Full episode transcript (beware of typos!) below:
Nick Jikomes
Professor Aaron Gruber, thank you for joining me.
Aaron Gruber 4:07
Hi, nice to be here.
Nick Jikomes 4:09
Can you start off by just telling everyone what you do and what you're interested in?
Aaron Gruber 4:14
Yeah, so my background is maybe a little bit different. I started in
an engineering discipline, went through sort of a computational route, and then ended up doing experiment experimental work for the past 20 years or so. And so, what I do is a lot of behavior, some pharmacology so getting drugs, this is in rodents.
And using technologies to record neurons as animals are doing things, and then using computational tools to try to understand how the brain is representing things and processing information.
Nick Jikomes 4:53
I see this you're kind of an engineer that became a neuroscientist. Yeah, that's a that's pretty Common, I think, given all of the engineering of the gadgets involved,
Aaron Gruber 5:04
yeah, this now, yeah, back in the 90s, it was a little more rare. But now, you know, there's a lot of cross fertilization between technology and biology. It's really amazing how things how quickly things are going.
Nick Jikomes 5:18
So a lot of what you study has to do with, you know, some pretty big questions in neuroscience, behavior, decision making cognition, you know, learning how all of these things interact in different ways. There's a lot of different directions we can go. One of the things I wanted to talk about first, was dopamine. And the reason I wanted to start there was it's a neurotransmitter in the brain that many people have heard of, probably most people who are listening have at least heard of it. And, you know, it's pretty, it's pretty well known. And they're sort of a cartoon version of dopamine that most people have in their mind. Right. So the pop science characterization of dopamine is it's the pleasure molecule, quote, unquote, things that feel good cost dopamine release, this makes us want to do them again, drugs of abuse, you know, hijack the reward system by causing a lot of dopamine release, which is what can lead to addiction. And so can you just talk about what dopamine is? And how accurate or complete is that sort of story? We often hear about it?
Aaron Gruber 6:19
Yeah, so you did a nice encapsulation. And it's, I think, it shows kind of the power of guess what we could call a meme, a conceptual meme. So the the person that originally proposed dopamine as a as a pleasure, mediator of pleasure, walk that story back a few decades ago, but once the genies out of the bottle, it's, it's hard to revert that. So most people in the scientific context that study it has long realized that dopamine itself, it's not pleasurable. And one of the reasons we know this is that one of the precursors for dopamine, called L dopa is the first line treatment for Parkinson's disease in which the dopamine neurons die off. And when you give someone l dopa, the essentially that crosses the blood brain barrier, that increases the amount of dopamine that the surviving dopamine cells release, and kind of normalizes the motoric and other deficits. It's not pleasurable at all. And so physicians, neurologists have to titrate benefit with, you know, how much will they actually tolerate, because it makes people irritable, makes them feel unwell. And so if dopamine really was the pleasure drug, then you know, you could go to the seedy part of town and, you know, buy some old job or whatever, and it would make you feel good that, you know, that doesn't happen. So it really seems that it's the opioid system is probably the best candidate for pleasure. But there's a lot of there's a lot of crossover. So you know, the fact of increasing dopamine can cause, you know, downstream release of other things that do have these mood elevating progress. So that's, you know, this is sort of a testament to how networked things are and how much crosstalk there is. So that that's the pleasure part. Now, there's an entire nother story with with dopamine, that is fantastic and beautiful. It integrates biology and theory and computer science. And I guess, you know, what I often tell my undergraduate students is that in a way, I kind of broke my heart, because it was so beautiful. And so this this story is this that, that that the signaling of dopamine provides what's called a reward prediction error signal. And that means that things that are unexpectedly good increase dopamine, you get a little, like a one second, little bit less than a one second kind of burst of dopamine. When things come just as you expected, it doesn't change. And when things are worse than expected, you get this dip. And so this is this linkage with dopamine as a as a learning signal. And so it turns out that if you, if you have this kind of reward prediction error signal, you can learn some really amazing things. And so there's a company maybe you've heard of called DeepMind, that was, you know, bought by Google for half a billion dollars or something. And so, this spawned this entire field, what's called Deep reinforcement learning. So this idea of reinforcement learning is all predicated on this idea that if you have this word prediction error signal, you can learn all kinds of stuff. And so this was actually shown in the early 90s. From Bartow, and he's retired and rich Sutton is at the University of Alberta. So if you do this, you can train a computer to play chess to do robotics applications. Like you can do you know, a really a surprising and amazing array of things. Meaning you can solve out rhythmically, these hard problems just by providing this word prediction error signal, meaning that when it does what you want, you give it a you know you give it a little a virtual dopamine hit. And when it does what you didn't, then you kind of take that away, you pause it. And so just by giving that you can learn lots of stuff. And so that came back to neuroscience. I mean, the original idea came in neuroscience several decades ago, but then it was only sort of formalized starting sort of in the 90s. And then that kind of fed back from the computer science people back in neuroscience and said, Well, hey, you know, if algorithms can use this word, prediction error signal to solve all these problems, then maybe that's what dopamine is doing in the brain, that it's a way you know, to teach a fundamental teaching signal in the brain to solve all kinds of problems.
Nick Jikomes 10:38
I see. So computer scientists had figured out that if you had this kind of signal, it was very powerful. You could you could write computer code that was able to do impressive stuff using this kind of signal. And then in animals, neuroscientists discovered that actually, there are dopamine neurons of turned out that instantiate this type of signal in, in the wetware of the brain. Exactly,
Aaron Gruber 11:02
yes. So yeah, Wolfram Schultz is credits. a neuroscientist working in in Europe was credited with doing recordings of dopamine neurons and a monkey and very simple experiment, you unexpectedly give them a reason, and no dopamine neurons would would fire up. If you play this game a few times, if you approach them and you haven't raised your hand, and you open your hand, in the beginning, they didn't expect it, and you would get the dopamine burst. But after a while, meaning 1520 times, the monkey would come to expect it. And now the dopamine neurons didn't fire anymore, even though we got the reason. And then, you know, scientists being what we are, we have to manipulate things. Now you come with your with your fist, and no reason it monkey expects the reason you open it, it's not there. Now there have been neurons pause. So normally, they're ticking along about one time per second. And when they activate, they, you know, they have to like, though generate these bursts of five or six action potentials very quickly, and that causes the release of dopamine. And then the pause just means that you've got a second or so without that action potential in it. And so that that was kind of the first you know, when people saw that they said, Oh, you know, holy cow, that looks a lot like this word prediction error signal?
Nick Jikomes 12:18
And is that go just in the direction of positive valence? Do they do this in response to the presence or absence of rewards? Or does it go in the opposite direction as well?
Aaron Gruber 12:28
Yeah, fantastic question. So uh, for quite a while the the emphasis was really on the positive part. But then when it turns out, so So that's, that's sort of the the beauty of everything right? That that it's this relatively simple thing, this word prediction error signal can solve all kinds of hard problems, without having to teach an animal or computer like how to play backgammon, or chess, just telling them you lost or your one, right at the end of it, right. And the the algorithm can kind of figure everything else out. But then, you know, as you pointed out, then, data that came in over time was, was inconsistent with that fundamental idea. And so one of those pieces of data are that some dopamine neurons, it turns out, respond to aversive things and to cues that proceed or first offense, one thing I forgot to mention is, it's not only just the cue, so like, you know, when the, you know, when the raising comes out, but if you give it a tone or something ahead of time, so if you go, you know, be and then two seconds later you present the reason, what happens is that the dopamine neurons, then they start to respond to the queues that proceeded, and they'll actually propagate backwards in time. The reason that becomes important is, for problems like addiction and overeating. And, you know, this is it ties into advertising, that, you know, the reason that brands become very, you know, it's thought the reason that brands become very potent, is that if you've had a positive association, you know, by having, you know, the symbols and the logos and icons, right, they want to sort of drive that system. And, you know, I'd say a wider view of what dopamine and associated systems do is that they capture attention. And that what it looks like is that they're really there to detect what's called salience, which, which just is a term that means something is important behaviorally to you. So for an animal that would be positive things like awards, but also things that might precede a shock or a startling tone or whatever. But, you know, the idea is that if you're driving down the street in a city, you really love McDonald's, you're driving on the highway, even if you're not necessarily hungry, all of a sudden, you see the Golden Arches sign, and that trigger is like, Oh, hey, you know, should we get McDonald's? Right that that seems to tap into that, that system, this sort of salience system of that information gets in your brain And then you decide whether you're going to act on it or not.
Nick Jikomes 15:02
I see. So is it fair to say that it's much better to think about dopamine as detecting salience and being involved with the motivation to act on what you're seeing?
