June 6, 2021
What is the Internal Family Systems model? What kinds of information do our emotions give us? How many agents live in our heads? And, if there's more than one, how well do those agents cooperate? What is operant conditioning? What is attachment theory? How does parenting differ from animal training? Is decision theory able to unify many different psychological theories?
Divia Eden has always been interested in understanding how minds work, and she currently spends most of her time unschooling her three kids. You can find out more about her at becomingeden.com or follow her on Twitter at @diviacaroline.
JOSH: Hello and welcome to Clearer Thinking with Spencer Greenberg, the podcast about ideas that matter. I'm Josh Castle, the producer of the podcast and I'm so glad you joined us today. In this episode, Spencer speaks with Divia Eden about internal family systems and emotional frameworks, parenting and pet training paradigms, and the potential of decision theory to synthesize psychological models.
SPENCER: Divia, welcome! It's great to have you on.
DIVIA: Thanks for having me. Nice to be here.
SPENCER: One of the topics that I want to discuss with you to start, is the internal family systems. Can you tell us what that is and why you find it interesting?
DIVIA: Sure. This is a therapy method that I came across about a decade ago and was life- changing for me. The basic thing about it is you think of yourself as made up of different parts. It's like a pretty good paradigm. The guy who came up with this, Richard Schwartz, my understanding and I've read about this some, is that he came up with it by listening to the people he was working with. He found that they were in fact describing themselves that way. Like part of me wants to do this thing but part of me thinks that I shouldn't do that thing. He went with it and ended up finding that there's certain questions that he would ask about people and their parts and how they're related to them where they give pretty interesting answers. Like, he started to ask things like, "Okay, well, how does this part of you feel about that?" He started to ask things like, "Oh, well, I'm mad at myself for not doing my homework." And you'd say, "Okay, well, can you ask the part of you that's mad at yourself to step aside, and then see how you feel towards that." It does seem to be a thing that people could often do and people will get these different parts of them to show up then step aside. Another thing he noticed was that sometimes he would ask about the parts that were active. They'd say, "Okay, well, this actually feels different. This feels more like me." And that was the thing that he called "self" in his therapy thing. There's a lot I could say about it. He categorized parts such as their exiled child parts that are managed by managers. Or if they're in too much pain, then a part that he would call a firefighter would come up. A manager might be something that controls my behavior so that people don't criticize me if that's something I can't handle. A firefighter might be like, "Okay, well, somebody did criticize me, and I can't handle it. So now I'm going to go check Twitter a lot and distract myself." I don't know if that's a basic explanation and there are some therapeutic techniques for working through this stuff.
SPENCER: Okay, I got it. He's noticed that people describe themselves in this way as though they have parts. He takes this very seriously, and just said, "Okay, let's just go with this." Then he starts noticing these patterns with the parts that seem to be doing or their motivations seem to be right. Let's go into exile a little bit, so what is an exile?
DIVIA: Yeah. I can give an example. This is one from my life that's easy to talk about. I remember this one. When I was about eight, it's a pretty — on the surface of it — trivial thing that happened. When I was in school, we were all trying to get some water and there were these cups that I accidentally knocked over the pile of cups and someone was like, "Hey, you knocked over those cups." But at least in my memory I hadn't realized that I did it, so I was like, "No, no, I didn't." Later, when I was by myself, I rethought it, and I was like, "No, no, I did do that." I remember feeling guilty about it and I don't really know what to do with it. I was like, "Alright, well, I guess now I'm just going to feel guilty forever." I forgot about it. Until when I was doing this stuff with my therapist years later, and I would say that as an example of an exiled-child part. There's some part of me that was overwhelmed by this, "Okay, I guess I just permanently feel guilty now." That I shut away because it was overwhelming then I had some coping strategies to deal with that.
SPENCER: Got it. Then somehow that part is stuck and they're still in that state?
DIVIA: Yeah. It's just not really capable of making updates that language people use. That part of me doesn't know that I'm an adult, and I'm not still standing over that pile of cups that I knocked over. Then I have a whole different life now and different coping strategies, etc.
SPENCER: How would that exile part manifest because presumably you're not knocking over cups very often?
DIVIA: No, but I did, in fact, have some really pervasive sense of guilt that I always felt and it wasn't just about that. This was one iconic archetypal memory that was wrapped up with it. If I look at just the exile, I think the way that it's operating is this bad feeling is still there. And this unfortunate self concept.
SPENCER: Got it. So that kind of exile, the theory says it pops up occasionally in different scenarios, and then you're going to feel a lot of guilt, because that's what that exile is about?
DIVIA: Yeah, that was right.
SPENCER: Okay, so where does the manager come in here?
DIVIA: There are some other parts of me that try — it's been a while I don't remember the details of this one perfectly — but that basically has some set of behaviors that I'm supposed to do so as not to trigger this guilt. Maybe I should be apologizing to people as much as possible or I need to be really scrupulous about what I'm saying isn't true and stuff like that, so that I won't trigger this guilt.
SPENCER: So then there's this other part, this manager part that's trying to protect that exile part. Is that how you think about it?
DIVIA: Yeah, that's how I think of it.
SPENCER: It almost begins to seem like being possessed by ghosts or something. It's like, the spirit of this manager part jumps in and picks over your behavior to try to protect this exile from getting hurt.
