Mazviita ChirimuutaGuest
Kensy CooperriderHostSo you would go on, as you already alluded to, you would go on to write this influential book published in 2024 on abstraction and simplification in neuroscience.
And it sounds like some of the issues you discussed there did sort of take root or were informed by this work that you'd done in vision science.
I mean, I've always thought that abstraction is inherent in any sort of scientific explanation, right? What do you think, if anything, is special or distinctive about abstraction in the case of neuroscience?

So I totally agree with you that abstraction is fundamental to the scientific method as we know it.

We use compact representations just to memorize things, just to go about our world.

And science is in some ways an exaggerated form of the kind of simplifications and abstractions that people use

But at the same time, you can ask this question, are some objects of investigation more amenable to the kinds of abstractions that scientists rely on? So the biggest success cases for these highly simplifying methods are are in fundamental physics and also newtonian physics where you're dealing with a set of relationships amongst objects which are localizable the objects of investigation are quite stable in other words you're dealing with quite simple objects moving to biology there's generally less of a payoff of the kind of methods that work well in physics, like developing formal theories that apply universally to the objects in question, as much less of that seems to work in biology.

And neuroscience in particular is a biological system with just a really high degree of complexity.

You also have component parts whose job, the job that they have to do is to modify themselves in response to experience.

That's what learning is and all animals undergo lifelong learning, constantly having to adapt to environmental challenges.

So you don't have a stable target and you can't just average out all of the complexity without also losing like functionally relevant detail.
When I think about these issues, this is probably an annoying cliche at this point for you, but I think about the classic Borges story and the idea that a one-to-one scale map of the kingdom is not actually useful, right? Like you need to simplify this down.
So I guess, to me, it raises all these questions of what is a useful level of abstraction if we grant that some abstraction is sort of a necessary evil, if you like.

I mean, we can index this to the stated aims of neuroscience as a research field.

So a lot of the motivation for funding in neuroscience is biomedical research.

I mean, psychiatric disease, neurodegeneration, and neurological conditions like Parkinson's, they're devastating conditions.

And so much sort of research is poured into the brain precisely in order to find out the things that we could use to intervene biomedically And if the kinds of abstractions that we're relying on are so heavily oversimplifying that we're not making progress in biomedical research, well, that's one indication that too much complexity has been stripped out and the actual processes like disease etiology or contextual effects that would actually be relevant targets of intervention are just not being captured by our theories and models.

So the neuroscientist Nicole Rust has a really nice book precisely on this issue.

It's called Elusive Cures, about how oversimplifying the brain has actually kind of inhibited progress in translational neuroscience.
So you would go on, as you already alluded to, you would go on to write this influential book published in 2024 on abstraction and simplification in neuroscience.
And it sounds like some of the issues you discussed there did sort of take root or were informed by this work that you'd done in vision science.
I mean, I've always thought that abstraction is inherent in any sort of scientific explanation, right? What do you think, if anything, is special or distinctive about abstraction in the case of neuroscience?

So I totally agree with you that abstraction is fundamental to the scientific method as we know it.

We use compact representations just to memorize things, just to go about our world.

And science is in some ways an exaggerated form of the kind of simplifications and abstractions that people use

But at the same time, you can ask this question, are some objects of investigation more amenable to the kinds of abstractions that scientists rely on? So the biggest success cases for these highly simplifying methods are are in fundamental physics and also newtonian physics where you're dealing with a set of relationships amongst objects which are localizable the objects of investigation are quite stable in other words you're dealing with quite simple objects moving to biology there's generally less of a payoff of the kind of methods that work well in physics, like developing formal theories that apply universally to the objects in question, as much less of that seems to work in biology.

And neuroscience in particular is a biological system with just a really high degree of complexity.

You also have component parts whose job, the job that they have to do is to modify themselves in response to experience.

That's what learning is and all animals undergo lifelong learning, constantly having to adapt to environmental challenges.

So you don't have a stable target and you can't just average out all of the complexity without also losing like functionally relevant detail.
When I think about these issues, this is probably an annoying cliche at this point for you, but I think about the classic Borges story and the idea that a one-to-one scale map of the kingdom is not actually useful, right? Like you need to simplify this down.
So I guess, to me, it raises all these questions of what is a useful level of abstraction if we grant that some abstraction is sort of a necessary evil, if you like.

I mean, we can index this to the stated aims of neuroscience as a research field.

So a lot of the motivation for funding in neuroscience is biomedical research.

I mean, psychiatric disease, neurodegeneration, and neurological conditions like Parkinson's, they're devastating conditions.

And so much sort of research is poured into the brain precisely in order to find out the things that we could use to intervene biomedically And if the kinds of abstractions that we're relying on are so heavily oversimplifying that we're not making progress in biomedical research, well, that's one indication that too much complexity has been stripped out and the actual processes like disease etiology or contextual effects that would actually be relevant targets of intervention are just not being captured by our theories and models.

So the neuroscientist Nicole Rust has a really nice book precisely on this issue.

It's called Elusive Cures, about how oversimplifying the brain has actually kind of inhibited progress in translational neuroscience.
The rest of this transcript — segmented and speaker-labeled, so you land on the exact moment something was said
Search every transcript — by keyword, by phrase, or by meaning, across every show Radar indexes
Trends — what is surging across podcasts, measured against its own baseline
Alerts — when a name you follow appears in a newly indexed episode
No account is needed to search Radar.