Aaron Gruber 15:13
Well, you know, we think so that's kind of my opinion. I mean, other people are much more still in the reward part. You know, that's the beauty of science is that different people can, you know, be led by different hypotheses? And, you know, we don't know, I mean, the truth is probably going to be somewhere between that, yes, it has something to do with with value that we haven't talked about yet. But, you know, there's other data indicates it's doing some other things, too. So, you know, one story that humans repeatedly discover about biology is that things get reused. And so in the beginning, it might very well been only for value, but then, you know, as time has come along, that same system now has been co opted to do you know, other functions as well. And we see that in lots of places.
Nick Jikomes 16:02
Yeah, I'm wondering if you could unpack the idea from a neuroscience perspective of value a little bit more, some of the things that come to mind for me are, you know, when you think about these dopamine neurons, and this reward prediction error signal, you know, we know, computationally this type of signals very interesting, it can be used for learning, we know that some of these neurons are going to respond to good things, or bad things. But also, in some sense, the information of the goodness and badness was already there, right? In these neurons, like, if they're, if you're giving an animal a treat, or you're giving a shock, in some sense, that already knows that it's good or bad. So how does, you know learning the goodness or badness of something tied to sort of instinctive or innately innately present notions of goodness or badness that are already there?
Aaron Gruber 16:50
Yeah, so that that's a really good question. And and maybe we can, I'd like to cycle back to that that question a little bit once we cover a little more territory. But let me you know, let me unpack sort of a canonical reinforcement learning, which involves value, and then maybe later, we can talk about some of these other things. So what you're saying is, is very important that animals have to have these neat systems to feel pain to feel pleasure, as I mentioned before, you know, probably involves other things like the opiates or opioid receptors and other things. So you have to have that. So so, you know, it looks like dopamine really isn't involved in adjudicating, you know, I really liked that, or I didn't. But let's we'll talk about value. So, again, the canonical story is imagine you have a choice between, you know, two things, a and b. And you don't know anything about ARB, right, you're completely naive, and you go, and so it's their food items, we'll just we'll just say, to make this easy, you go and you taste a, you had some expectation going into it, even though you hadn't tried it before. And this is this is something you actually need for reinforcement learning to work, you need a bit of optimism, because if you thought everything was terrible, you've never tried it in the first place. So you have to say, well, you know, there's a, maybe it'll taste good, right? I'll try. So you taste it. And there's some system that gives you some hedonic pleasure, right, it's sweet or fatty, or whatever. And so you, you that that somehow in the brain is translated, or I guess, triggers dopamine response. So if it was better than you were expecting, then you get this dopamine response. What happens then is that you presume that the sensory systems that were involved in seeing a that you know, somewhere in your brain, so you have the the immediate sensory, visual, olfactory, whatever. But then you have these Association cortices. And whether there's, they kind of assemble the sensory things into some internal representation of the thing. And then the thought is then the amount of dopamine that that when dopamine gets released, that kind of stamps it with an amount of value, meaning that you have this pattern of activity that's representing the sensory elements, but that some of the neurons also that are involved in this pattern are going to signal the value of it. And so when dopamine increases, those neurons, the strength of their synapses, and so they they give an encoding of the value. Now, say you go to be you take B and it's a sort of man, you're it's not not as not great, not as great as a, so you have a different set of cells that were on for B. Say, you know, it's kind of exactly what you expected. So the dopamine cells didn't turn on those value cells then don't increase their representation of value. Now, what now when you have a choice that you come back, so imagine, you know, you taste a and b a few times, each time you taste a, right your assessment value goes up, the value for b kind of stays where it was right each each time you take You get a little dopamine that increases incrementally devaluation until it reaches some state where the tastes and the hedonic pleasure you get from it matches your expectation, then you don't get any dopamine change anymore. And so it stabilizes at this level, okay? Now, when you have a choice later on, you're going to choose between a and b, right, you're at a party, they're out on the table, you know, so there's these Asian snacks that you've never seen before, and you go off, like there's a, I got to decide between a and b, well, what you can do is you can use the difference in valuation. And so if you learn that a had a higher value than b, then reinforcement learning, the choice element of it says that you should tend to pick the things that have higher value. And there's different ways you can do that you can exclusively pick the thing that has the highest value. And that's called a greedy system, or you can kind of do it probabilistically, meaning that the bigger the difference, the more likely it is you're going to pick a over b. So that's the that's the reinforcement learning story. And you can do lots of amazing things with that. And the trick is just how do you convert it? So donec thing to a dopamine to a value? That's a very active sort of research question.
Nick Jikomes 21:16
So how well would you say reinforcement learning captures what we actually observe in animals when you give them two choices.
Aaron Gruber 21:25
Okay, so you know, it's a beautiful theory, and it works really well in some contexts. So the the one in the literature has the colloquial name of an Armed Bandit. And so that's where you have multiple devices, and the One Armed Bandit is, is a slang term for a slot machine. And so what that connotes is that you have these different choice options or different machines. And when you press the button, you have a certain probability of getting a payout. When you do this in humans, and monkeys, and even rodents, you get behavior out, that's very consistent with reinforcement learning. So you get choices that kind of track the the probability of getting paid out, or you can also do it with reward value. The problem comes in when you start to do different types of experimental designs, then it doesn't work as well. And you know, that this is where I think the story, you know, at least in my development of a scientist, I had to step back because the, you know, from an engineering perspective, the reinforcement story is, right, it's beautiful, it's elegant, it's, it makes predictions that you can test. But then, you know, when you get into the actual biology of the brain, you know, I think you really have to think more evolutionarily. And you have to think, in terms of the environments that animals came from, and that the brain really evolved to solve those problems. And, you know, historically, there weren't really an armed bandits that weren't, you know, kind of these other things. And so, you know, we're kind of looking at and those kinds of designs or maybe some of the outliers. So it's not, you know, the term for that is ecological, so ecological would be the environment and conditions in which an animal species has been living for a long time. And so they've kind of adapted to these things.
Nick Jikomes 23:23
I want to unpack for people a little bit more how you think about what a choice is at the level of the brain, so that at a high level, it's intuitive, right? What choices we all make choices all the time, if you see two items of food on the table, a and b, we know what a choice is, we, we we sit there, we might contemplate it for a moment, and then we go, okay, A is better. For some reason, I'm going to pick that one. Now, as a neuroscientist, how do you think about localizing the, where the choice is happening? Or what the mechanisms are? Like, for example, you know, I can imagine that one way to think about it is, well, if I want to reach for A versus B, that requires two different motor programs to execute in the brain. And so maybe the choice is sort of whatever network or circuit is, is, you know, looking at both of those motor programs, and then committing to one of them and or inhibiting the other one, how do you think about choice at that kind of mechanistic brain level? And where do you actually look for it happening in the brain?
Aaron Gruber 24:26
So it's a fantastic question. It's a really deep question. You know, I thought, you know, going into all this when I was younger, I thought, well, you know, obviously, I know what the choices that that's easy. There's a psychological phenomenon that's referred to as the Kruger Dunning effect, if you're familiar with this, but this is that if you ask people, if you ask people to rank their competence in something, and you actually plot it against how much experience they have in that domain, people with little experience tend to overestimate how much competence they have that domain, but then as they gain more experience, they actually rank their competence lower. And then at some point, you know, hopefully have this inflection and you become the master. And then, you know, you can evaluate that you really do know what you're talking about, and you really have the space to talk about it. You know, a good example for me is that mountain biking, which is something I've picked up in the past couple of years, right, you know, I know how to ride a bike, it's just riding a bike down a Woods path, right? Well, no, there's a lot of skills that are different. And, you know, it's not until you get out there that you realize, oh, you know, whoa, crap. I'm, you know, I'm a newbie. And so anyway, going back to your, you know, the question you're asking, is very much an important one. And what I want to do is kind of go back a little bit. And for anybody who's interested in this, there's a really excellent book, and it's called the evolution of learning memory systems. And this is by Elizabeth Murray, Steve wise, and Kim Graham. And so what they've done is that they they asked this kind of question this, this evolutionary and ecological question. And so they traced invertebrates back, and so of the earliest vertebrate meeting animals with the backbone shut up about a half a billion years ago. And so that's before flowering plants, that's before trees, you know, the only things around these kind of like, funny things, right? So it's a really long time ago. And so they asked the question, you know, what, you know, what did these animals have to do, and they were aquatics they're swimming around. And the idea is, is that, you know, they have motor systems that they can swim around, they can catch whatever their prey was, right. So that when they detect the the smaller things, they can eat it, when they see a shadowy thing, they run away from it. And so, you know, you can you can think of this as responses in a general term, but, you know, in a way, those are decisions, right? Do I keep swimming? From what, you know, do I continue to try to harvest food? Or do I run away from this thing, because I see a shadow. And in very simple organisms, these things are all very kind of automated, right, the sort of prey drive and Predator avoidance. And, you know, as systems have evolved, really, what's happened is that the same architecture has persisted, but But it's become more elaborated. So even those very early organisms had homologues of even the medial prefrontal cortex, it had a simple visual cortex, it had, you know, some sub cortical stuff, the basal ganglia, and you know, these other lots of these other sort of deep structures. And what's happened over time is that this is really just sort of, you know, elaborated. And in the extreme case of the great apes, where we have this massive expansion of the neocortex, so the, you know, the folded up kind of tissue, that's, that's out on the outside. And so in an ecological context, we really kind of have to think, you know, I think using that as a context to interpret behaviors and responses is, is quite important. Now, okay, that's a lot of pontificating on these things, it might sound esoteric, but, you know, it still comes back in play. So in my mind, and I think there's pretty good evidence for this, you can have some choices are just mediated by your, the, your motor systems, so that those are the circuits in your brain that, you know, control your limbs and, you know, other other muscles. And so we have some good data for that. And we can talk about that some more if you want. There's other systems. So, for instance, if you are having a conversation with someone, and you're walking up to a vending machine to get a snack, and you just, you know, you sort of punch in a button, or two, to get to get your snack, right, that you might think of that as something that's more driven by your motor systems. And so maybe, you know, every most of the time you press B 15, and that's, you know, whatever snack that you're accustomed to having. And if you do that enough, that response is mediated by what's oftentimes referred to as the habit system, it's probably more appropriately called the sensory response system. But this is a system that involves,
it doesn't really involve a lot of your neocortex, just a little strip of motor cortex. And it's largely mediated by the set of structures called the basal ganglia, which are evolutionarily again, very, very old and very important for executing motor outputs. So that system is perfectly capable to generate a complicated output to to get your snack. And what we find in the lab in animals and in humans is that you tend to rely on that more when other brain systems are preoccupied. So again, kind Looking at a higher level view, what I mean by that is that the brain has multiple systems that can guide your choice. So you can think of one as kind of a motor system, like I just described, you can think of one as more a cognitive system, where, you know, you're debating over the calorie content of each snack and whether the chocolate was ethically sourced, and you know, these more cognitive things, you have other systems that are more probably emotionally guided as well. And so you have these multiple systems that are interacting, and at any one time, one of them might have a larger influence on your choice than others, depending on many different factors. So, you know, when you ask, you know, what's a choice in, you know, the, you know, where scientists would say is that some sort of response, which is, is vague, and it's intentionally vague, because otherwise, you have to make presumptions about what's doing what what part of the brain is, you know, calling the shots. And that seems to depend on a lot of things.
Nick Jikomes 31:01
I see. So so the choice is related to the execution of a motor program, and not the execution of potential alternative programs. But you know, what's happening just upstream of that can?
Aaron Gruber 31:14
Like, it can but then, you know, imagine another scenario that's largely cognitive. So imagine, you know, choosing a car or vehicle, right, it's not something that really involves, you know, it's much different than a vending machine problem. And so, you know, it's unlikely that your motor systems are going to be largely the ones responsible for generating the output in that case. So that's probably going to be more, you know, we like to think that's going to be more cognitive thing, because you're going to look at, you know, gas mileage, or is it a hybrid, or, you know, I have five people in my family, so I need something with a back seat. And, you know, this, this gets more into the psychology literature, but But you know, what it ends up is that a lot of times people have already made an emotional choice about something and then they just rationalize it. So, you know, there's, there's not really a utility argument to be made for, you know, a red two seater Ferrari for half a million dollars, right, when a Honda will do the same thing. So, you know, it's a really, it's an important question, I think it's an active area of research, like how much of it is sort of emotionally driven versus sort of cognitive system versus other types of systems that can generate a choice.
Nick Jikomes 32:33
But But it sounds like, you would not say that there's one physical locus of decision making in the brain, you know, where, quote, unquote, a decision is happening? Depends on the type of decision and all these other variables? Yes. So when we use terms like cognitive versus emotional versus motor, how, and this is going to be a somewhat vague question, we all have an intuition for what those mean, at some sense, we're sort of imposing imposing those things on the brain. How, how distinct are those concepts in terms of how cleanly they can be separated? We actually look inside the brain, like there's a circuits mediating emotional stuff versus cognitive stuff, and how much of it is just sort of linguistic? convenience for us?
Aaron Gruber 33:21
Yeah, absolutely. I mean, that that, that is, I'm using unabashedly jargon. And, you know, it doesn't, it doesn't fly very well, in a in a very strictly academic sense. But, you know, it's convenient to use that kind of jargon, because it does convey kind of a lot of information, you know, with very compact language. And, again, you know, your question is very excellent. And if I, if I had to guess that if I hadn't placed a bet, I would say that the motor systems are fairly separate from the other ones. But that sort of this idea of, of cognitive and emotion, I think, are very intertwined. And that there's not a clean separation. You know, we know this, there's, there's a lot of good evidence for this. And so this is something you know, I'm quite interested in and you know, it's something we can certainly talk about some more if you're interested in.
Nick Jikomes 34:15
One of the things I wanted to ask you to help us think about is like the way the different types of ways that we humans learn compared to many other animals. So for example, we and our dogs can both learn from direct sensory feedback of positive and negative reinforcement signals, right? So if you think about training your dog, you know, you give it a treat, when it does something good, you positively reinforce what you want it to do say in response to your command. And it sort of learns right there in the moment incrementally from from getting that reward. We can do that type of learning as well. A lot of the learning that you know, we've done throughout our lives is like that. But humans and certain other animals can learn and make decisions. In other kinds of ways, it seems we can actually generalize things that we've learned and make intelligent decisions and novel contexts that we've never actually been in before. So we can sometimes we can know the right decision in a new context, even though we've never actually been in that context and gotten any kind of direct reinforcement before. So can you talk a little bit about sort of that difference between, say, the human and the dog? And what's what might be going on there?
Aaron Gruber 35:24
Yeah, absolutely. And again, that's a great question. And I'll say right off the bat, that I think animals capacity to do, what you just talked about, is greatly underappreciated. So I think many kinds of animals actually can do that. And I think, even as back as the 1940s, so there was a very pioneering behavioral scientist. Ever. Geez. Told me, sorry. So he told me, he did these very clever experiments with animals with rats. Where, and this is a good segue for this. So he would have, let's see, the simplest version of this would be that there would be a starting position, and then there would be a place with food, and with the animals learned just a few trials is that there was one track that would go like this kind of like a, you know, a sea shake. And so they would do that a couple times. And that's all they needed. And then what he did is he put them at the starting point. But now he has these dead ends, that kind of pointed radially in all different directions. And so if the animals just learned, hey, to get my food, I need to go directly west, right, and then go north, and then east, right to get my food, you might expect that to happen. And some animals did that. But the majority of the rats, what they did instead is that they went on the dead end, that would have made a straight line for the food. And so So Tony came was the originator of this idea of a cognitive map. And he meant this in kind of a spatial domain that, you know, animals very quickly. And we now know that animals do this, almost as soon as you bring them into a room, they develop sort of they, they develop a representation in their brain of locations in the space. And so the idea being is that, you know, the rats even though even though they learned that the Luna food is over here, only by their experience of running this way, they could use that and knowledge of the space to say, well, if I ran this way, this would be the fastest way to get there,
Nick Jikomes 37:38
they could literally triangulate the shortest path. I'm sorry, they could literally triangulate the shortest path, even though he had never taken it before. Exactly.
Aaron Gruber 37:47
Right. And so that's this idea of having knowledge and being able to act as if you have knowledge, with no without actual prior experience of doing that. And so this is where this is where I think the brain is even the brains of many kinds of animals still exceeds what we can do with artificial intelligence and machine learning. Even though artificial intelligence and machine learning, it's great at very specific tasks. You know, it can now exceed, you know, finding pictures of cats and large images and that kind of thing. But it's, that technology is largely still beholden to the training set that you give it. Whereas the brain, what it what it seems to be really good at doing is acting in novel situations, which you mentioned before. And that's really important because the the natural environments are tremendously complicated, right? So if you got into the woods, there's all these contours, there's leaves rustling, there's branches, and tree trunks, and scuttling critters, most of which is completely irrelevant to your behavioral well being. And if you had to sit there and attend to all these things, you wouldn't be able to do the things you do. So you know, imagine you're you know, you're foraging for mushrooms or you know, your grubs or, you know, whatever it is, most of the information that's coming in is irrelevant. And you know, the best estimates are that that's, that's something like 10 to 20 gigabits per second, like if you added all the sensory, the bandwidth of all the sensory information that's coming in, and it's just tremendous. And it's so much that you can't attend to it all. And again, the problem that the brains have to solve is that only a few things in the environment are actually relevant to you or salient. So this comes back to this the dopamine system that what might be happening is that that's one way of, you know, out of all this sea of noise that's happening or relevant information when something important happens, right? You need to pay attention to it, whether it's good or bad, you know, a predator or a food source. Yeah, so So that's that part of it. And now I can talk a little bit about this idea. So How do you act when you're in a novel situation, and again, just about every situation we're in is novel. So for instance, you know, one of the just kind of humorous examples I give is that if I'm invited to go give a lecture at a new university, you know, I wouldn't show up in my underwear drinking a beer. But sitting in my underwear having a beer is perfectly acceptable behavior in my house. So but how do I know that right? And I never tried that I never got the, you know, the negative consequences with Dean's yelling at me, and embarrassing videos on the internet, and all all the consequences that would happen. But I know, you know, I'm pretty confident that that's what would happen. And so it's just, you know, it's an example of that the brain embodies knowledge that it's gleaned through experience, but not necessarily direct experience, right? Because again, I've never done that. And that's, you know, I surely know, that's not, that's not a good thing to do. And I've not even seen other people do it. Right. And so this knowledge, one way that they're one term that people use to describe this is what's called a schema. And it's your schema is how different things in the world relate to one another, that you can use in, you can think of it like a simulation. So I can anticipate, hey, you know, Wouldn't it be funny if, you know, if I did this thing? And I can think, well, you know, if I did this thing, you know, these people will, they might get offended? And that wouldn't be good, right? So you can actually think through to do these sorts of things and make those kinds of choices. And that's typically what's referred to as a model based choice. So you've got a model of how things in the world relate to each other that you can use to estimate, what's the outcome going to be. And that can be value that could be negative consequences. Do you mean to pause? Or do you want me to get back to why humans are seem to be different?
Nick Jikomes 41:53
No, no, let's, let's keep going on this trip. Okay.
Aaron Gruber 41:56
So one thing, so you pick dogs, which maybe was an unfortunate animal to pick, dogs are really interesting, because they are, to my knowledge, they're the only animals other than great apes that can infer what humans are looking at by their eyes. So you know, if you spent a lot of time with your dog, and you're looking at something, your dog would go look at it, try that with a cat, or a bird or any other animal just isn't gonna care. That's not That's not what they're what they're about. And so, you know, the reason why I think humans capabilities and other great apes as well, I mean, we're not. We like to think we're a lot different than the great apes, but probably probably not as much as we think. That the, to answer the question that you asked earlier, you know, why do humans have this, you know, seemingly, very enhanced capability to generate this, these schemas and these mental models of the world. And the reasoning seems to be again, this is pretty speculative. But the reasoning goes something like this, that humans are very social, well, that there's, there's actually two explanations, one of which is that our ancestors spent a lot of time in the trees, eating fruit, and that that accounts for our very good spatial awareness and color vision, because fruits that are right, typically change color. And so there's a utility in being able to tell which fruits are right, based on color, because then you guide your foraging towards the brighter fruits, which have more sugar, you get more calories, you have better fitness in your environment. So that that is sort of the the explanation for why humans color and sort of spatial awarenesses is good, is that it's an artifact of tree dwelling and fruit eating. But these other sort of more cognitive things, the best explanation that I know of is that it's an, it comes from the fact that we're very social animals. And so, being in a social group, it's quite important to be able to infer what's going to happen of your actions, because when you become dependent on a group, you really, it's not so much true in the modern day. But if you're, you know, in a small clutch of animals, you know, out in the jungle with lots of predators, and you know, food scarcity comes around as the seasons change, or whatever, you need to rely on the other members of your group. And so therefore, I need to be very attuned. So for the same group, and I'm doing a certain thing, and now I see you start to grimace or whatever, I now have to infer, like, Well, what did I do or say that's like pissing you off? Because I don't want to do that. Because if I do that, and you know, at a time when you have the food and I don't have the food, then you want to share with me, and so that's bad. So that's sort of the argument where the ability of having sort of these very broad and rich mental models comes from is that, you know, that we have to do a lot of a lot of inference, particularly in the social domain. And that, because that that capability was very successful in the social domain, then it applies itself well to other things like mechanical things. And, you know, we haven't been using tools that long, you know, in the context of how long vertebrates had been on been on the earth, but we've been in social groups for for very, very long time, or you know, our ancestors.
Nick Jikomes 45:34
So when you talk about something like a schema, some sort of model the brain can construct, its, you know, it's obviously a pretty complex thing, but complex phenomenon has to involve the direct sensory information, or at least some of it coming into the nervous system right now in the moment, that has to involve probably integrating information for memory, these sort of different value systems and things. So it's a very, very complicated entity that that we're trying to describe, how do you actually start to experimentally touch something like that and study it in the lab?
Aaron Gruber 46:09
Yeah, so this, you know, this is where technology is enabling. So you know, these ideas have been around for a while, but they've been largely theoretical, and very hard to test directly. Because, you know, the the standard toolbox, you know, prior to 20 years ago, maybe you could record one or two neurons at a time, you could do brain lesions, you can give from pharmaceuticals. The, you know, the the difficulty was setting the brain in that way is that, you know, if you damage a certain area and ask, well, you know, to somebody, does this part of the brain? Is it involved in that kind of behavior or function? You take it out? Maybe it was, but other regions can take over that functionality? And so it looks like there's actually very little deficit. So in the, in the field, read, there's two questions, is it necessary? And is it sufficient for particular behavior. And it's actually quite rare in the neocortex to find things that are necessary and sufficient, outside of the primary, outside of the sensory area. So there, you know, there are some, you know, language specific centers that if you take those out, then you can understand speech or produce speech. You know, there are some specialized things in humans like that. But most other animals have that language. That's not necessarily true. But that's not to say like, if you have a stroke and motor area, right, you can have motor, you know, deficits and those sorts of things. Certainly got lost
Nick Jikomes 47:40
on the question, you know, so what are what are some of the tools and approaches you actually use in the lab? Right, yeah,
Aaron Gruber 47:45
so the tools and approaches Yeah, so. So there's a lot of redundancy in the brain. And that, you know, the the upside is, is that that makes you robust against damage. So if you do have a stroke, or you get hit in the head, or something like that happens, you know, you're not other parts of the brain can take over that functionality, which, you know, helps you to survive. It makes it hard to study it with the classic tools. But the new tools that people have developed now are sensors that can now record 1000s of neurons at a time. Some of these are optical. And so we can now I just use these tools, I didn't develop any of them. This takes teams of geneticists and engineers and you know, tremendous amounts of work. But, you know, the tools have now trickled down to users like me, one of which are proteins that you can use a viral vector to express. There's actually this similar technology to like the COVID vaccine. So if you get little plasmids of RNA, and you get those into the cells, they'll express those proteins. And so you can do that. And some of the proteins, for instance, will fluoresce when cows when they come in contact with calcium. When neurons activate, they typically let calcium come into the cell. And so what will happen is that the cell that's active will fluoresce when it's active, and then it will stop flexing when it's not,
Nick Jikomes 49:07
you can do something like a virus or genetic engineering to breed animals that have their neurons have proteins that literally light up so that you can see what they're doing when they're active.
Aaron Gruber 49:19
Yes, yes. And so doing that you can record many 1000s of neurons at the same time, once you start doing that. Now, you know, once you have that information, now, you can use machine learning and other computational tools to try to detect or piece together these more complicated things like these schemas. So things that are not you know, a direct consequence of you flash the light on and some of these neurons, you know, turn off. That certainly happens, but what happens is, you know, you turn the light on, there's a few neurons that reliably turn on. There's other sets of neurons that sometimes they turn on, sometimes they don't, those same neurons, sometimes they turn on when a sound comes on or not. So it's not completely obvious, like what they care about. But, you know, again, using some machine learning tools, we can now start to ask the question, well, what, you know, what kinds of things are do these cells care about, and it's usually combinations of other things. And so when we put those combinations together, now we say, ah, you know, here's a cell that, you know, it cares about when this light has come on, but it's only after a reward. And so we can now infer that, you know, these kinds of cells are encoding, you know, some retrospective memory or something like that.
Nick Jikomes 50:34
So there's some neurons in the brain, say, say, in the patch of cortex that you're recording hundreds or 1000s of neurons from, they might be relatively simple in that they have a very close correspondence to one particular thing in the environment. So if there's a light in one side of, of the environment, you know, when the animal sees it, that neuron reliably fires, but there's other neurons that are somehow integrating all of these different channels of information. And so only when something happened in the past in a particular way, and the animal sees something and it's moving in a particular direction, or something, that neuron will be
Aaron Gruber 51:08
active. Yes, so and that's yeah, that's a really good explanation at the single cell level. And to find those cells, you have to sample you know, lots and lots of cells. So you need these, you know, 1000s, we're not quite up to hundreds of 1000s. But But technology is is progressing rapidly. The optical is one technology, the other technology is essentially the same technology that's used to make microprocessors. And what you can do is you can make essentially a super thin shaft, or several shafts that will have 1000s and 1000s, of electrical recording sites on them. So in that case, you don't have to genetically engineer your animal at all, you just you stick them in, and now you get these electrical signals. And so, you know, again, that's, it's enabled by, you know, technology in, you know, engineering,
Nick Jikomes 51:59
I see. So, so I guess the name of the game is that when you're studying something as complicated as the brain and how it makes decisions, or something like this, the way that you approach it, using a variety of different techniques that could be optical, that could be electrophysiol. Physiological, you're, you're recording the activity of many, many neurons across a sizable chunk of the brain at the same time, you're actually able to do this in animals that are awake and doing stuff. And then you're essentially just using computational techniques and analytical techniques to try and literally decode what those neurons must be caring about.
