SPENCER: Slime Mold Time Mold, welcome.

SLIME MOLD TIME MOLD: Hey, Spencer, thanks for having us.

SLIME MOLD TIME MOLD: Hi, Spencer. It's great to be here.

SPENCER: So you've done something very unusual, which is that you've created a whole new paradigm for psychology, and some people have really been taking note of it. Adam Mastriani said the following, "It's the most provocative thing I've read about psychology since I became a psychologist myself. This is probably the first time I felt surprised by something in the field since 2016." That's a lot of praise. So what I'd love to do with you today is explore your paradigm for psychology. Also, talk about some of its elements, and some of the ways I agree with it. Point out maybe points of disagreement. I have my own ways of thinking about psychology. They're in some ways similar, some ways different. They're not formalized and written down yet, but I've been working on them for years, and so I think they'll be really interesting to explore.

SLIME MOLD TIME MOLD: Hell yeah.

SPENCER: Awesome. So let's start at the beginning. A paradigm, you argue, is supposed to have entities and roles. What do you mean by that?

SLIME MOLD TIME MOLD: Yeah, so a paradigm refers to a large variety of things, but we'd say something like a mature paradigm, or a scientific paradigm, that's going to get you anywhere, that's going to lead to progress, has to be mechanical. But mechanical is a word that gets tossed around a lot, and we think it's better to be more specific. What does it really mean for something to be mechanical? Our basic argument is you need to have some idea of what the building blocks of your corner of the universe are, what are the entities involved? And then you need to have some kind of very specific proposal, or at least the start of a specific proposal, for how they interact. And those are the rules.

SPENCER: So for physics, for example, what would the entities and rules be in some part of physics?

SLIME MOLD TIME MOLD: Well, not being a physicist, we'd say things like, you know, mass, charge, distance, and there are specific rules about how those interact.

SPENCER: It's a different formula about the way that charges affect each other across distance, and that kind of thing. Those would be the rules.

SLIME MOLD TIME MOLD: Yeah, entities and rules is a tiny bit of a slogan or catchphrase. If you unpack it a little bit, we'd probably say something like entities, properties, and rules. You need to have some kind of list, maybe not exhaustive, but you want the start of what kind of things exist in the universe, or the part of the universe you care about, what properties they have. Because particles have a mass and a charge, so those are more like properties. They're sort of distinct from the entities themselves. And then the rules about how those interact, how they change over time.

SPENCER: And why does psychology need this?

SLIME MOLD TIME MOLD: Part of the problem with entities and rules is that we're still working on trying to figure out how to convey it, because there's a sine qua non about it that is hard to put into language. It's like saying when you're building a car, why do you need to know how the engine works? I'm really not sure how to answer that question. The engine works because it is the interaction of many smaller parts according to some kind of fixed rules. And if you don't believe that, I don't know how you can think that the engine works. So understanding it needs to involve some understanding of what parts are under there and how do they relate? Does that make any sense?

SPENCER: Yeah, basically, you need to know very specifically what we are even talking about, and then what are the relationships between the things that we are claiming, right? But some people might argue psychology as a field already has entities and roles. You look at the Big Five Personality. Well, conscientiousness, is that an entity? Or IQ, is that an entity? And then there would be rules around this, like people who are conscientious, "Well, they're going to be more orderly. If you check their rooms, you're going to find they're more organized." And people who don't have high conscientiousness.

SLIME MOLD TIME MOLD: Yes, it also seems like they're abstractions. And again, this is one of those things that we're still working on, how to put it into language. Those are abstractions in part because you make a judgment about whether or not someone's conscientious based on their external behavior. It's like saying that a car is fast or that a house is big. Cars are fast and houses are big as a consequence of the little moving parts inside them. The systems that give rise to these situations or behaviors. And it's fine to say this car is fast, that car is slow. This car makes good gas mileage. This car is terrible, but they are that way for a reason. And while abstractions can be useful and maybe even profound, they're not mechanical understandings of these issues. So whatever makes people extroverted or introverted or conscientious is something inside their mind, maybe inside their brain, depending on how you think it's best to analyze it, but you can't get very far appealing to abstractions alone.

SPENCER: So would you argue that most of the concepts in psychology that are used today are actually abstractions? They're not entities or rules.

SLIME MOLD TIME MOLD: Absolutely. In fact, I'd say I would challenge anyone to say anything that's mechanical in Psychology Today. I'm not closed-minded about it, but if there's a concept in psychology that people think is mechanical, I'd love to hear what it is.

SLIME MOLD TIME MOLD: Let me give you some background on this project, too. We started writing about cybernetics after we wrote a book about obesity, where we basically started thinking about human behavior in terms of thermostats and set points. We couldn't stop; it's a really kind of addicting idea, and we started seeing it in so many different places. The Mind in the Wheel, which is our newest book available on slimemoldtimemold.com, is really just exploring those ideas much further and asking a bunch of questions.

SPENCER: So your paradigm is a cybernetic paradigm for psychology, and you're saying that through your obesity research, you started using a cybernetic paradigm, and then you started saying, "Oh, wait, this is a really powerful idea that has much broader applicability than obesity." Is that correct?

SLIME MOLD TIME MOLD: Yes, exactly. It's very well said.

SPENCER: We're not going to focus on the obesity research here, but do you want to talk about the obesity research and how this came into play?

SLIME MOLD TIME MOLD: Sure. It was the pandemic, and we thought, "We should collaborate on stuff." We started a lot of different projects, and some of them went nowhere, but one of them was pretty interesting, we think. It was taking a look at the obesity epidemic in terms of a couple of perspectives: one unusual and one not so unusual. The not so unusual one is the idea that weight regulation is homeostatic. That it's cybernetic. This is something that laypeople don't usually buy into, but this is the dominant perspective in the research community.

SPENCER: Could you unpack that? Because I don't know if people know what that means.

SLIME MOLD TIME MOLD: Sure, the idea is that, and obviously, in reality, it's much more complicated than this. But the idea is that your body has a set point, a weight that it wants to be at, and it defends it vigorously in both directions. If you eat a lot or do things to gain weight, the body will take action, like making you raise your body temperature, making you fidget, driving you to eat less, to bring your weight down. If you weigh less than you need to, then your body will do things like make you use less energy and make you try to eat more, to try to defend it in that direction. And from this perspective, what happens in obesity is that, for some unknown reason, the set point is raised. So while your body's natural set point might be 180 pounds, if you were to reach a hand into the brain and turn that thermostat to 300 pounds, suddenly you're going to gain a lot of weight, and it's going to be really tightly defended.

SPENCER: If you look at shows like The Biggest Loser, where they would have people compete to lose weight, a remarkable finding is that years later, after people had lost tons of weight on the show, many of them regained the weight, or most of it. And so that kind of suggests maybe there's a set point kind of thing going on here, that the body's kind of defending a set point.

SLIME MOLD TIME MOLD: Yeah, and the set points defend in the opposite direction too. When they do overfeeding studies where they have people intentionally eat 10,000 calories a day, people do gain weight temporarily, but they lose it very quickly, going back almost to right where they started. And so that and a bunch of other evidence has convinced the research community that, for the most part, weight gain and obesity seem to be the result of a change in your set point. And so we combined that relatively orthodox perspective with a slightly unusual one, although we weren't the first to suggest it, which is that the obesity epidemic might be caused by some kind of contaminant in the environment. So this has been proposed before, but it hasn't been investigated. We thought there was still more to find out, and so we looked into it. And then the only really original thing that we contributed was we looked into lithium, in particular. We think that there's a good chance that lithium might be the contaminant, or at least there's evidence to suggest that lithium might be the contaminant behind the obesity epidemic, that lithium, or possibly some other contaminant or contaminants, are the factor that is changing people's set points so that they defend a higher weight than normal.

SPENCER: Yeah. For anyone interested in that in particular, you can go read a long series of articles about that that they've written. They're very interesting. I wasn't persuaded, but maybe we could talk about it in another episode. I do think it's very interesting, and I think you make a lot of interesting points there. But, okay, so then, just to clarify further, the word cybernetic, what is that referring to here?

SLIME MOLD TIME MOLD: Yeah, so cybernetics is the interdisciplinary field, or at least was conceived as an interdisciplinary field studying negative feedback or control.

SPENCER: Right. So it's basically this idea that the body or the mind wants things to be at a particular level, and then if they're deviating from that level, the body is going to detect an error signal and then try to get back to that level, is that?

SLIME MOLD TIME MOLD: Right. No, much more. That's correct, but it's much broader than that. So the earliest work in cybernetics, although it wasn't called that yet, was in mechanical control systems. So I think Christian Huygens worked on controlling the speed of windmills — I can't remember his name — but centrifugal governors or, in general, mechanical governors were essential to the development of steam engines. If you have a steam engine, you want it to be hot, but not so hot that it explodes. And this is actually a pretty tricky balance to strike, and so they used mechanical governors to regulate the temperature of their engines, making it so that they could keep their steam engines running without blowing up. If you're running a boiler, you have to be pretty careful about it. And so the earliest work in this was actually mechanical, and there was some early mathematical work in it by James Clerk Maxwell, and then Norbert Wiener and Arturo Rosenblueth ended up sort of pulling this together. I think Wiener used it in some of his, I think it was anti-aircraft gun designs. And they were like, it seems like there are a bunch of principles here which sort of exist across fields. There's definitely some control problems in engineering, but also there are control problems in biology and there are control problems in psychology. So let's make an interdisciplinary field that covers all of them, so we can sort of compare notes and see if anything comes of it.

SPENCER: So from this broader field of cybernetics, what are you bringing into your cybernetic theory of psychology?

SLIME MOLD TIME MOLD: Yeah, so in some ways, I think it's deceptively simple, but I think that's often how science works, right? It doesn't seem like rocket science for Sir Isaac Newton to say he thinks that the laws of motion on earth and in the heavens might be the same laws. That's not really a galaxy brain moment. It's just a change in perspective. But it's a change in perspective that's really important. Similarly, it's not like there are some very particular technical observations that we're bringing into psychology. We're just saying, "What if you start thinking about things in terms of negative feedback? What if you start asking questions because that suggests very specific questions?" As soon as you start saying, "What parts of this are negative feedback?" You start asking things like, "What's the rate of feedback? What's the set point? What is being controlled?" We think that those questions are really generative.

SPENCER: And just for the listener, could you explain the difference between negative feedback and positive feedback?

SLIME MOLD TIME MOLD: Sure. So in negative feedback, you have a number, an error signal, which the system attempts to send towards zero. It's a bunch of different parts that act together to take some number and try to send it to zero. In positive feedback, you have some number or value, and it compounds upon itself instead. So it gets bigger and bigger, and it goes towards infinity, or as large as it can within the limits of the system.

SPENCER: So then you want to say that a number of elements of the human mind have a kind of negative feedback style, which is distinct from claiming they have positive feedback. And there are also a lot of other claims that can be made that are neither positive nor negative.Is that right?

SLIME MOLD TIME MOLD: That's right. It's not that psychology has normally been built around positive feedback, and we think negative feedback is better. It's more that psychology hasn't really been examined in these terms or thoroughly examined in these terms. We think that there's a bunch of low-hanging fruit that you can pluck as soon as you start asking these questions.

SLIME MOLD TIME MOLD: I think staying grounded in the metaphor of a thermostat is really useful for thinking about cybernetics. We just talked through the weight set point. But you can also be like, "Okay, I have a thermostat in my body for salt, another one for sleep, and another one for social things — maybe the number of people looking at me. I get both upset if more people or less people look at me because I'm trying to hit my set point."

SLIME MOLD TIME MOLD: Yeah. Another way to answer your question, Spencer, might be by going back to entities and rules. If psychology is, as we argued, largely abstract, and you need some kind of specific mechanical proposal to bring it into being a mature science, then definitely one of the benefits of cybernetics is we can say, "Here's an actual mechanical system. You can literally have machines that do simple versions of these tasks," and we're going to bring these in as a proposal for the kinds of entities and rules that are going on behind the scenes. When you see something like someone cleaning up their room or not cleaning up their room, when we have these abstractions like conscientiousness, we can say, "Is it possible to build a mechanical theory using these mechanical parts?" While it's not the only proposal for the entities and rules behind the human mind or behind minds in general, it is one of the very small number of proposals we're aware of that seem like they could be functional at all.

SLIME MOLD TIME MOLD: One of the most interesting things about working on this project is that responses generally fall into two categories. People are basically like, "Yes, cybernetics, I've already heard of it. What's new?" And then the other half of people are like, "I've never heard about this before, and it makes a lot of sense."

SPENCER: Yeah, it's a funny kind of thing, where sometimes when someone proposes something new, people can't agree whether it's obvious and trivial.

SLIME MOLD TIME MOLD: Old and boring.

SPENCER: Crazy [laughs], yeah, exactly. So it can be very polarizing, and just to clarify, when you say mechanical, do you mean you could literally write down a series of rules that it follows, or do you mean something else by mechanical?

SLIME MOLD TIME MOLD: So I think that there are different degrees of mechanical, and this is part of why we use the phrase entities and rules more than we say mechanical, although there are reasons to use both. I think at the most basic level, mechanical is just looking at surface level behavior and trying to say, "What parts underneath are causing that behavior?" So instead of just being like, this car is fast, this person is conscientious, you say, "What parts of their mind are giving rise to this conscientiousness? Why are some people conscientious and not others? What's creating this behavior?" And then, more specifically, you can try to have particular proposals for entities and rules. Does that make sense?

SPENCER: Yeah, it does. The one thing it makes me wonder about is it seems like you've got to bottom out somewhere, right at the bottom. You have to have something that doesn't have a causal explanation.

SLIME MOLD TIME MOLD: Yeah. Great point. But I think that this is true in every science, right? You can bottom out physics at some point and be like, "These are the rules as low as we've found them." And you can be like, "Great, what caused the rules?" And they have to shrug their shoulders and say, "You know, who knows?" I think that's just irreducible. You can always push it down to another level of what's happening under the hood here. And everyone is going to, obviously, push it as far as they can. But making it mechanical at all is what leads to the benefits of having a mature science.

SLIME MOLD TIME MOLD: To me, the idea of a paradigm is about pointing to what the most interesting questions are. I think that one of the reasons cybernetics is so compelling is because it highlights how understanding drives can help us explain human behavior.

SLIME MOLD TIME MOLD: There's another point here, which is, I think you're right to ask, "Is this going to give us the ability to actually write out a series of rules that are testable?" You said something along those lines, and I think the answer is, eventually. This is part of the reason that we start by talking about the history of chemistry. In the early days of chemistry, they were not able to describe quantifiable rules. They eventually got there by starting with mechanical theories of chemistry; they started postulating ideas about what was going on beneath the surface. Eventually, if you pursue that far enough, you do get the ability to write down very specific rules. We don't claim that we can do that yet. We just think that this is the first step on the path towards being able to do that. Aspirationally, yes, you can sit down and write down the rules, but that requires a lot more research.

SPENCER: Got it. So what are some of the entities?

SLIME MOLD TIME MOLD: The entities are governors; they're control systems. We can say some things about the parts. In a basic, sort of stereotypical governor, you have an input function which creates a percept, a type of signal measuring the thing you care about in the world. Maybe I should explain it in terms of the thermostat.

SLIME MOLD TIME MOLD: For anybody who is listening, I would really recommend going and looking at part one of The Mind in the Wheel because I definitely think that the visual aids are very helpful if you are not familiar with cybernetics.

SPENCER: Yeah, and we'll link to that in the show notes.

SLIME MOLD TIME MOLD: Fantastic. So yeah, stereotypical world's most simple thermostat. Obviously, in reality, you can make much more complicated control systems. The human mind, at least, is biological, so it's going to be complicated in all these additional ways, but starting with the stereotypical basic thing, a thermostat cares about the temperature of your house. Your house is some temperature. The thermostat has an input function, some kind of mechanical device that detects the temperature of your house, and it turns it into a signal. That signal is sometimes called the percept, so it's an estimate, a measure of the temperature of your house. Separately, a human being has set the set point, which is the desired temperature of your house, let's say, 72 degrees Fahrenheit. Then there's a component called the comparator, which is really simple. It compares the percept to the set point and asks, "Is the house at the temperature that I want it to be? And if not, what's the difference?" The comparator compares those two numbers, and it creates a new signal, which is the error signal. It's literally just the difference between the set point and the percept. If the house is too warm, it'll be above, and if it's too cold, it'll be below. The error signal is passed to an output function, which can be very simple; it can just be, if there is an error, if the house is too cold, you turn on the furnace. Output functions can also be very complicated. Obviously, in humans, for us to do something like control our social status, or even just control our blood sugar, the output function is really complicated. We have to do a lot of stuff to accomplish those goals, but it passes it to an output function which takes some kind of action on the world. If the thermostat is built well, then the output function will do something that will close the loop, so it will reduce the error signal towards zero, and it will hopefully keep it there.

SPENCER: Okay, so let's apply that now to a specific human governor that you hypothesize.

SLIME MOLD TIME MOLD: Sure. We can go back to our obesity research. We think it's more complicated than this in reality, but assuming that there's a set point for weight, if your weight is set to 300 pounds — obviously it wouldn't be in those units — and there's a system that exists for regulating your body weight, then it's going to defend that. If you go on a crash diet, it will make you really sluggish. The output function will take actions, too. We think of it in terms of the output function voting against you doing things like going for a run; it will vote against you getting up off the couch. If you reduce your calorie intake, it will take steps to reduce your calorie output. It will do things to defend your body weight. Conversely, if you're at a lower weight and you overeat, your output function will vote for things like fidgeting. It will vote for things like jumping around the room, going for a run, going out to see people.

SPENCER: But people might think, okay, so you're kind of talking about these very biological things. They may have trouble connecting to human psychology. So what's an example that really is fundamentally psychological in nature, where you can still see this idea of a governor?

SLIME MOLD TIME MOLD: Eating is psychological.

SLIME MOLD TIME MOLD: Yeah. We see this as fundamentally psychological. It's hard to know how to make this case because eating is a behavior. Where do you decide to eat from? Where do you feel hunger? This is all in the mind. You need to use your brain to be like, "Am I hungry? Okay, let's make a sandwich." That doesn't happen separately.

SPENCER: Yeah, and I agree with that. On the other hand, it's a lot easier, I think, for people to immediately see it for hunger, and then it works that way. Then let's say an emotion.

SLIME MOLD TIME MOLD: That's why we use that as an example to ground the rest of the paradigm in.

SLIME MOLD TIME MOLD: Yeah, we also think hunger is an emotion, and obviously, this is at some level semantic.

SLIME MOLD TIME MOLD: Yes, plenty of people don't like us calling that using the word emotion. So if you feel more comfortable thinking about it as a feeling, that's fine, but hunger is absolutely a feeling.

SPENCER: Right. So let's do another feeling that maybe seems more psychological in nature. Even though I agree with you, hunger, there's obviously a strong psychological component there. But I don't know what a good example would be, fear.

SLIME MOLD TIME MOLD: Fear seems like a great example. Here's an example. Have you ever fed sparrows in the park?

SPENCER: Feeding them bread crumbs? Yeah.

SLIME MOLD TIME MOLD: Yeah, exactly. So unless they've been fed a lot, you see a counterplay between two of these drives. Obviously, the sparrows are hungry. They want to eat the bread, but they're also afraid of you, and they don't know what you're going to do. So if you scatter the bread crumbs too close to your feet, you'll see a bunch of these sparrows reach an equilibrium. They will come and approach you, but they will hold off at a certain distance before eventually approaching slower and slower until they grab their bread crumbs one at a time. You can interpret this however you want, but we interpret this as a tension between at least two governors. One governor is saying, "You should walk forward, get the food, you're hungry, eat it." And another governor for fear, which is saying, "Don't approach the big, strange, scary animal that's huge and could kill you, might be hunting you." These reach an equilibrium, which for a while leads to inaction or hesitation.

SPENCER: And what's the case that fear actually works on a governor system where you've got these signals, a set point, a comparator, an error signal, and an output function rather than being modeled by something else?

SLIME MOLD TIME MOLD: What's the alternative? I guess maybe we'd say something like, and this is one of the difficulties when talking about paradigms at all, because at some level, a perspective is a matter of faith, at least until it's been supported. Even things like conservation of mass. The early chemists were like, "We're going to start by assuming that mass is conserved. This ends up being a really useful assumption, even if you eventually do end up in some places where it's no longer true." For example, they were saying, "We're going to consider things to be an element if they can't be divided, if they don't have any smaller pieces." This ended up being a really useful perspective, even though it ended up actually not being true. When they discovered isotopes, that was a big shock to the system. But thinking about things in terms of elements continued to be useful even after the definition of element fundamentally changed.

SPENCER: So in your paradigm, you want to say that things like fear have a governor. And so they work through negative feedback. What does this actually imply about fear, rather than a much less detailed model of fear? Someone might just say, "Oh yeah, we have this fear when there's something dangerous, we feel fear, and we try to avoid the thing that we're afraid of." But you have a more precise model. You're saying, "No, fear is actually controlled by a governor system. A negative feedback loop." So what are you actually claiming about fear that a simple model wouldn't claim about fear?

SLIME MOLD TIME MOLD: That's great. I think first we want to say it's good to have a model at all, right? I don't think that there's any competing mechanical model for fear out there, not that I'm aware of. People are just discussing it abstractly. So saying, "Here's a potential mechanical model is good," but you're right, there are more specific claims. There are claims like everyone's going to have a set point. That means there's going to be a point, and it may not be a strict threshold, but there's going to be some kind of point past which fear is the error signal in this case. There's going to be a point of danger past which you see behavioral changes start to kick in.

SPENCER: Fear corresponds to the error signal when there's a gap between the set point and what's actually measured. Is that right? That's how you think about fear?

SLIME MOLD TIME MOLD: That's right. You should see things like, in principle at least, it would be possible to discover what the set point is. You'd be able to show that different people have different set points, or that maybe being in different situations, being exposed to different substances, changes your set point or reduces the error signal. It also suggests the error signal might eventually be quantifiable. It suggests other things which are not necessarily linked to it being a governor, per se, but it suggests that fear is meaningfully distinct from the other values being controlled, and it will compete with other motivations and trade off against them.

SPENCER: So with obesity, we could think about the set point as being a set point in body weight or something like that. For fear, what is the thing that is being set? Is it the danger level?

SLIME MOLD TIME MOLD: Yeah. So we'd say there's some kind of perceptive danger. And obviously, the question of how that's generated is a good and open-ended one, and it's probably very complicated. That's compared to some set point for danger. When danger is above the set point, or possibly below, you never know how these things are actually wired up. When it is past the set point, you start generating an error of fear. As fear gets bigger, it exerts a larger and larger influence over your behavior.

SPENCER: So then, would this assume that some people actually have an amount of danger that's above zero, that they're sort of trying to get to in some way? If they're at zero danger, there actually is some part of them that's trying to increase danger level.

SLIME MOLD TIME MOLD: Yeah, great question. We think so, although we also think that this would likely be a different governor. There are neurological reasons, and I think phenomenological reasons, to suspect that when you have control on both the lower and the upper bound of the signal biologically, these would be different governors. We think, for example, there's at least one governor keeping you from getting too hot, and a separate governor keeping you from getting too cold. While fear makes it clear that we have a governor keeping us from getting into too much danger, there may actually be a governor that encourages us to go out and get into a certain level of danger, but that's an open research question. "Is there actually a danger governor?" It certainly seems like there might be. People definitely have behavior where they go out and put themselves in danger, but that might be coming from a different source of motivation. If you know you can get status by showing that you're not afraid of danger, maybe some kind of status governor is voting for that instead.

SPENCER: It could be the case that the set point of danger for everyone is zero, and all they're trying to do is kind of get down to the zero level. Or it could be that there are individual differences, where some people are set at zero, but for some people it's above zero, and they're more danger seekers kind of thing.

SLIME MOLD TIME MOLD: Yeah. So fear is going to be a governor that's trying to keep danger below some level, probably not zero, since we don't seem to be afraid when we're just walking down the street, even though we're in some danger. Although obviously, we don't really know how the mind constructs its perception of danger. In that, I think, basically, well, there are some people with knockouts. Basically, everyone tries to keep their danger level below some amount.

SPENCER: I see. So it could be some small amount that you're below that amount, you just don't feel any danger. The danger of walking down the street normally. There's always some danger. You could get killed by a car or bike or something, but most people don't feel afraid because it's a small enough danger level.

SLIME MOLD TIME MOLD: Yeah. Similarly, there's a thermostat, and it's trying to keep your house above, or no lower than 72 degrees. It's not generating an error signal when your house is at 80 degrees because it's above that level. Or a steam boiler. It's got a governor that's trying to keep it below, I don't know, a thousand PSI. I have no idea if that's a reasonable number, a thousand units. When it's below a thousand units, it's just totally inactive. Similarly, when you're below a thousand units of danger, your fear governor might be like, "I don't care." Then when it gets above a thousand, you're like, "Okay, we're generating our signal that is perceived as fear. We're going to start trying to exert some kind of influence on behavior to try to put us in less danger." There might be, like you suggest, a separate system trying to keep you above some level of danger, but if so, we think that would be a different governor.

SPENCER: Now my understanding is you want to apply this to many aspects of human psychology, right? So we talked about fear, we talked about hunger, but how many are we talking about? Are we talking about hundreds of these potentially?

SLIME MOLD TIME MOLD: Yeah, potentially. I think it's a lot like the search for the elements, right? You start out with four, and you're like, "Maybe there are more. How many more? Turns out, more than a hundred, and it definitely seems like people have dozens of different sources of motivation. We've had dozens of different things we need to do to keep ourselves alive. And maybe there are hundreds, perhaps thousands. We think that probably our intuition is that the mind is ultimately going to be reducible to a finite number of motivations. But it's also possible that that number is really, really large. The fact that we have a sort of keystone things like, "Don't get too cold, don't get too hot, make sure you drink water, don't let a lion eat you," suggests that there is, in fact, a relatively finite list. But who knows?

SPENCER: And you mentioned in your writing that even hunger might actually have many different governors. Could you explain that?

SLIME MOLD TIME MOLD: Yeah, so this seems like one of the strongest cases for the idea. There might be hundreds and hundreds of different governors. It's easy to think about hunger as one drive, but it's pretty clear that it's not when you look at it more closely. Have you ever had the dessert stomach effect, where you eat a big meal and you say, "I'm so full," but then somebody says, "Do you want this chocolate lava cake?" And you say, "Absolutely, I would love some." This suggests that even though a bunch of your hunger governors are exhausted or fully satisfied, you still have other drives for other things. A better way of saying this is, if you put out a salt lick, deer and moose from the local woods will come out and lick the salt. It's clear that this means they must have a desire for salt. They have some kind of psychological drive for it because they're not getting anything else out of licking the salt lick. For humans, it seems like we definitely have a drive for, if you have an empty frying pan in front of you, you will add butter, but you don't just eat the butter, even though it's a source of calories. Then you add some bread, or you add some salt, you add some chicken, right? It seems like there are a bunch of different, at least a couple different hunger drives, and maybe dozens and dozens. If you're pregnant, you'll get all kinds of weird cravings. Or if you're a vegetarian, you will get deficient in certain things.

SPENCER: Yeah, it's interesting because it's scientifically debated to what extent people actually have cravings for things they need. Let's say you're low in vitamin C. Will you crave citrus fruit because it has vitamin C? I think there's actually a lack of consensus on this right now, or at least a lack of consensus on how far it goes. Everyone kind of agrees that there's a general hunger drive we have, but not everyone agrees it could get to the specificity of the level of a single vitamin.

SLIME MOLD TIME MOLD: Yeah, it's a great question. We're kind of surprised that there hasn't been more work in this direction. Maybe it's just because once you frame it this way, it does start seeming like a really fundamental psychological question. We also think that you can come up with methods to investigate these things. It is possible to design studies where you can answer questions like, "Does an animal deficient in vitamin C actively seek it out? Or can it not tell the difference between water fortified with vitamin C and water without?" You can start getting very objective or very clear answers to very basic psychological questions. Maybe it's because it's normally seen as nutrition, but we think that these are basic psychological questions.

SPENCER: Now, fear, to me, seems easier to fit in your model. But what about something like anger? How does that actually fit in? What is the thing that's being set with the set point?

SLIME MOLD TIME MOLD: Yeah, great question. The thing that makes this easier to interpret is remembering that, at least in our model, all of the emotions are error signals. In this case, you have to say,"What aspect of your life is out of alignment, not under control when you're angry?" I would say that this is something like social control. When you are not in control of a situation, then you tend to get angry.

SPENCER: You're saying it's sort of social control. Is the thing that you want to have at least a certain amount of it? When you get below that, then you get this error signal. The error signal is like anger itself. Then your body takes action to kind of get you back into social control.

SLIME MOLD TIME MOLD: Exactly. And it's definitely possible that anger does seem one of the less immediately applicable. Anger seems to be one of the worst fits in terms of this model.

SPENCER: [laughs] [inaudible]

SLIME MOLD TIME MOLD: But actually that seems really productive to us. The fact that it's a bad fit for this suggests that it must be something else. And so then you start saying, "Wow, unlike all these hunger and unlike fear and unlike maybe social status, and unlike feeling hot and feeling cold, anger seems kind of weird. Is anger a different kind of thing psychologically?" This is not the kind of question that people normally ask in psychology. So this is why it seems really generative. Can we look at anger and say, "Yeah, actually, it does make sense to see this as an error signal, like these other ones?" It's homeostatic. There's some kind of anger governor. Or is anger really a different class of signals that operates on different rules? It's a fascinating question, and I think we would continue to sort of err on the side of ultimately, it can be understood as an error signal coming from a governor, just because we think that that's sort of the strongest null hypothesis for any of these emotions. But it might not be.

SPENCER: One of the things I really like about your paradigm, because there are aspects of it that I agree with, aspects I disagree with, but what's really nice is you're making crisp claims that then we can actually talk about, "Okay, well, does anger fit this model? And if not, there should be an experiment that can be conducted to kind of refute that it works this way."

SLIME MOLD TIME MOLD: Exactly. And to circle back to entities and rules, this is the reason that mechanical science is necessary, and this is why you have to get all the way down, or as far down as you can, into the entities and rules if you want to get those benefits, because it forces you to construct specific claims, which means you get specific questions that are super generative, and it means that you suggest all kinds of questions, such as the question you're asking about anger, whether you intend to or not. And so it's just tossing us all these really productive research directions and only questions, such as, what do we go after first? What's easiest, what's most tractable, what's most interesting?

SPENCER: This might be a good moment to talk about a way that I disagree with you, which is that I actually think your model works really well for things like hunger and keeping the body temperature regulated. I think it's very much like a thermostat. I think my own view is that typically called emotions — not emotions in your theory — but things like anger, fear, sadness, and things like that, I think of them quite a bit differently, as not trying to keep some kind of set point, but rather, I think of them more like your ears and your eyes, the way that, through evolutionary history, eyes evolved, ears evolved, and each of them evolved to solve a set of problems. With eyes, you can see food, which makes it easier to find, or you can see predators, which makes it easier to get away. Ears solve an overlapping set of problems, but they're different. Eyes and ears are different. And ears are different, but they're both hyper-specific in a way. Eyes are there to perceive light, and ears are there to perceive sound. I think of the emotions as being like that, that they're these very specific, evolved mechanisms that are trying to detect particular settings. When this kind of setting is detected, you experience the emotion, and at that moment, there's a bunch of behaviors that change, and the stronger you feel the emotion, the more it pushes those behaviors into that different mode. The reason it does that is that pushing the behavior to that different mode in that kind of setting gives you a survival advantage. For example, the way that I think about anger is that it typically occurs when you feel like there's a particular agent that is threatening things that you value, and when your brain detects that kind of setting, you start to feel angry. As it ramps up more and more, there's a predictable set of anger behaviors that start getting activated. The more you feel angry, the more those behaviors get activated. When you're extremely angry, there's a very stereotypical set of behaviors that people engage in when they're angry, such as shouting, clenching fists, and being much more prone to start physically fighting or threatening, etc. So that's how I think about it, which I think is a bit different. I'm curious to hear your comments on the way I look at it versus the way you look at it.

SLIME MOLD TIME MOLD: Yeah. So that doesn't seem totally incompatible, right? There's no reason that the — what did you call it — the medical thing that's being measured.

SPENCER: So I think of it as there's some part of your mind, not necessarily a physically distinct part of your brain, but a part of your mind that's kind of scanning the world for this type of situation, and in this case, for anger. The way I've described that type of situation, it's monitoring for whether there's an agent that's threatening things you value, and when that's detected, you start getting this anger signal. That anger signal comes with a characteristic set of behaviors.

SLIME MOLD TIME MOLD: I think that can definitely fit with a lot of my understanding of cybernetics. It stands out to me how much you're using the word setting and you were talking

about different modes, which resonates with me, and it sounds like maybe you're proposing that for the anger thermostat. It's a feeling of other people having more power than you. Is that what is controlling?

SPENCER: I would say, threatening things you value, which is slightly different.

SLIME MOLD TIME MOLD: I think that matches quite nicely with lived experience in terms of the thing that I see people get the most angry about is definitely politics, which really does speak to that power differential and feeling like other people have control over things that you value.

SLIME MOLD TIME MOLD: It's definitely an open question whether it is constructed in a negative feedback sense, but I don't think that there's any reason that it's necessarily in conflict. You can have a governor that's detecting something. We don't need to pin it down exactly, comparing it to an acceptable level of that thing, outputting a signal which we interpret as anger, or we experience as anger, which drives certain behaviors. I think maybe one of the things here, or at least the way you describe it, I don't know if this is something that you feel strongly about, is the idea that all of these behaviors are innate rather than learned, but I would expect that even anger responses are highly socially determined. People from different cultures, even different subcultures, will express anger in very different ways, even if it always ultimately leads to physical violence.

SPENCER: Yeah, so if I try to kind of put a point on in what way do I actually disagree with you, in what way does what I said not fit into your model? I think that the key thing is, I don't think there's a negative feedback loop, which I think is key to your model. If it's not key to your model, then I'm not sure what is key to your model. So would you say that that is really essential?

SLIME MOLD TIME MOLD: Why do you think that it doesn't have a negative feedback loop? When you are angry, you take actions, and those actions make you less angry, right?

SPENCER: No, I don't think that's the case. I think they can often make you more angry, actually. A standard thing that happens in anger is that you take actions that actually make the other person angry, and then you both keep making each other more and more angry, and eventually, you get in a physical altercation. I think of that as very characteristic of anger.

SLIME MOLD TIME MOLD: Although, eventually anger goes to zero, right?

SPENCER: Oh, sure, sure. But I think typically, because you're no longer detecting that there's a threat to your values occurring from an agent that you can interact with.

SLIME MOLD TIME MOLD: But that's what I mean. You detect that there's no longer a threat. This is just like anything else being homeostatic. You detect that the temperature is now in line with or above or below the threshold.

SPENCER: I think a key thing here is, how would you program an agent? Imagine you're making a computer program for a little simulated world, like Adam Mastroianni brings this up in his blog post about your work of programming the Sims. It's funny because I use that exact kind of way of thinking. When I think about psychology, a lot of it is, "Okay, you're programming a little SIM creature to behave as much as possible like a human. How do you actually write the code?" If I understand your paradigm, the way you write the code is that there's a thing that's being measured, there's a set point for it, there's an error signal, and then it tries to get back to the set point. The way that I would write the code, the way I describe this for anger, is more like there's a sensor. It's a sensor first, or a very abstract quantity, like there's an agent that's trying to threaten things you value that you can interact with, and then basically as that signal ramps up, there's a specific behavioral vector that affects your normal behavior pattern vector and changes it. The stronger it is, the more that behavior vector changes the current vector. So I think the code is actually different.

SLIME MOLD TIME MOLD: That doesn't sound different at all to me. Or at least it certainly could be implemented as a control system.

SLIME MOLD TIME MOLD: Yeah. I don't see anything in there that feels contradictory to me.

SPENCER: If you think about the idea of modeling the motion of the planets. So back in the day, people didn't know the right way to model the motion of the planets. Some people said, "Maybe they're circular orbits," but that didn't fit the data very well. Then people figured out, Okay, if you make a circular orbit within a circular orbit, you can model it better, and you can do that if you keep repeating that, a circular orbit with a circular orbit with a circular orbit. It gets even better." People sometimes make fun of this idea as, "Oh, that's so ridiculous". But actually, it was an incredibly cool idea because it was sort of the basis of what we now call Fourier analysis, which turns out to be a very powerful and flexible way to model a signal. You can model pretty much any periodic signal, a repeating signal, using sideways within sideways within sideways. So it's a very powerful concept. Now, the problem with it was it was so powerful that it ended up just mathematically modeling things, not getting at the underlying reality very well. Once you switch to elliptical orbits, you could see, "Ah, that's actually much simpler, like Occam's razor; you don't need circles within circles within circles." But that doesn't mean that the repeated circles idea was a bad idea. It was a very flexible and powerful idea, almost too flexible and too powerful to actually see the underlying reality. So I guess what that makes me wonder is this idea of cybernetics: is it so powerful that it could kind of model anything? And if it can, then is it really making any claims? Since I don't think I would program, thinking about cybernetics, I don't think I would program that Sim to have anger in the same way I would based on my model. But then you're saying, maybe there is a cybernetic model that gets exactly the same output as what I'm describing.

SLIME MOLD TIME MOLD: This sort of gets to maybe a crux of the questions around where psychology is going and what you need. A lot of people, when they hear about this idea, come back at us and say, "This is interesting. I'm pretty convinced, or at least I'm intrigued." So what prediction does the theory make that we can go and test? I feel like your question is coming from a very similar position, and that's not how we think about it. We think that it's true that theories do eventually get to the point where they're very robust. You can make called shots; you can make a prediction that doesn't line up with anything seen before, but that's not what they look like in the beginning. A really good paradigm asks really interesting questions. It's really generative, and it asks questions in a way that lets you substantially disagree and go after empirical questions with a real sense of direction. That's what it offers. I think there is a sense in which the epicycles: you can always construct a way to explain things in terms of control, probably, for the sake of argument. But I think that's true of any paradigm. You can divide up the entities of the universe in any way you want. The question is, is it useful? Does it have a good ratio of simplicity to usefulness? Since we're relatively early on that path, we don't know for sure, but it's very exciting to start seeing things this way because it leads us to questions that seem super generative. Whether or not it ends up being the case that thinking about things in terms of governors is the place we will end up, it seems like a great place to start because we have these kinds of conversations. Do people have different set points for anger? If it's cybernetic, if it's a governor, then they have to, or for some reason, all have the same set point. What's the set point for anger? is a reasonable question, whereas, if you're right, maybe that's not a reasonable question, and you'd see something else.

SPENCER: Okay, so imagine I put forward a theory or paradigm for human psychology. I said, it turns out my theory is that humans are Turing machines, and now all we need to do is figure out what code we're running in our minds. For those listening who don't know what that means, a Turing machine is a model of computation, and anything that can be computed with a computer can fit into a Turing machine model. I would say that's not a very good paradigm because it can capture anything, but it also doesn't tell you very much. There are some people who think human behavior can't be modeled as a Turing machine. There is a group that says that. It's not claiming literally nothing, but it's such a general and powerful paradigm that it's making very little claim, and now you still have almost all the work to do, which is to describe, "Okay, but what are the computations the mind is doing?"

SLIME MOLD TIME MOLD: I totally agree. I think that's bad for two reasons. First of all, it's bad because, as you say, it's fully general. It can't really be disproven, because this, as far as we know, is just a feature of all computation. But I think it's also bad because it doesn't suggest any research directions. Even if you're absolutely on board, then you're left wondering, so what? Whereas cybernetics, at least as we've applied it, does immediately suggest a lot of different research directions. You have to ask things like, how many governors are there? What are they controlling? What are their set points? You can start asking questions about what their output functions are. Do different people have different gains on their output functions? We would not be at all disappointed if a hundred years from now, people had rigorously pursued this perspective and found out that actually, nothing in the mind is cybernetic, because it's not really a paradigm about how things will turn out. It's a paradigm about how to investigate these questions.

SPENCER: Okay, so could your paradigm be disproven? Or do you think that that's not the sort of thing one does to a paradigm?

SLIME MOLD TIME MOLD: Yeah, paradigms can't be disproven, because you can choose to divide up the universe however you want. It's just that some ways of dividing up the universe end up being useful and tractable, and others don't.

SPENCER: So it's more a claim about essentially a way of looking at things that you think will be a fruitful direction for the field, rather than a claim about the human mind fundamentally.

SLIME MOLD TIME MOLD: Precisely. And it certainly could be the case that this ends up being correct in some sense about the human mind fundamentally, but it seems like it's a fruitful direction.

SPENCER: Unlike a Turing machine, there are some things that can't be modeled cybernetically. If a lot of human behavior or action did not fit a cybernetic model, wouldn't that suggest the paradigm wasn't a very good one?

SLIME MOLD TIME MOLD: It's a good one if it gets us to a mature science of psychology.

SPENCER: Okay, so it practically might help, even if most human behavior can't be modeled that way.

SLIME MOLD TIME MOLD: Compared to most approaches to psychology that are available right now, this one is at least somewhat mechanical, and it seems really suggestive. If somebody has a better, more mechanical paradigm, or even just an alternate one, we think that would be worth pursuing too. But it's hard not to look at the juicy questions suggested by this direction and think, that's a good direction to try.

SPENCER: One thing I really like about your paradigm is that it does immediately suggest some things that are a priori surprising. For example, that happiness is not an emotion, according to the theory, which I think most people would say, wait, isn't it an emotion?

SLIME MOLD TIME MOLD: Yeah. To make the case for the idea that social emotions are, in fact, also negative feedback loops, I would say something like this: evolution is pretty parsimonious. If it has a piece that's working, I don't see why it wouldn't just continue to apply that elsewhere. You have really simple organisms that evolve to control things like the salt level and the sugar level of their blood, and how much danger they're in, really basic stuff like that, control their body temperature, and they have a brain that helps them do all these things. When it comes time for them to exist in a social network, or when it comes time for them to do other more complicated things, they already have this pretty sophisticated motivational system that trades off between a bunch of different negative feedback loops. Why would evolution bother to evolve this whole new set of things when it could just adapt a new control system, add another governor to control this other signal? Obviously, it might not do that, but it seems like a good starting assumption to be like, "Yeah, we have these governors. Let's just keep copying that code to support this new need."

SPENCER: But I suppose if it turned out, for example, that positive feedback loops are really useful for modeling some things, that would be one reason why it might deviate from that.

SLIME MOLD TIME MOLD: Yeah, potentially.

SPENCER: Let's go into happiness a little more. Could you explain how it fits into your paradigm?

SLIME MOLD TIME MOLD: Yeah. So happiness doesn't seem like the other subjective experiences. It doesn't seem like it's an emotion because happiness, successful behavior doesn't drive it to zero. You're not trying to reduce your happiness, whereas behavior is always trying to reduce your feelings of cold, your feelings of being too hot, your feelings of hunger, your feelings of shame. And so happiness seems like it's probably a different kind of signal because it doesn't seem like it follows the same rules.

SPENCER: How does it fit into your model, exactly? What is it in your model?

SLIME MOLD TIME MOLD: Yeah, so we suspect that happiness is a different kind of signal and that it probably is related to regulating explore versus exploit. This is a basic sort of trade-off in machine learning. If you have a number of good options, but you can also go out and look for other ones, do you explore and try to find new options that are better than your current lineup? Or do you exploit the best of your current set of options? So if you've moved to a new city and you've gone to just two or three restaurants, do you want to go out and explore the other 500 restaurants in the city? Or do you want to go to the best restaurant you found so far? In general, one of the things that happiness might be used for is to calibrate this. We think that happiness is generated through a function of when the other errors are corrected. If you're really hungry and you eat a sandwich, that makes you happy, or if you're really lonely and you see friends all of a sudden, that makes you really happy too. This is how we think it helps calibrate explore versus exploit.

SPENCER: So essentially, all of these other governors we're talking about could produce happiness as you correct the errors in them.

SLIME MOLD TIME MOLD: That's right, and it's probably not as simple as correct one unit of error, get one unit of happiness because, for example, eating a sandwich that you've had a lot, even though it's really good, doesn't seem to make people super happy. So it may have something to do with the difference between the expected correction and the actual correction. If you expect a large correction from something, it may not make you very happy, sort of a hedonic treadmill situation, but we think that this might calibrate explorer versus exploit, because in general, if your happiness is high, that means that you're correcting a lot of errors. You're very successfully navigating your environment, whereas if your happiness is low, it means that your strategies are not really working, and maybe you should try new strategies. So that's one proposal. It's also possible that happiness's main role is in learning. So this is a signal that we experience sort of as a side effect of when a new strategy is being stored. So if you do something that is good and you didn't know about it already, it makes you happy, and the happiness is sort of the signal that carries it to your memory, or it's affiliated with the signal that carries it to your memory.

SPENCER: I find this really interesting because something I've long wondered about is whether there really are multiple positive emotions. As far as I know, every single theory of emotions that's out there says that there are multiple negative emotions, for example, that anger and sadness are importantly distinct.

SLIME MOLD TIME MOLD: Yeah, everyone agrees they're different.

SPENCER: But it's really unclear on the positive side. If you think about positive emotions, feeling relaxed, feeling happy, feeling awe is generally considered a positive emotion. It's much less clear that we're talking about actually different emotions. Clearly something like relaxed is also talking about the absence of certain negative emotions, like feeling stressed, for example, or anxious. But it's not clear whether it's talking about a distinct positive emotion. And so it sort of seems like not. So I think that's interesting. Then your theory, you've got a clean way of explaining why there might only be one, although I suppose your approach doesn't guarantee that there's only one, because maybe different types of areas being reduced could produce different sorts of happiness signals. Is that right?

SLIME MOLD TIME MOLD: Absolutely. We definitely start from a place being like we've looked at it, and at least the model that we've proposed, there's only one positive emotion. And in fact, we think it isn't an emotion because it's a whole different class of things. There's one positive signal, and it's possible that's wrong. It's definitely possible that there are other things, like awe might be more of a one-off, like anger, if they are their own class of things. But yeah, our intuition lines up with yours.

SPENCER: One thing we haven't discussed yet is how the governors work together, right? Suppose there are many of them, maybe hundreds of them. How does that lead to actual human behavior?

SLIME MOLD TIME MOLD: Yeah. So you only have one body, and it means that your governors have to sort of share the road. They've got to come to a compromise about what kinds of behaviors are going to be carried out and the order that they're going to be carried out in. We think that there's some kind of parliamentary procedure, which we don't fully understand yet, where governors submit bids or votes for certain actions or plans of action, and the strength of these bids is some kind of function of the size of the error signal. If you're really, really tired, then the bid to go to sleep or at least lie down is going to get more votes. You're more likely to do that if you're in a lot of pain or if you're in a lot of danger, you're going to get a lot of votes for the things that will help you rectify the behaviors that you think will help you rectify that situation. It's probably not just a function of the error signal, because there are obviously some error signals, like suffocation, that demand really high priority. It's possible there are different functions for each governor based on things like the size of their error signal, but there may be other elements as well.

SLIME MOLD TIME MOLD: We also expect that different people will have different set points. One of the most fun things to think about from this paradigm, in my opinion, is that people can be in the same situation materially, but based on their set points, they might behave quite differently.

SLIME MOLD TIME MOLD: That's right. They might have different set points, and they'll have different strengths. They'll get more votes for the same amount of error signal. Or possibly they'll have a faster or more powerful connection between a governor and the voting mechanism, leading that governor to have more or less influence over the votes. The exact parameters are something we still don't really know about, but we would expect that this is a major source of individual differences.

SPENCER: Right. This could explain things like personality, how different people have different personalities.

SLIME MOLD TIME MOLD: Yeah. My whole personality is wanting salt.

SLIME MOLD TIME MOLD: Yeah. For real, it's clear that we have something different about our salt governors. We both really, really like salty foods, and we'll eat pure salt off a plate. Most people don't do that.

SLIME MOLD TIME MOLD: You catch me pouring salt on the back of my hand and licking it off.

SLIME MOLD TIME MOLD: So, which parameter is this? Do we have an unusually high set point, or is there some other parameter that's set unusually? Not clear yet, but we think that's evidence of an unusual setting. And again, once you sort of sit down and try to build a model of it, you would have to actually define at least a proposal for those parameters, and then you could go out and try to test them.

SPENCER: So would you suggest that if you were trying to construct a model personality using your paradigm, you'd think about each of the parameters of these governors? For each person, you could think about, well, what's their set point for Governor 1, what's the person's set point, what's the sensitivity of the set point, and so on. If you go through all the different governors, and then you map all the different parameters, the sum total of all those things would be a person's personality, essentially.

SLIME MOLD TIME MOLD: Exactly. So every governor has K parameters, at least one of those is going to be the set point. We don't really know how many there are. And then there are also going to be a set of parameters about how the governors interact with the voting mechanism. And maybe there are also parameters of the voting mechanism. Together, these form at least a large part of personality. This is very different from current theories of personality because it's really highly dimensional. Not just that there are five different personality categories. There are many, many different precise ways in which your personality can differ.

SPENCER: Do you have a theory of why, for example, when people are asked to describe themselves, and then you take all the words they use, and then you do a factor analysis, you get kind of these five signals a lot of the time, or most of the time, of conscientiousness, agreeableness, extroversion, openness, neuroticism?

SLIME MOLD TIME MOLD: Yeah, that's a great question. I guess we can say two things about that. First of all, it's not totally clear to us what the null hypothesis is. Is a five-factor solution sort of just the default for doing a large factor analysis on this kind of data set? It could just be that if you had some sort of randomly generated corpus, you would tend to get five factors out of it anyway, and so this says nothing about the human mind, and maybe something more about very large data sets.

SPENCER: My experiences know that you could get three, you could get one, you could get seven. But, okay, interesting point. What were you going to make another point about?

SLIME MOLD TIME MOLD: Yeah, I think the other thing is that we would interpret it more linguistically and socially. It seems like there are five dimensions that we tend to care to talk about in language. Because we have all of these words in modern society about things like conscientiousness, because it's one of the things that we care to talk about. It's important to us. We don't have a ton of adjectives for things like, "Do you have a very strong salt or sugar drive?" Although we can talk about it, there are phrases like sweet tooth, but we don't have a whole panoply of adjectives for it, because even though people differ in this way, it's not culturally relevant. In contrast, if you went back to medieval Iceland, they had a ton of different words for being brave or being cowardly, because it was really important to their culture. It's something that they wanted to talk about a bunch. So you'd probably find a different set of factors if you analyzed Icelandic adjectives corresponding to however many dimensions they care to talk about the most. So it's not necessarily a good read on personality per se, but rather the types of differences in people that are relevant to talk about.

SPENCER: Yeah, there's some debate whether you get different models of personality if you run analyses on different cultures. So a lot of what they call weird cultures, like America, you tend to get the big five. I think most people agree. Or sometimes people will say you get six, actually the Big Six. But then, you know, maybe if you go to Papua New Guinea, you don't get the same thing?

SLIME MOLD TIME MOLD: Yeah. Although, I guess I would sort of sidestep the question and say the real problem with the Big Five and every other theory of personality is that it's abstract. It is, at some level, really, genuinely not a scientific theory. Abstractions are fine; they can be useful, but it's not making any kind of appeal to the mechanisms that are happening underneath. So it's, at best, a superficial theory of personality.

SPENCER: Yeah, it's kind of an empirical observation, more than a theory. This is what we do and we find this result.

SLIME MOLD TIME MOLD: Exactly, it's fine if you're a naturalist to say there are cows and there are sheep, and they seem like different kinds of things, but it's very, very different from doing biology. And it's really, really different from trying to, even before we knew what DNA was, say, "Okay, why are sheep different from cows? What is the fundamental thing going on that makes one of them grow this way and the other one grow that way?"

SPENCER: I would say that IQ is a lot like this too. It's an empirical observation, essentially, that if you give people a wide range of intelligence tasks, there tends to be a correlation between them, and you can do this operation to compute this number, and that number predicts different things. There's way more consensus on that observation or series of observations than on, "What is that thing, IQ?" Which is actually still debatable and often debated.

SLIME MOLD TIME MOLD: Exactly, and this is why we challenged you at the start by saying, "Is there anything from psychology that's mechanical?" There are many things in psychology that are useful. There are many things that are robust and reliable. There have definitely been some problems, but there are things in psychology that are really robust, and yet, I don't think there's any part of psychology, or very few parts, that are mechanical. They're all basically, like you said, observations. They're all abstract, they're all superficial.

SPENCER: I guess I'm not exactly clear on what it would take to make it mechanical. Let's take the theory of ego depletion, which I don't really believe, just to be clear, but let's suppose it's a theory. It's out there, and I think it says, "There's some quantity of the mind that gets depleted as you exert willpower. It's sort of like a battery. As you exert willpower, it goes down. If you later encounter things that require willpower, you'll be less successful at doing those. If you're given a plate of cookies and radishes, you'll tend to choose the cookies more because you'll have less willpower left." They even made more specific claims, like giving people glucose can increase the battery more quickly, so that later they can use willpower more effectively. So what about that is not mechanical?

SLIME MOLD TIME MOLD: Yeah, great. It's definitely a step in the right direction. That's not literally 100% superficial. I think what's missing is entities and roles, and this is why we tend to use that phrasing slightly more than mechanical. It's okay to say that there's something like a willpower gauge under the hood, but I think it's better than the general state of the art in that you're at least postulating about something. The main thing that's missing here is that it's not trying to provide a set of building blocks for the mind in general. There's one proposal for one value, a willpower value. You can say things like when you eat more sugar, it goes up. When you eat less sugar, it goes down, or it's related to calorie expenditure. But it's not a proposal for the class of things that exist inside the mind. It's a specific proposal rather than a proposal for a class of entities. Like I said, there's a sense in which it is like much better than not proposing anything at all like, it's better than the Big Five, because the Big Five make no reference to sort of the inner workings at all, putting your neck out and being like, "We think that there's a scalar value for willpower, and it sort of has these rules attached to it, is good." When chemistry started, "Let's try to assume that everything is particles with mass and velocity, and it bootstrapped from there, even though that ended up being not entirely correct." You want to have as comprehensive a proposal as possible for classes of entities that will appear over and over again and the rules that govern them.

SPENCER: I go back to thinking about programming Sims and imagining you're programming them. You could quite easily imagine trying to program ego depletion into your Sim. It's just a single number, and as they expend glucose, it goes down, and then they don't resist the cookies or whatever. With the Big Five, you could also try imagining these five numbers you're going to put in your Sim, and how often it organizes its room will depend on the conscientious number. It feels a little bit different to me because the five numbers of the Big Five personality model seem much more abstract to me. For example, an issue with them is that they seem to mix beliefs with behaviors and patterns, and a lot of different kinds of entities get combined.

SLIME MOLD TIME MOLD: Yeah, absolutely. Maybe, as we said, we're still figuring out the best ways to talk about this. A good way to think about it is whether it's generative or exception-based. Both the Big Five and ego depletion appeal to entities, to some kind of thing under the hood, but it's definitely more of an exception-based system. You add this entity on top to try to explain this particular situation. We think of a real entities and rules paradigm as the entities that we think are going to populate the world, and then we keep extending it out to explain as many phenomena as we can in terms of these tools. Does that seem like it resonates?

SPENCER: I think it's trying to do a different thing. Because ego depletion is just trying to explain one little aspect of the world by creating a new entity just for that thing in the world, rather than provide a paradigm for what's going on generally.

SLIME MOLD TIME MOLD: Yeah, or maybe something to pick on, even more is grit. Some people can go through this hard situation. Other people can't. Why? Because their grit number is different. On some level, this is not superficial, because you're at least pushing it down to a lower level. But I think on another level, it's absolutely not very specific. You can say grit goes down when you do hard things and goes up when you do easy things or something, but you're really just adding a noun in between. It doesn't really have any specificity about its properties or the rules that govern its interaction, and it's maybe even still superficial. I don't know. It's still quite difficult to think about. We're working on it.

SLIME MOLD TIME MOLD: Yeah, I mean, it's interesting to me that you brought up glucose as part of this ego depletion study, and yet I feel like you have a hard time seeing hunger as an emotion.

SPENCER: I would draw the distinction between emotion as the way people typically use the word, right? And I don't think the way people typically use the word considers hunger an emotion.

SLIME MOLD TIME MOLD: Sure.

SPENCER: But I do think hunger is clearly highly psychological. I think that's gotta be the case. I don't know what else it could be other than psychological, although it feels very physical, and so I think it's easy to lose sight of the psychological aspect, or something like fear, which is really clear it's psychological. With hunger, you might feel like it's more rooted in the body rather than the mind.

SLIME MOLD TIME MOLD: Yeah, definitely. I think it's interesting to bring to the conscious mind that type of bias that a lot of people have when they approach this idea.

SPENCER: Going back to grit for a second, I think we can see a challenge that the field of psychology has with the way it defines things, because there's a debate: is grit the same as conscientiousness? You have conscientiousness from the Big Five model, and then you have this newer construct, grit, which is trying to explain people sticking with things, like, "Do they succeed in difficult training in the military, or do they give up?" Grit is proposed as this kind of more specific thing, but we actually did a survey of academic psychologists and asked them about this very question: what is grit in relation to conscientiousness? Because they're known to be correlated. The empirical observations show a strong correlation between them, and we found that about 50% of people who had an opinion said that grit is actually just identical to conscientiousness; it's just a renaming. Then 50% said it is not identical. So, yeah, getting really precise with what we're talking about is actually difficult in psychology and creates lots of problems. You have half the field saying this is not a new entity and half the field saying no, it is a new entity.

SLIME MOLD TIME MOLD: It's hard to know even where to start with that. The responses to this set of questions on this questionnaire are correlated with responses to this set of questions in this other questionnaire. What are we doing? What's going on here? Those questions fundamentally don't seem scientific to me, and it is quite hard to say why that is not scientific and other things are, but it really seems like you're barking up the wrong tree.

SPENCER: Yeah. So let's talk about this with regard to the replication crisis in psychology, where lots of major findings have not replicated when people redo the studies from scratch. How do you see your ideas about paradigms relating to things like the replication crisis?

