That Neuroscience Guy - That Neuroscience Guy Lecture Series - Part 2

Episode Date: May 4, 2025

In today's episode of That Neuroscience Guy, we pick up where we left off last week with Dr Krigolson's lecture titled "Why we do the dumb things we do". If you haven't checked out Part 1 yet, please do!

Hi, my name is Lone Wolf Krogolson. Hi, I'm Brayden Allen. I'm a neuroscientist at the University of Victoria. I'm a neuroscience undergraduate student at Indiana University. And we're the Neuroscience Guys. So on today's podcast, it's going to be a bit different. I'm going to give you the lecture that I gave to the students at Indiana University.

Why we do the dumb things we do. Now, this leads us to a simple model of decision-making that goes back to Huygens. Always choose the highest expected value. There you go, that's all you need to know about decision-making, compute your expected values and always choose the highest expected value. Well, of course it's a bit more complicated than that.

There's a problem with Huygens, this idea of just compute expected values and always choose the highest expected value. What if your highest expected value doesn't exist anymore? I love pizza, so let's stick with pizza. It's a great example. You phone up your favorite pizza place and you find that they're closed.

They've gone out of business. So what do you do? If you've only ever ordered from that one pizza place, you have no idea where to order from. So this leads us to a concept called exploration. Exploitation is when we always choose the highest value. Exploration is when we deliberately choose an unknown or something with a lower value. And the reason we do this is we want to get a better estimate of the value of the choice. So you choose a pizza place you've never ordered from

before to find out what the value of that pizza place is. So you explore, you make these choices, and you explore because of unknown values, you choose things that you've never tried before. And this is a good idea. Like try it. Try that thing on the menu you've never had. Go to a place you've never been.

Because what if it's better than what you think is the best thing already? And if it's worse, well guess what? It affirms what you believe, and you just don't do it again. Now the other reason you explore, there's two reasons, is changing environments. What if the world changes? Here in Victoria, I'm not going to name any names,

but there was a pub that I used to really like going to, and they had the best nachos in the world. The nachos were fantastic. But then they changed their chefs, and the nachos there now there are garbage so Because I had explored a lot and I spent a lot of time wandering around trying nachos When I found out the restaurant was closed

I was very quickly able to pivot and try a different restaurant, right? Like I was able to get out there and say well, hey, okay So I can't I can't go to this place for my nachos, but I know this place used to, you know, this was a place I would have said the second best nachos were at. So you explore because of unknown values, you don't know what's out there, and because of changing environments. Now, how much should you explore? On the slide deck, I've got a graph that shows basically how much reward you win if you explore 30% of the time, 10% of the time, or 1% of the time.

Well it's an impossible question to answer because it depends on context. The take-home message is you should explore. And what we've learned through research is that early in learning, when you move to a new city and you're trying to figure out where the good pizza is, you should explore a lot. But once you've figured out the value of all the pizza places, you should explore less.

You shouldn't go to zero, because what if the world changes? But you should explore less. So how much should you explore? I can't tell you that, and neuroscience can't tell you that. But the reality is you should explore less. So how much should you explore? I can't tell you that in neuroscience, can't tell you that. But the reality is you should build exploration

into your decision-making. Now there's research that supports this. There's research by my former PhD student, Cameron Hassel. We published a paper called, What Do I Do Now? An electroencephalographic investigation of the explorer exploit dilemma. We published in a journal called Neuroscience in 2013.

And basically in this experiment, we use something called the balloon pumping task. I think it's got a fancier name than that, but let's just call it the balloon pumping task. And basically what you see on the screen is a balloon. Now, typically it's a circle. Our programming skills in my lab aren't good enough to draw actual balloons, so we used a circle.