DIVIA: I do think there's something surreal about it from a certain perspective. I think someone had a really good tweet about it that we live in these mashups of our emotional memory, dreamworld, and what's actually happening to us. It's weird to realize that that's what's happening for me when I do.
SPENCER: And then what's the deal with firefighters and where does that fit in?
DIVIA: The main book that I've worked with doesn't distinguish super cleanly. But I would say, the distinction there is the manager is trying to prevent the feeling from being triggered at all. Then the firefighter is a different type of protector that once it is triggered too much, it tries to make it so I'm not overwhelmed with the feelings. Usually, with some distraction. In the worst case, it could be really destructive behavior like drinking way too much, or something like that. Ones that I'm pretty familiar with are going to check social media, or it could be something healthy too but something to relieve that pain after the fact.
SPENCER: Got it. If you are wracked with guilt, the firefighter's like, "Oh, I gotta protect this exile," jumps in, and gets you to check social media or something to distract you from your own experiences and pain.
DIVIA: That's right.
SPENCER: Then the true self is something in the middle, according to the theory, that's able to observe all these other parts. Is that right?
DIVIA: Yeah. I don't know that I put it exactly as middle but I had a pretty good moment with this. This was when I was working with my therapist years ago then I was saying something about a conflict with another person. I was like, "Oh, well, I don't know if I can be there for the other person. I didn't say it in the qualified way. I can't be there for the other person if I'm really upset." My therapist was like, "Hmm... Interesting that you think that. How do you feel towards that belief?" I was like, "Well, it seems true and I don't know how I feel towards it." She walked me through some exercise of stepping back from it. It was weird. I believe it was the first time, in real time, that it had been so dramatic. Where I was like, "Okay, just a few seconds ago, well, that just seems true." Now from this different perspective, I'm like, "Interesting, that's a belief with some function in my life. I have some evidence for it and some evidence against it. It's maybe load bearing in these ways, and less blended (that's another IFS jargon term). You can be blended with these things where you think that that's all you are, or you could be more like take it as object."
SPENCER: When you talk about it that way, I started to see these connections between IFS and cognitive behavioral therapy, or even mindfulness. Where in a given moment, you can feel like a thought is totally true. Then you can step outside of that thought and be like, "What's the evidence for and against?" Or you could just view that thought mindfully. One thing I wonder about what IFS is, is that it's just another way of doing a very similar thing to what cognitive therapy is doing where instead of thinking of it as, "Okay, in a moment, you can believe a thought. But then you can step outside of it and you can question it and challenge it." In IFS, you just assign a part to that thought and you step outside of it. Maybe you assign a visual image, or maybe it makes it more vivid in a way than CBT does, or maybe it's trying to accomplish something very similar. I wonder what your thoughts on that are?
DIVIA: I have less deep experience with CBT than I think you do so you might know better than I would. But something that IFS does talk about that comes up a ton with me and many of the people that I've done this with, is the idea of an analytical part where that is really not the same as self and is more concerned with consistency, checking, and looking for evidence for and against which all can be great, healthy and extremely useful. But when I'm doing the IFS process and this comes up with someone, I'll be like, "Okay, that's great. I'm not saying there's anything wrong with that and would the analytical part be willing to step aside so that we can have a different type of high bandwidth connection to the emotional side of what's going on here?" I do find it very useful because there's a tendency with many people to be in the habit of analyzing very quickly without, necessarily, all of the information that would be useful to actually analyze. Another one of those questions that comes up a lot in IFS with different parts is, "Do you think that I understand just how bad this really is?" Or "Do you think that I understand just how important this really is?" Whether it's the strategy or the pain or whatever, but let's say it's the strategy of being really careful to never say anything that might be misconstrued by somebody, that's a potential strategy. I could start picking apart that strategy but in the IFS model it's not gonna work super well if I hadn't passed first to the checksum of things like, "Do I understand just how important this really is?"
SPENCER: Important it is to me? Or what do you mean by that?
DIVIA: From the perspective of the part.
SPENCER: I see. So adopting that exiled perspective, or something like that?
DIVIA: Yeah, that was right. Or the protector's perspective, either way.
SPENCER: Protector is another word for the manager and firefighter, is that what they do?
DIVIA: Yeah, that's right. Protectors are the managers and firefighters.
SPENCER: Got it. Let's think of this a little more. Suppose you work within this framework. Let's set aside the question of, "Is this true?" Let's just assume for a second this is a useful way of framing things. What do you do with this? How do you work with your parts?
DIVIA: People usually have a trailhead. It's something that they can tell they're really stuck on, where they have internal conflict about it. Or they think they should be doing one thing but they keep doing another thing. Or they're not really sure what they're supposed to be doing. Then there's an iterative process where the idea is that you want to go first deeply into one part and really untangle things from there. The way that you can get this actually full emotional picture of what's going on is if you can get all the other parts to step aside so you can pick one of the latest. So let's say somebody keeps getting angry at their partner for something their partner is doing. They'll be like, "Okay, well, how do you feel towards the angry part?" They might say, "Okay, well, I feel frustrated about it." You can say, "Okay, well, can you ask the frustration to step aside so you can get to know the anger from a more open place?" Maybe it will. