math is there for people who benefit from it— totally fine to skip :)
This is an excellent piece and pretty much covers why I was originally interested in pursuing neuroscience/psychiatry/neurology. In grad school, I read a lot of classical ethology on my own to see “mechanistic” explanations of behavior without any actual mechanisms— I now see those are like the algorithmic frameworks you outline. In the end, I think David Marr nailed the brain with its three levels of computation, and the molecular level you detail here would be at Marr’s implementation scale.
Thanks. Yes, we've written about Marr's levels academically. The more I think about it, the more I'm not actually sure that it always works as cleanly as we would like. I think it's partly because we don't have as crisp of an understanding of the neural code that the line between algorithm, computation and implementation can be a bit blurry-- but your point generally is well taken and it's helpful to at least attempt to draw distinctions.
Good point. As you also mention with Shannon, information is about reducing uncertainty, so explanations can only go so far when that line is blurred.
Exactly! All I know is that depression is (almost certainly) not low serotonin and schizophrenia is not high dopamine :)
It might be worth taking some notice of the discussion in philosophy of science (starting in the 1970s) of 'levels of explanation' and various models of 'why-questions' in science. (Why did the automobile accident occur? For some purposes we say "the driver of car A was inattentive" and for other purposes 'the momentum of car A was thus-and-such while that of car B was...".)
I'm thinking especially of Bas van Fraassen's model of why-questions and Peter Lipton's account of 'contrastive explanation', both of which try to capture something about the interest-relative nature of questions (and their appropriate answers), among other things.
~ your “model” response of how the brain works is incorrect; two systems, yes - but the brain doesn’t fire on predictions and rewards the way LLM’s do.
Perhaps a redefinition of mental illness - what’s considered healthy, what isn’t? Intrinsic vs extrinsic motivations and internal reward systems based on unpredictability ~ spontaneity, for example, the big ideas that so often find their way into our consciousness that we feel need exploration and explanation, to what end?
Misinformation might actually be the biggest threat to psychiatric conditions - especially if it’s being used to inform what’s needed in the field [it might be better if you were open about how this content was created…and why…especially if input is being asked for]
Assumed intelligence without omniscient refence puts your work at risk.
I don’t believe you’re asking for an explanation of what you consider illness; if anything ~ you’re asking for an algorithm that explains what it means to be human…