And basically you have a choice to pump the balloon up or quit and take the money. Now the key is every time you pump the balloon, you gain more money, but there's a chance the balloon will burst. And what we did in this is when we studied this, we found that people would make a bunch of rapid pumps in a row, boom, boom, boom, boom, boom, and then they would pause. And those rapid pumps we believe were exploitation. This was just going highest values to pump, highest values to pump, highest values to pump, highest values to pump. And then every once in a while

people would pause and we believe that's when they were thinking about exploration. They were making the decision, should I pump the balloon or should I not pump the balloon? All right. And what we did is we measured brainwave activity, EEG, while people were doing this. And again, it's a bit hard to describe the results, but on the slide decks, if you look at figure A at the top left, and you look at the brainwave response, this is basically two decisions to explore or exploit. There's this massive difference when people are exploiting and exploring, and we were able to localize it to a couple key parts of the brain.

Sort of part of it was pre-multi-ripped, part of it was pre-frontal cortex. And this is your brain trying to decide what to do. So we were actually able to see the explore-exploit dilemma happening using brain waves. So again, we believe this is what's happening in the brain. Another study that I did with this was a study happening in the brain. Another study that I did with this was a study I did with Cameron.

I was the first author and was my post-doctoral supervisor, Dr. Todd Handy. It was called How We Learned to Make Decisions, Rapid Propagation of Reinforcement Learning, Prediction Errors in Humans. Now I'm not gonna spend too much time on this one, but in a nutshell, we had people make a series of gambles. So they had to choose between a red square and a blue square. And this is a choice. And people make a choice

and then they find out whether they win or lose. Now the key to this is exploitation is if you figure out that red wins you just keep doing red but blue could pay you more right you don't know and we gave people a series of squares so there was yellow green blue green red blue all these combinations and every once in a while we changed it all up all right so people were constantly having to decide to explore or exploit and we did some fancy stuff with it We use some computational neuroscience to model the task I was trying to show off to be honest that I knew this computational stuff But basically what it meant is that we could mathematically determine the you know

Whether people were exploring or exploiting. And the model predicted their behavior. So we were able to figure out, this is what people are doing, bunch of fancy math, like I said, not gonna go deep into that. But at the end of the day, if you looked at the brainwave response,

we saw that early in learning when people rewarded, they had this massive brainwave response to the reward that went away at the time of reward as people learned that, hey, this is the high value option. But what was cool and unique with our study was that we saw that once people had learned the high value choice, when it appeared there was this massive brainwave response and that brain wave response it grew. On the first trial it wasn't there because they didn't know about it but once they had learned that the item had value that

brainwave response grew and basically that was them figuring out the value so they could make the choice. They could decide to explore or exploit. So that gets us back to this model of decision making. Always choose the highest value option and then you have to sometimes explore. All right, that's a that you have to build that into the equation if you will. But it's still not perfect because guess what? There's a problem with value. I've always liked this one. It's called the cheeseburger dilemma. A McDonald's cheeseburger costs about 99 cents, right?

So what's a McDonald's cheeseburger worth? 99 cents. However, you know, I've got a long haul flight to Germany coming up in a couple of days, be about 10 hours from Vancouver to Munich, and I'm not the biggest fan of airline food. So if you were on that plane and you had a McDonald's cheeseburger, you could probably sell it to me for $5. I'd pay $5 for that 99 cent hamburger as opposed to the airline meal. Right. And it just shows that the value is hard to assess.

It depends on context. All right. I'll give you a better example of this, which is a common thing that happens. Let's say you're going to buy tires for your car. All right. And those tires are $800. But you find out that if you drive for 20 minutes to the other side of town, you can get the same tires for $700. So $100 off. And if you ask people, would you just get the tires at the place near your house or would you drive across town? Well, save a hundred bucks most people choose to drive across town

But if you frame this problem a little bit differently, you can frame it this way you can basically say Well, hey, you're gonna buy a new car and if you buy it right here at the dealership near your house It's thirty thousand dollars. But if you buy it right here at the dealership near your house, it's $30,000. But if you drive across town, you can get the same car for $29,900. It's $100 less because it's got that tire discount built in. Will you drive across town to buy that new car over there? No. A lot of people say they're not gonna do it. The reasoning is simple. They just say, it's only a hundred bucks, but it's the same hundred dollars.

So again, it shows that people's ability to estimate value is a little bit skewed. It's skewed by any number of things. And complex choices. Which degree program should you choose in university? Should you do law or medicine or history? How do you estimate the true value of these options? Should you leave a significant other? How do you estimate the value of the happiness of staying

versus the potential happiness of staying versus the potential happiness of leaving? So people, the problem with Huygens, or one of the problems with Huygens is that people are very bad at estimating value. But there's also a problem with probability, right? I sort of hinted at it in the sense of like, if people were really good with probability, casinos wouldn't exist, would be my thinking. But, and you can see this, you know, again on the slide deck I've put up a blackjack hand, you know, the player's got 12 and the dealer's showing a seven.

You know, what's the chance of winning this blackjack hand? What's the chance you should hit? Now some people, if you're really into blackjack, you actually know the odds, but in Vegas, you know, they do things to change the odds. They remove cards from the deck and things like this. They play with five or six decks at a time. You know, most games have chance of this way.

People are very bad at estimating probability. You know, they tend to be a bit overconfident in their beliefs. Another example that I've always liked is it's estimating the probability of dying. So here are ways you could possibly die from a shark attack, a car crash, a plane crash, a lightning strike, Ebola or a bee sting. So the rank the what's the most likely way you'll die? What's the least likely way you'll die? Shark attack, car crash, plane crash, lightning strike, Ebola, or bee sting.

Well, if you look at the actual statistics, you're more likely to die from a car accident than any of these other things. And the year this data was taken was about a 1 in 9,100 chance, a 1 in 9,100 chance of being killed in a car accident this year. Your chance of dying in a plane crash is 1 in 11 million. All right? So you're far safer to fly than to drive. And the ones in case you're curious, and if you don't have the slides, car is most likely than it's shark.

In your lifetime, one in 3.7 million. I guess you can put that to zero by never going in the ocean ever. Beasting or an American one in 5.2 million. Lightning strike one in 9.6 million. D dying in a plane crash 1 in 11 million, and dying from Ebola, and this was a year when there was a minor Ebola outbreak in the United States, 1 in 13.3 million, yet there was an Ebola panic. Interestingly, this is true of COVID too. You're still far more likely to die in a car accident than you are from COVID. You're actually more likely to die just walking across the street.

So the problem is that people are very bad at estimating value and they're very bad at estimating probability. So we have trouble with decision making. The reason we struggle is because we're not sure whether we should explore or exploit. Like, is this a time to exploit? Is this a time to explore? We can't assess value accurately, and we're even worse at estimating the probability of outcomes. So what do people actually do?

How do we actually make decisions if we can't do expected values that well? Well, it turns out we try. I'm a firm believer that our brain does try to estimate these things. It's probably just not that good at it. But there are some other options. And we've talked about this before. Daniel Kahneman made this famous.

He's got this great book, Thinking Fast and Thinking Slow. Basically fast decision systems are your gut hunches. We've talked about that on the podcast before. That's your just intuitive response, right? And slow decisions are these more analytical things that you do. And I'll give you some classic examples of this. examples of this, you know, what's two plus two? And four probably popped into your head right away. You know, complete the following phrase, bread and butter, for a lot of you, would have popped into your head. You know, and in the slide deck I've got some other examples, but for a slow decision, you know, what's 13,678 divided by 13?

Well, you're probably not, if you're getting that off the top of your head, well, well done, and you're probably a savant of some kind. But for most of us, you might be able to solve the problem, but you'd go, okay, long division, 13, and you draw the squiggly line, and you'd go 13,678, and you'd go, well, 13 goes into 13 one time and and off you would go. And you'd probably you'd come up with some kind of an answer but it's an

analytical decision so this is our prefrontal cortex engaging and doing this stuff. And what's interesting about this is if you look at like a game like chess, you can see both systems. If you're very good at chess, you know, you're probably playing with your fast system some of the time. You're just making very simple moves, opening moves, especially your first move is just boom, you know what you're doing. But you know, and if you've never played chess before, you might spend a lot of time on that

opening move because you're not sure what to do. And again, in the slide deck, if you look, thatduralscienceguy.com on the blog, there is a chess board in there somewhere. And if you're an experienced chess player, you know it's white's move. Well, you'll know that white has got the king in check. So you would actually know that it has to be black's move at this point in time,

and what would black do here? And away you go. And you would see very quickly that black is in a lot of trouble. If you've never played chess before, you might struggle with what you're supposed to do here. Now, if you want to compare these systems for fast decision systems, you know, fast thinking is very common. We spend most of our life on autopilot making fast decisions. It's easy to do. It can be emotionally biased.

It can be error prone. And slow thinking, on the other hand, is rare. You know, it's very hard and deliberate. It's your rational system and it's reliable. It generally gets you what you wanna do. Where in the brain does this all go on? Well, basically system one, this sort of gut hunch system, we believe lives in midbrain structures.

Things like the ventral striatum, the basal ganglia region, midbrain structures that are basically hardwired just to give you a response. So for a given stimulus, here's my response. Now it would involve other brain regions than the midbrain, but generally it's a very reflexive system. System two, on the other hand, is the prefrontal cortex, right?

That's the front part of the brain. We've talked about it a lot. This is your analytical system that engages when you're being challenged and you have to really think about things. And again, research highlights these things. Again, I'm pulling papers from my own group. We published a paper back in 2019. It was one of my former PhD students, Chad Williams, led the project. It was called Thinking

Theta and Alpha Mechanisms of Intuitive and Analytical Reasoning. We wanted to see if we could map these processes with EEG. And lo and behold, we could. We had people do a very simple decision-making task. It's the add one or add zero task. I won't go into a lot of details, but you can Google that. But at the end of the day, when people made analytical decisions, the EEG pattern we saw over the front of their head was very different than when they were making gut hunch

or intuitive decisions. So we could actually see with brain waves whether someone was making a gut hunch or a rational or logical decision. And we're hoping to do this in real time. It's something we're working on now. Imagine wearing one of those mobile headbands I've talked about.

And when you make a gut hunch decision, it beeps with one sound. When you make an hunch decision, it beeps with one sound. When you make an analytical decision, it beeps with another sound, just so you know the type of decision you've made. It'd be a fantastic warning system in a bunch of situations I can think of. So most of the time, we're on autopilot, right?

We're relying on heuristics, rules of thumb. This is our intuitive system. But sometimes the prefrontal cortex kicks in to help out. All right. I personally believe that it's not one system versus another. I think the way this works is you're wandering around on autopilot making intuitive decisions. And the prefrontal cortex is just keeping an eye on things. And it engages if needed and to the extent it's needed

because sometimes our logical system can figure something out pretty quickly and other times it needs to give considerable effort, right? So when you believe they're two separate systems, we call that dual process theory, I personally subscribe more to the continuous model, although as an academic I have to admit there's evidence for both sides. I just want to thank you again to the Neuroscience Club at Indiana University and a special thank you to

my new friend Braden for inviting me. It was such a great, great time coming back to Indiana, and I really had fun giving the talk. Just a reminder, there's the website, right? Thatneuroscienceguy.com. These slides are gonna be there. I'm gonna post them as soon as I'm hit stop record, so they'll be up for you to look at. There's also links to Etsy and Patreon. Don't forget, you can buy merch, all right?

You can donate to the podcast. The money goes to graduate students in the Craig Olson Lab, supporting them in their neuroscience training. Follow us on Instagram, X or threads, at thatneurosciguy. Jennifer's putting up lots of cool stuff on Instagram about the podcast. Check it out. And of course, send us a message on X or threads or even at thatneuroscienceguyatgmail.com because we really want to know what you want to know about the neuroscience of daily life. And of course, as ever, the podcast. Thank you so much for listening. Please subscribe if you haven't already. My name is Olav Kregelsen and I'm that neuroscience guy. I'll see you soon for another episode of the podcast.

Thanks for listening.