SLIME MOLD TIME MOLD: Sure, we think that the replication crisis is definitely bad, but not really for the reasons that people normally think. The replication crisis deals with issues of statistics and methods. "Are you doing the right analyses, and are you designing your studies in a way that will yield meaningful results?" We think that the focus on this is, on some level, good. It's definitely bad to p-hack your data. It's definitely bad to design studies in a way that leads to ambiguous results. You want your methods and your stats to be good, but also these are not important things to work on if the questions that you're asking are fundamentally meaningless. If you're purely asking about abstractions, or if you're just shuffling around different nouns, then it doesn't actually matter if your methods and your stats are really good; you're not going anywhere. It's not clearly the case that people were really focused on questions of meaning and then the replication crisis swept in and pushed them towards thinking about methods and stats. But there is this sort of substitution effect where, when people dedicate their time and energy to trying to make psychology better, they tend to think about it. For the last decade, they've tended to think about it in terms of, "Can we do our stats right? Can we not fabricate? Can we design our studies well?" Those are the wrong questions to ask because it doesn't matter how good your stats and methods are if you're asking the wrong kinds of questions. You need to be a mature science, which means we need a paradigm. All of the effort towards making psychology better should be about asking better questions because, while you do need good methods and stats, good analyses yield progress, those are only useful if they're asking reasonable questions.

SPENCER: So is the idea that without a paradigm, "wrong questions" are being asked. While yes, to make progress, you would need to do good statistics, it doesn't really matter if you ask the wrong questions because that's not going to get you anywhere. We need a paradigm that sets us up to ask sensible questions that can enable progress.

SLIME MOLD TIME MOLD: Exactly right. Back when they were in the paradigm of the four humors, doctors might argue about the best way to try to figure out how much phlegm somebody had. It doesn't matter how good your methods are; it's the wrong questions. Back before modern chemistry, you might ask, "Did the world start out as all water or all fire?" Again, it doesn't matter how good your methods or your stats are; it's just the wrong question to ask. When you're not a mature science, you should really be focusing on questions of meaningfulness. I mean, not exclusively. It's also bad for other reasons that people are out there fabricating so much data; there's so much outright fraud. But you're never going to rescue psychology into a mature science purely through methods and stats. It's just not the right direction.

SPENCER: It seems a paradigm is really useful. Not all paradigms are equal. How would you think about what makes a paradigm better than another paradigm?

SLIME MOLD TIME MOLD: You got any thoughts on this?

SLIME MOLD TIME MOLD: A good paradigm is one that you can't get out of your head.

SPENCER: You mentioned earlier that a good paradigm brings questions to the forefront, like search generative in that way. Would you add that as well?

SLIME MOLD TIME MOLD: Yeah, definitely. We ended up writing this whole book because we started thinking about things in terms of cybernetics, and we realized this is kind of great. It leads me to ask all of these interesting and new questions and have new and exciting thoughts. That's why we really wanted to write it up and share it publicly and have these kinds of conversations with people, because we think that this is what science is really about: having discussions about the questions and then going and trying to figure out what's actually going on here.

SLIME MOLD TIME MOLD: Yeah, I can add a little bit to that. A paradigm, we think, has been attempted in psychology before. Cognitive psychology is one of them. We think that these are not really paradigms, but proto-paradigms, because they don't have entities and rules. They're not fully mechanical. Behaviorism was a little better in that it was specific in some ways, but ultimately that was more of a proto-paradigm, too. For something to be an actual paradigm, it needs to be specific enough, and it really needs to have a proposal, even if only at a sort of starter level, about the entities and rules that make up that corner of the universe. The other way it's usually presented is that a paradigm leads you to an era of puzzle solving. There can be different opinions about how to put that into words, but I would say a paradigm identifies specific openings that seem like they can be filled, but they're tractable with the tools available. Chemistry did this with the question of the elements. Once they started discovering new gaseous elements, they realized there might be a lot of different elements. They said, "Well, we can at least prospectively discover the elements by trying to divide substances. Anything that we can divide can't be an element, because elements can't be divided. Anything we can't divide might be an element." You start getting these questions of whether any given substance is an element or not. What are the elements? Is light an element, or is it some other kind of thing? When chemistry started moving into its mature phase, there were a bunch of these specifically open questions. Once the periodic law was discovered, there were literally gaps. The periodic table for a while was literally a puzzle, asking, "What element is here? What can we learn about its properties? Can we find some of it?" A paradigm needs to be mechanical enough to be a mature science. It needs to have a proposal for entities and rules, and I think it needs to provoke this puzzle solving.

SPENCER: So what would you want to see for your paradigm in the future? Are there things you want to work on about it? Are there ways you want people to use it?

SLIME MOLD TIME MOLD: Yeah, absolutely, and to piggyback off of the last answer, we think that there's a really pretty obvious starter set of puzzle-solving questions that people can work on and that we hope to work on. The most obvious one is, what's the list of drives? There's clearly more than one. There's a difference between eating and sleeping. We think we can make a pretty strong case that there are a bunch of different hunger drives. We think that compared to other questions in psychology, these are fundamental questions. By that, we mean they're questions that as soon as you pose them, you can't get away from them. Similarly, is gold a compound or an element? That's a fundamental question about chemistry. Once you ask it, you have to know. So asking, how many kinds of hunger are there? Is anger an emotion? Is it governed by a negative feedback loop, or is it something else? Is happiness governed by a negative feedback loop, or is it something else? Is happiness the same kind of thing as anger or different? They seem like fundamental questions, and they seem like questions that we can potentially answer. At least for the hungers, you can sit down, and we are coming up with a set of protocols where you can try to distinguish between different hungers. There is going to be, at some point, a solved puzzle where we think there are however many kinds of different hunger, and here's what we think they are controlling for. So yeah, those fundamental questions. Actually, I'll pause so you can respond.

SLIME MOLD TIME MOLD: Yeah, I think the immediate next steps are going to look a lot like rolling some balls down inclined planes, type research using both human studies, and we are interested in doing some animal studies as well, and also modeling things on computers. I think ultimately the thing that gets me most excited about this paradigm is the idea that we could do things like cure depression and addiction and things like that.

SPENCER: You're saying that if your paradigm led to a much more precise definition of something like depression or addiction, then you could actually much more easily develop treatments for it, potentially.

SLIME MOLD TIME MOLD: Yeah, exactly. For something like depression or insomnia, when you look at people's lived experiences of those conditions, it actually ends up looking like a bunch of different things. Symptoms of depression can be sleeping too much, but also not sleeping enough, or eating too much, but also eating too little, depending on the person. We think that by having a better understanding of the fundamental human drives, we can tease apart these different problems that people experience and help to solve them.

SLIME MOLD TIME MOLD: To sort of zoom out, I would say there are two main directions that seem obvious to me. One is this puzzle-solving phase. The obvious puzzle-solving questions right now are, what is the list of drives, what are the parameters of each of the drives and the motivational system, and how do we measure these things? If we want to know what Spencer's set point for his potassium drive is, we'd love to have a method of measuring that. So what's the list of drives? What methods are needed to study that? What are the different parameters on those drives, and how do we measure them? That's the puzzle-solving side. The other side is modeling, trying to use these ideas to build very specific models, build them out in software, kind of like you were saying Adam suggested about The Sims. We can build a little model of how we think the mind works. If we disagree with Spencer about that, we can build two models, and then we can see where they disagree. Can we develop specific studies? Can we put people in a situation where the two models disagree and see what they do? Eventually, that will lead, and it's hard to know exactly the trajectory, but eventually that will lead to a better understanding of things like depression, anxiety, addiction, and maybe to better treatments.

SPENCER: Before we wrap up, final question for you, why do you go by Slime Mold Time Mold?

SLIME MOLD TIME MOLD: The summer we started the blog, we were looking for a name, and we separately had been doing this bit where somebody would see something, and they would point it out, and then you'd rhyme with it.

SLIME MOLD TIME MOLD: We were doing cockney rhyming stuff, yeah. So like a raspberry, that was a bad one. But a raspberry, it's called raspberry because raspberry tart and tart rhymes with fart. So we were doing a riff about slime molds, and we got slime mold time mold, and we liked it. And here we are today.

SLIME MOLD TIME MOLD: Yeah, one of us pointed out a slime mold on a walk through the woods, and the other one was like, "Oh yeah, Slime Mold Time Mold." And then we were like, "Oh yeah, that's a good blog name."

SPENCER: That's an awesome origin story. Well, thank you all so much for coming on. We'll link to your work in the show notes. Generally a lot of fascinating ideas.

SLIME MOLD TIME MOLD: Thank you, Spencer.

SLIME MOLD TIME MOLD: Thank you so much, Spencer.