Aaron Gruber 52:34
Yes, yeah. And one of the reasons it's challenging to understand the brain, I mean, one of the beauties of the brain is that it never looks the same twice. So even if you, you try to keep everything as identical, as you can, more, you know, successive times, it never looks identical. In the past, it was thought that was just noise. But I think the contemporary thinking is that that's really just a property of the system that is probably important. And I think, you know, to add on a little bit, I think we can use, I think a good analogy is the weather. So if you imagine you want to predict the weather, you know, say somewhere in the central United States, you know, in the past, it would be like, you know, we could have one, one temperature recording in Bismarck. So say you wanted to know, the temperature in Cincinnati, Ohio, you can you can only sample one place in the United States, and you're trying to understand what's going to happen, you know, somewhere else, you know, where the reality is, is that it depends on the pressure and, you know, the air currents and all these things. And so what we're doing now, just like, you know, weather prediction has gotten a lot better, because now we can have lots of sensors all over the place. And we have better models of how the weather is changing over time, right. So it behaves as a dynamical, it's called a dynamical system. So anything with more than three parts, tends to exhibit sort of these chaotic types of dynamics that become hard to forecast. So you can forecast fairly well in the near term, but the farther and farther out you go, the less and less likely we're able to forecast it. And that's because very, very tiny changes in the initial conditions change sort of the trajectory of the event. And so the brain looks like that. Right? So what we're doing now is it's like putting lots of temperature sensors all over. And now looking at how does the instead of wind now it's electrical activity, how does it propagate through? And so it's quite a different thing than like a computer or, you know, a mechanistic kind of thing where thing A happens and you know, the cracks turn and you reliably get this other thing out, is very much different. It follows these dynamical systems sort of properties. And so, you know, going back with what we're talking about the, you know, the single neuron, so there's some neurons that reliably act To date, because of direct sensory input. That's kind of like the initial wind currents. You know, some of them are in North Dakota, some of them are in Florida. And by the time you get to the association courtesies, right, they're having, they're having some effect. And, you know, it's not always the same. But what's interesting is that it tends to have the same effect, which enables sort of learning, associative learning and these other things, as well as decisions.
Nick Jikomes 55:29
So one thing I want, I want to talk about is how things like certain neuropsychiatric disorders can impact cognition. And we've sort of built up this concept of a schema in our toolkit, which is just some sort of complex model that the brain constructs for a given situation. And it contains lots of different types of information, sensory information, information about value, it's connected to memory systems. Before we get there, is there anything else you want to cover? That you think we about cognition that that would be important to convey? Before we talk about the impact of say, depression or anxiety on these things? No, I
Aaron Gruber 56:05
think it's a very good segue, actually.
Nick Jikomes 56:08
So when we think about depression, anxiety, again, this is something where I think we all have an intuition for it. These are bad things. When someone is depressed, you don't feel good. You don't want to be depressed. And you also tend to look at the world in an overly negative way, I think everyone probably has a pretty good intuition for that. Can you start to talk about how something like depression or anxiety would impact some of these aspects of cognition that you've talked about? And how, what you know what, actually no, they're at the level of neurons?
Aaron Gruber 56:39
Yeah, absolutely. What I want to preface this by is that, you know, what we call depression and anxiety is probably several different things, meaning that different things in the system can go wrong. And so, you know, to different people, if they you know, if, you know, they present with anxiety, it might be because of different reasons. For instance, there's one structure called the amygdala, which probably many people have you heard about, that, when it becomes hyperactive, it tends to produce anxious states. So things like post traumatic stress disorder, PTSD, is strongly associated with hyperactivity, the medulla. And so it's a system that it receives a lot of direct input receives output of the hippocampus, which is an essential memory system. And it seems its job seems to be to form relatively simple associations, right. So you know, like in combat, if you saw a certain type of, you know, vehicle, right before you hit a landmine, or something, you know, and then you're in a completely different context, you're in North America, and you see that vehicle afraid, you get this, you know, stress response. And so the, you know, we think that the amygdala does that, that kind of thing. And the reason you get the stress response is that there are, you know, many small little nuclei that are in, you know, around the brainstem midbrain, that secrete certain neurotransmitters and hormones that trigger stress response. Now, other types, and this is the, this is the, so that's kind of one type that I'll put out there, it's not necessarily the third, they're clean types, and it's one thing or the other. That's just kind of one form that seems to show up. But I want to talk about a different form that ties in better with the seamless. Other people that present with depression, anxiety, if you if you put them in a magnetic resonance imaging machine, and you look at where blood flow is occurring, and you can give them stimulus or you can ask them questions or whatever. And if you look at how the blood flow is changing, we can get what's called a functional MRI. I don't do this work, but I think it's important to, you know, to sort of discuss this and then we can talk about it in animal models. And what you oftentimes see is that there's a brain region called the anterior cingulate cortex, that sort of on the midline, so it's sort of medial and the prefrontal cortex, it becomes hyperactive. And this correlates very strongly with anxiety, and especially what's called negative rumination. And so what this is, is this is unwanted compulsive thoughts about something bad that either did happen, or you know, could happen. So, you know, bereavement is a good one, like you just can't get over, you know, a level of dying or, on the other hand, it could be you know, I'm afraid of flying because, you know, might get hijacked or you know, something that's very low probability has never happened to you that you really shouldn't have cause for worrying about it, but you do and you just can't, you just can't help it. So when those things when those things are happening It looks like the anterior cingulate is involved in in that. And so this is one brain region that I've been studying in animals. And it's a little bit special for a few reasons, one of which is that it's one of these parts of the neocortex that's phylogenetically conserved. So, whereas you know, a rat doesn't have this lateral prefrontal cortex that's unique to the primates, or sorry, the great apes, and other other monkeys as well, the medial parts are conserved. So you can find a humble lot of that in, you know, rats, mice actually don't know about alligator. So I don't know too much about the other animals anyway. So the anterior cingulate, though has, is, has sort of special connectivity with the hippocampus. So going back to what you were mentioning before, about how these schemas sort of, you know, there's some product of your experience, either direct experience, or things you've read on the internet, or have heard from your friends, or all these different things, right, they, they get sort of processed in your memory systems, and then exported in some way to your neocortex. And it looks like the anterior cingulate has a very strong connectivity with the hippocampus. So it somehow is very strongly linked with with memory systems, or, you know, the one of the key components of memory system. But the St. Pierre singlet is also well connected with these sort of brainstem structures that trigger
anxious like behaviors. So freezing, it's also connected with neurons in what's called the lateral hypothalamus that release CRH, that corticotropin releasing hormone, which gives you the adrenaline response. So it's hooked. And it's hooked up to other cells that control heart rate, and breathing and all this other physiological stuff that you feel with stress. So what it looks like is that this area is somehow, you know, at this nexus of, you know, learning systems, but it's more than just learning. It's it's the knowledge that has been acquired, and sort of these mental models or schemas. And so the this interesting of the cortex seems to be important for schemas. And you know, again, it's connected with sort of the stress a factor so that things actually produce the physiological symptoms of stress. And what's, you know, the sort of state of the art, what we know is that antidepressants, for instance, reduced activity there. For people with major depression, who do not respond to drugs, drug treatments, the last line of defense is actually to go in and make lesions of this anterior cingulate. So you go in with an electricity surgeon goes in with an electrode, and actually applies electrical current to essentially both ablate that material, and the success rate is like more than two thirds. So these are people who, you know, could barely get out of bed. You know, we're not responsive to several different kinds of drugs to talk therapy to anything, you go in and zap it, you know, like, hey, alright, things things aren't so bad. Which, you know, I think is why, you know, myself and other people are quite interested in, you know, what that brain region is doing. One thing I'll add to this is that, you know, even though it seems to be a key player in anxiety and depression, to really understand what's going on, we have to know what it's doing sort of in, in other contexts as well, like, what's his main job? And then we have to know that in order to know, how does it go wrong?
Nick Jikomes 1:03:46
I see. Yeah, kind of comes back to the, the notion you touched on earlier of thinking in neuro ecological terms, what is what is actually the circuit or this area doing? Why did it evolve? What is it doing for the animal, its natural state, and that can perhaps help help you guys make sense of what it's doing when it goes awry?
Aaron Gruber 1:04:06
Yes, exactly. And if I can, but a bit of a speculation, but one of the one of the kind of neat things that, you know, has come to my attention. So, one of the things hopefully, like that, one unfortunate thing about sciences is that we end up getting what's called siloed. Meaning that, you know, I might study this anterior cingulate cortex, somebody else's studying, you know, what, olfaction or something, you know, something else, and that we don't necessarily those communities don't necessarily talk a lot, partly because, you know, there's now hundreds of 1000s of research articles in each one of these disciplines. And it's, it's become, there's just too much information for anyone person to, to synthesize. And so we're kind of kind of forced in in that way. But what's interesting about the story coming back to the anterior cingulate cortex, what's, what's interesting about that is that in one of these other silos, people have been looking at what's called observational learning. And so if you put, again, someone in a functional magnetic resonance imaging system, and so first of all, if you give them pain, so if you, you know, if you poke them on the finger a little bit or whatever, you'll see part of this, the cingulate cortex light up, it's actually a little bit more mysterious. It's not exactly the same anterior cingulate, but it's, it's a continuum, just to be, you know, clear about this. So if you poke around, you'll see that parts seem to respond to pain. They also respond to videos of people in pain, and also, for people that are in precarious situations. So you know, if you've ever seen an image of, you know, again, for me, like in mountain biking, if you see someone, you know, biking on the skinny little trail, that's right, on a cliff, right, it makes me sweat, you know, it makes me quite uncomfortable. Again, even though I haven't done that, I've not gotten hurt doing that. But, you know, I, what I know about the world is that there's gravity, and, you know, trails have gravel on them, and I know, they're slippery, so it's quite conceivable, you know, it might be about to see some dudes, you know, fall off the fall off the cliff. And, you know, coming back to why do people why do people do that, again, you know, it's this idea that there's a social component to it. And so, you know, particularly, you know, what we know about primate behavior is that they're dominant, you know, that is sort of this, you know, social dominance hierarchies, and there's the, you know, the alpha male, and, you know, people will challenge the alpha male, so from a FOMA, you know, upcoming teenager monkey, and I see my buddy go over and challenge the alpha and get his butt completely kicked, it's like, well, maybe I shouldn't do that, right. Because if I do an A, I might get hurt. And be, you know, by doing that, it's going to lower my social standing. And so, you know, the other animals in my troop might not share as much food or, you know, I might not get as much benefit. So I think that's where this, this comes from that what what our brain does, is, it's now so attuned to taking in information, so that we don't necessarily have to make the mistakes, so we can learn from the mistakes of others, so that we ourselves don't, don't mediate them. And I think, again, this is this is speculation. And the reason I think that we get this physiological responses is that it's important for the learning process. So we talked about reinforcement learning before that you actually, you know, you need to feel like if it tastes good, you know, the value goes up, if it tastes bad, or hurts, right, the value goes down. And so by watching these videos, by actually embodying you know, by actually embodying a negative aspect of state of fear, worry, whatever, we're actually learning from that, and that that's a useful component of, of learning. Yeah. And then actually, you know, feeling the pain, you know, if my, you know, teenage monkey buddies, getting his butt kicked, I'm actually feeling, you know, empathy for him. But that's actually helping me or might is driving my learning systems without having to actually suffer the consequences directly.
Nick Jikomes 1:08:23
Yeah, it makes sense. You might think, you know, I'm someone who's like, really sensitive to, you know, if I'm watching a movie or something, and I see someone get a paper cut, like, I almost can't stand it. Now, it doesn't obviously, literally hurt, like me getting a paper cut does, but there is this sort of real physiological response, even if it's not fully what I would get if I was in that situation. And, you know, naively, you know, one might think, Well, why do i Why does that type of response in me need to be baked in at all? But if what you're saying is true, it almost sounds like the physiological side of it, the actual feeling of the simulated the thing that I have in mind, or that I'm viewing is somehow necessary for for the learning component of this, that that needs to be there.
Aaron Gruber 1:09:07
Yeah, absolutely. Like, I think, you know, the under again, speculation, but you know, I would, I would put a good wager that this is true. But that, you know, this systems can have exaggerated responses. So, you know, a paper cut is sort of, you know, is a great example, it's something you don't want to have happen, but it's not, it's not life threatening, either. It's not the same as you know, watching people do insane stance, and feeling that if we can actually cycle back I think this actually makes a good segue. So back back to depression. So you know, one. So what does that mean? Well, in depression, linking this back to the schemas is that depression and anxiety are oftentimes associated with what are called negative schemas. So these are schemas in which you expect bad things to happen, even though you don't even know the evidence you've had shouldn't really necessarily suspect that, you know, one example of this is that, you know, if I'm sitting in a room full of people, and I'm kind of glancing around, then I see someone who started looking in my direction with a scowl. If, you know, if you have, you know, a positive schema, you might think, oh, that, you know, what happened to that guy, maybe, you know, maybe he just got dumped, or, you know, he lost some money in the stock market, you know, we don't know, who knows. But if you have a negative schema, you might think, Oh, gee, that guy, you know, he doesn't like me. Or, you know, it's something sort of internalized that there's some negative thing about you that's causing this to happen. So, you know, that would be an example of a negative schema. I mean, also, the example I gave before about flying and having a, you know, flying, your chances of dying in flight are much less than other modes of transportation, pretty good driving, but, you know, the sort of the, the fear states in the negative schemas aren't necessarily driven by facts. And I think this is a testament to this overlap between or intersection between learning and memory systems and emotion systems.
Nick Jikomes 1:11:05
So, so two people could have the exact same schema, the exact same mental model of something they're looking at, or something that they're imagining, but one person might just be biased towards, you know, pulling that in a negative direction and framing it in terms of fear, or, you know, just thinking thinking about something negatively, everything else is the same, we might both be thinking about the same airplane in the same vacation. But you know, you're thinking about the positive, the positive side of where you're going when you get on the plane. And I'm simply thinking about the negative side of how much I hate flying.
Aaron Gruber 1:11:37
Yeah, so I would actually phrase that a little bit differently. And I would say that, you and I might have the same data, so pretend that we have the same data, but that your schema is a little bit negative than mine, because of other experiences you had that weren't related to that data. Right. So by data, I mean, you know, here's a, here's a little dossier on everything we know about flying. So pretend you and I didn't know anything about airplanes, we got this pamphlet on, you know, here's, here's how it works, here's, you know, safety records and all that stuff. The you know, sort of the, the schemas that we develop, or the internal models we develop are going to be colored by all sorts of other things, other life experiences, also by genetics, and other things that aren't even in our control. And that, then having that, you know, having constructed that model, and then we ask, you know, someone asks you, and then I, you know, how do you feel about this, and I'm like, Ah, you know, instead of taking, you know, a boat across the Atlantic for a week to get to one, then you know, you can now do it in six hours, it's amazing, I'm going to see Big Ben versus, oh, that's up in the air. And this is a new technology, you know, if that thing falls, you know, dead. And so I think I think that is, that's how I would think about it, you know, and again, this, it starts to get more into the psychology literature, or the psychology and that I'm not nearly as well versed in, but you could, you could very much imagine that these aren't fixed things. And so they can wax and wane with, you know, the variety of neurotransmitters and hormones that you've got. So you know, what you might kind of feel one way about something one day, and then you sleep poorly. And then the next day, you know, we know that sleep disruption affects all sorts of hormones, and neurotransmitter concentrations. So you might think differently about it the next day. So these aren't necessarily sort of fixed entities.
Nick Jikomes 1:13:34
The other thing I wanted to get your perspective on is how, you know, different drugs affect these kinds of things, how drugs can sort of tip the scales, one way or the other in terms of how we make decisions, what decisions we're predisposed to making, and things like, you know, schemas and how our, our modeling the world, there's a, there's a variety of directions we could take on this. I know that one, one drug class that you've looked at somewhat are stimulants, amphetamines. And I'm wondering if you could comment a little bit about how that class of drugs affects things like decision making, and and attention. I think something like Adderall, for example would be something that a lot of people have heard about. But I know that there's some interesting, some interesting effects of these things that differ in terms of, you know, the dose you take, or whether or not you're taking them acutely or chronically.
Aaron Gruber 1:14:23
Absolutely. And so, yeah, it gets very complicated, because the direct action of the drug can be different than if you've so first of all, it often depends on dose. So one thing that shows up all the time in pharmacology, is that a little bit of something can have a different effect than a lot of something. And so, you know, a good example of that would be caffeine or alcohol. So a little bit of caffeine, a little bit of alcohol can actually increase attention, increase, you know, perceived energy. But, you know, if you then go Have twentieths presses on a day or something, you know, your your attention actually falls off. So st the same thing happens with amphetamine. So low levels of amphetamine increase attention vigilance. Which again, yeah, these are psychological terms that I think we're coming close to understanding more in terms of what does that mean in terms of neural neural function. And it can affect how people value things. And as you increase the dose, those effects kind of revert. So at low dose, oftentimes, it has a very facilitating effect on attention, memory, retention, all those sorts of things. But at a higher dose, it has a deleterious effect. One really interesting thing, and this goes to how memory occurs, is that, so let's talk about learning for a minute. So if you teach, so you're going to teach a rat to do something, you know, say when the light comes on, he presses a lever, he gets some reward. If you give him a little bit of amphetamine, he can oftentimes learn that more quickly. And so we can measure that in two ways, like how many trials is it take to get to the criteria, or you give them a certain number of trials? So let's say we do it 20 times on day one, and then we come back on day two, and just ask how quickly can they do it? Or how well did they do it? If you give them amphetamine, low dose, and do the same number of trials, and then asked the same, the next day, it looks like they've had more training, right, so they actually look better at it. What's really interesting though, is you can actually, you can train them without amphetamine, give them the 20 trials, wait a half an hour, then give them infant, I mean, bring it back the next day. And you also see this facilitating effect. And the the most likely explanation for that is that a lot of learning occurs in your downtime. So when when you're sleeping, but even quiet sort of restfulness, what appears to happen is that at least in some cases, the brain tends to recapitulate the activity that was going on to when you were doing a task. And so it looks like amphetamine can affect not only the direct experience, but also sort of the brain's recreation of that afterwards. So this probably this is one reason that you know, good sleep is important, and probably just having downtime, you know, hanging out in a hammock or
Nick Jikomes 1:17:29
whatever. Yeah, that's interesting, the brain is recapitulating. What happened to it in the moment and how it does that. And whether or not there's a drug in the system can impact how that's happening and whether or not whether or not the memory is consolidated as much or or gets weakened, for example? Exactly. Yeah, I know, this isn't your direct area. But that actually, so that that reminds me of this this other topic that I've discussed many times on the podcast, stuff that's getting a lot of attention in the popular media as well, with some of the concepts you were telling us about earlier. And that's what psychedelics have been doing in various clinical trials. The thing that I wanted your take on here is the following. You know, what strikes me as is, is that you've seen these very striking clinical trial results so far for things like depression for things like addiction, that are very complex disorders that we really don't understand. And they're very personal, right, the way that one person's depression or addiction is going to manifest is not going to be identical to someone else's. And you've seen some of these drugs in the context of psychotherapy have very, very strong effects at alleviating, say, a severe addiction or severe depression. And you've seen comparable at least directionally similar effects with different drugs. So for example, you know, addiction in addiction effects for something like Ibogaine and psilocybin, they're both strongly psychoactive, but they're operating through different mechanisms, they're different drugs. You know, something like MDMA and psilocybin for things like, you know, depression or PTSD. And so directionally, you're seeing these these similar effects, they're alleviating these very negative, very complex neuro psychiatric disorders. We don't know exactly how all these things are working at a mechanistic level. But you know, it sort of reminds me of the things you were talking about earlier around schemas and our ability to modulate you know, sort of the the positive or the negative valence on the schema. And I'm wondering if you thought at all about, you know, how or why drugs like this might be having this type of effect on something like severe depression.
Aaron Gruber 1:19:36
Yeah, that is the trillion dollar question. Yeah. Yeah. I mean, that that really, you know, if we could answer that question, we would know so much about the brain and we'd be be marvelous. I think it's a, you know, unpacking what you've just asked touches on a lot of very important very fundamental things. So let me preface this by saying that what a lot of the deleterious things on schema, so the things that tend to push towards depression and towards anxiety towards the negative schemas, things like stress, things like psycho stimulant abuse, things like inflammation, which turns out to be a big 101 of the things all these do is they reduce spine density in a few areas, including this anterior cingulate cortex and hippocampus. They do a lot of other stuff too. So
Nick Jikomes 1:20:36
they reduce the number of connections and certain parts of the brain.
Aaron Gruber 1:20:40
Yes, and what may be happening. And again, there's some evidence for this. But again, you know, everything in neuroscience, it's, you know, it takes a lot of evidence to say things to say anything with a lot of confidence. But my I'll give you my take, and again, this is partly speculative. What we do know is that spines, the spines are lost, what I believe is happening is that the spines that remain essentially become dominant. And to be could become dominant in such a way that now it's hard to change that schema. So as I was mentioning before, I mean, these schemas, they can be updated, they, you know, from day to day, even they can kind of modulate a little bit, right, they there shouldn't really be be fixed, like as new information comes in, you should, you should update that. But if you start to lose a lot of spines, and then the ones that are left over, what happens is that you, you these schemas sort of crystallize into something, and if that's something isn't good, or you know, is associated with a negative effect of outcomes, what happens is that as the information comes in, it sort of runs the schema, you can't really change the schema. So imagine that something like the anterior cingulate cortex, the anterior cingulate cortex, kind of filters that, and then it starts to activate these deep brain structures that you know, trigger these sort of anxious, depressive sort of phenotypes. Right? And so that would be, you know, one way to think about this, well, how would you get over that? Well, if you could, you know, for instance, make it more likely that you could then update these schemas, that would be one thing to do. One thing that all the one thing that all the psychedelics do is that they regrow spines in the neocortex like crazy.
Nick Jikomes 1:22:29
Yeah, the last, the last guest I had on was Alex Kwan. And he actually showed us some of that data where you give the mouse psilocybin in this case, and new spine start forming, there's a net gain a number of spines forming in certain parts of the neocortex. But what was especially interesting to me was, this continues to happen after the drug is gone.
Aaron Gruber 1:22:48
Yeah. And so we don't understand the molecular mechanisms. So there are a number of molecules that are floating around that serve as growth factors, essentially, they serve as molecular signals to you know, essentially polymerize, the cytoskeleton inside the neuron, which then forms the synapse, and then a number of other things have to happen to make to make that functional. But it turns out, there was a phenomenal paper a few years ago that showed that many different psychedelics do this. So psilocybin LSD, very short acting, one called DOI and a few others. It turns out that ketamine does this as well, even though it's site of action. So this this, the psychedelics, the main psychedelic action happens because of agonist agonism of the what's called the Five HTT a receptor. So that's one of the serotonin receptors. So when you agonize that you get the psychedelic first if you bought that, even if you get a psychedelic, you don't get the psychedelic effects. Whereas ketamine blocks an NMDA receptor, which is which is quite different. Well, what's interesting is that the ketamine induced spine growth depends on the five HTT to a receptor. So even though its primary site of action is a different receptor plate, it's somehow is also tickling that five HT to a we also know that ended the precedents, also correspondence, but much, much slower. So it's really tempting to speculate that that's why it takes antidepressants so long to work, whereas psychedelics, you know, tend to go very quickly. Now, we have, we have no idea why and there can be. So I've been putting a lot of thought towards this recently. And there are several different explanations. So one of which is that you could turn down the effectors meaning those very deep brain centers that trigger avoidance and freezing and increased heart rate and you know, all those physiological symptoms, it might somehow turn down their excitability maybe they become hyperactive number two could somehow be affecting the memories slash schemas, like maybe somehow, you know, attenuates those. A third possibility is that what they might be able to do is help. Competing memories of brain systems become more competitive against the negative schemas. So what I mean by that is that you're not, you know, what the general consensus is, is that, you know, if you're going to play chess, you have like a chess schema that you're using, right. But if you're going to go and do something else, that, you know, you can somehow swap these mental models in and out as you need to. And what we also know is that when you, when an animal or human learns, has a bad experience, and generates fear response, that fear response will attenuate over several days, like a week to two weeks, you know, if you shock a rat or a mouse in a chamber, and then you put them back in there, the next day they freeze, you keep putting them in there day after day, what you see is that the freezing is less and less and less. That's not because they're forgetting, because if you give them a reminder, even a very brief reminder that memory comes back, lickety split, what seems to happen in the interim, time is that some competing safe memory comes on that sort of suppresses all that, you know, the the fear response. So again, it could also be that what psychedelics are doing is they're potentially eating other ways, you know, for a better way to describe this, but they're potentially eating some of the way to look at the problem so that you're not predominated by this sort of negative way of looking at it. And when you do that, then you're not sort of triggering the activation of those stress effectors. Right now, it's all speculative. It might be one or more of those things, or something I didn't even mention. It's sort of the wild, wild west for psychedelic research right now.
Nick Jikomes 1:27:00
Yeah, but but the common thread is, whenever it's been looked for, it sounds like there is this increase, or at least change in the amount of structural plasticity that's happening in response to these drugs. And that's, that's probably what you would expect if they were to have this kind of effect, right? Because at some level, it must be true that if you're going to get over something like depression, or addiction, circuits need to change, they need to be sort of hooked up differently. However, that manifests. So if, if there is increased plasticity, that's at least making it possible for such a change to occur.
Aaron Gruber 1:27:35
Yeah. I agree. But how it happens, that's a Yeah. Hopefully, we'll know within years, but it's, it's hard to hard to forecast and neuroscience is interesting, because it also brings up, you know, may be of interest to your listeners is, brings up an ethical question as well. You know, certainly someone who has severe trauma in their past, and, you know, here's a way to help them. Right, I think, I don't think anybody would really argue that that's sort of an ethical use. But you sort of get into this slippery slope, when you can start to modify memories or responses to memories. And, you know, this is where pop culture and other people may be a little ahead of the game. So one of my favorite movies is Eternal Sunshine of the Spotless Mind. Which I forget how many years ago that was, but at that time, it was completely, you know, fantasy fantasy land. But who knows with with current technology, how things are going, you know? Is it okay to just go in and modify people's memory?
Nick Jikomes 1:28:45
Yeah, that's interesting. And I don't think we're too far from from that?
Aaron Gruber 1:28:50
Well, because imagine, you know, imagine if you knew, hey, I could, you know, I can make a lot of money, but it's going to cause suffering of a bunch of people. But afterwards, I could just have that erased from my mind.
Nick Jikomes 1:29:03
There's people who won't do that. Right. So
Aaron Gruber 1:29:05
you know, I mean, well, there's this link of, I don't know how strong the evidence is, but it's certainly you know, in pop culture, the idea that, you know, CEOs sort of lack empathy, and they kind of have to, because they have to make decisions that are, you know, often going to have negative consequences for employees or other people.
Nick Jikomes 1:29:25
Yeah, almost the more utilitarian mindset might be negative in the short term, even if it's beneficial in the longer term or whatever. But, you know, one of the things I did want to ask you about and again, we're gonna step away from, you know, the actual science and your research a little bit, but but I think that's okay. If you're okay with it, I wanted to ask, ask you, someone like you, you know, what your general take is on the impact of mobile technology and social media on cognition. And, you know, I think everyone probably has an opinion about this that's directionally similar, but I thought you might have an interesting perspective. You know, I would speculate that many people listening probably have had a similar experience to me, which is that over time, as our cell phones got more and more intriguing, and, and powerful, as we spend more time online and engaging with social media, speaking for myself, you know, my ability to sustain attention on one thing deeply, for a long time has degraded somewhat over time. And, you know, when I start to think about people, basically anyone younger than me, you know, I'm sort of about, I think, as young as you can be, where most of my childhood was spent in the absence of these technologies, when I start to think about people who've grown up with, you know, that cell phone in their pocket the whole time, you know, what is that doing to their brain and their ability to, to have a cognition that similar to mine? How much different could you know, the way that people that are just growing up now? How much different is their cognition going to be their ability to sustain attention? And to think about things in the way that I implicitly assume everyone else thinks about thinks about them? You know, what's your take about on how social media, in particular, is affecting or degrading our ability to focus and attend and decide?
Aaron Gruber 1:31:13
Sorry, what'd you say? Now? Just know exactly. I mean, I Well, let me let me answer a related question first. So that kind of follows from what we've been discussing so far, which is that, you know, social media in particular, I think tends to, or has enabled information, filters or information bubbles, unlike what people have seen before. And I think we have no idea what the consequences are. And I think, you know, I, I worry, again, this is pan waiting speculation, this is not is not me, as a scientist, this is me sort of as a general person and parent, sort of worried about this, the state of things, that, you know, what I was mentioning about schemas is that it's all a function of not only your direct experience, but what you're reading what you're hearing from people. And so, you know, how can you possibly, I don't know how you explain sort of the bifurcation that you see on a lot of topics, right? Masking versus anti masking? Like how can people be a, you know, again, take if everybody had the same information, you wouldn't think it'd be so polarized, where you have some distribution where most people think this, some are on one extreme end or the other, but it's really kind of a bimodal thing that it's like, you have to do it? Or you're Hitler, or, you know, if you do it, then you hate freedom. Right, it's really great that the thinking around it is very, almost diametrically opposed. In Yeah, you can see that in a lot of places. I mean, I think, in the, in the political sphere, particularly. And so I think, you know, again, I think social media is one of the enablers of that, I think it's one of the reasons that we want to say, manipulation of information on social media platforms, is really a problem and really needs to be addressed. So, you know, things that were going on with certain platforms before. The last, I guess, the last two US elections, you know, it's really problematic, where, you know, you can pay to either bump up or bump down certain information, whether it's true or not, you know, probably has a big influence on the way people see the world and interpret things. And, you know, that has huge influence on not only politics, but it's, it's the name of the game in terms of selling people things, and, you know, commercialism. So, you know, at some point, we really got to think about, you know, what's, what's it doing was to learn long term consequences. Now, once your other question, which was, was it due to attention spans? I have no idea that's really outside of what I know, directly. You know, again, the, the sort of what I would say sort of one of the popular interpretations is that you get your phone out, because you want to see, you know, you're hoping to see a light on something or you're hoping to see, you know, an interesting story. And when that happens, that's something salient to you, you get this little dopamine hit, and that's reinforcing. Again, the idea that don't mean No, it doesn't say whether it's good or bad, but that it's relevant to you. And so in that case, it doesn't really matter if someone's saying, you know, this politically, this this person who's politically juxtaposed to your positions, something bad happened to them, or they said something outrageous, right? It doesn't necessarily mean if it's good or bad, but it's just as relevant to you and your schemas and your, you know, the group that you associate with And so you know, how much does that interfere with getting the the business of what you need to do during the day done? You know, for me personally, it certainly is the Lord. And people really like information. And now that we essentially have this unlimited bandwidth of information of various quality coming at us all the time. Yeah, it's hard to concentrate on other things,
Nick Jikomes 1:35:22
are you? How intentional Are you? So given your background and the nature of your work? How intentional are you with, you know, putting in guardrails to limit access to your phone and things like that? Or if you have kids? Do you have any? Do you have any guardrails that you put up for them?
Aaron Gruber 1:35:39
I do. And I, you know, I know the whole argument of nature versus nurture is way outside of the scope of what we're talking about. So I don't know the reason but I have a 14 year old daughter. And you know, that's at the age now, where she reports at school that many, many of the kids in class when they have downtime, she said they're just on the phone. And so it's been hard for her to make friends. We just we relocated school districts recently. You know, so I certainly see that even though she herself doesn't have any aspirations to but I absolutely, you know, we put put limits on stuff. So we have limits on screen time. We, you know, I use the parental controls and all the devices to try to limit it. You know, me personally, my kids are more interested in things like Minecraft, which I think it's not, I don't know, doesn't really have all the problems of what we're talking about, but so long as it's within reason. But social media. Yeah, I mean, you know, I think I hate to say it, I don't like to make this comparison. But it's almost like drugs, where, you know, I keep having to relay this to my kids, too, that we now know that the prefrontal cortex in humans isn't fully matured until now, big 20s used to be thought it was like 18. And then it was 20. And it just keeps getting pushed back and back. And so you know, if you start to have things, drugs, or experiences that start that influence that, before it's fully developed, I think you get different outcomes. To me, it's kind of funny, right? So me, growing up, I'm a product of the early 70s. So I didn't even have a computer until I went to college. I didn't have a cell phone until I was like 30. So to me, that's very, it occurred very late. Neighbors in my neighborhood are in their 70s. We're just having an internet issue recently. It's very intermittent. And I asked both of my neighbors, one is 65 villains, again, is in the 70s. It was like, Oh, are you having a problem with your internet? And each one of them said, I don't know. I haven't used it for a couple days. And I was like, what? could check anything? So you know, I think, definitely, we can see the the progression? Well, I really do, I've tried to limit it on myself. I tend to creep back to spending too much time, mostly on news aggregators, more so than other things. But even that's too much. So it takes an active, active battle.
Nick Jikomes 1:38:19
I don't want to take too much more of your time, Aaron, you've been very generous. Any final thoughts you want to leave people with about any of the general topics we were talking about within the realm of cognitive neuroscience? Or any, any resources, books or places on the internet, people might go to, to learn more about this stuff, generally speaking, if they if they don't have a hardcore neuroscience background?
Aaron Gruber 1:38:43
That is a really good question. The, I guess, one, one passing thought I would like to leave people with is unfortunate. One is a sort of tale of caution, which is, you know, be aware of what you read, unfortunately, a lot of journalism I see on sciences is pretty off. And it's really hard to negotiate back from fiction without having some expertise in it. So, you know, definitely, when you read something, and especially if it's making bold claims, you know, verify the source, try to look into it a little bit before you take it on space, which is something I actually have an exercise that I have my undergraduates do. So that's the take something in the news and then kind of do a critical review of it. And I wish there was a really trustworthy source of information. I will say that the NIH National Institutes of Health, they're the chief sponsor of a lot of the biomedical research that goes on in United States, they do a good job. So they're, they're quite unbiased information. So that's a good place to look look for things. You know, I also think I think the future is pretty, you know, Brite in terms of fixing some of the problems that we've got, like depression, anxiety, I would, you know, again, caution people, I think psychedelics are probably going to be very efficacious. But, you know, beware of too many bold claims or I think there's going to be a gold rush in terms of people trying to profit off of it and, you know, kind of making bold claims and those sorts of things. So, do do your due diligence on that.
Nick Jikomes 1:40:32
All right, Professor Aaron Gruber, thank you for your time.
Aaron Gruber 1:40:35
Thanks, Nick. Was the pleasure talking with you today?
Comments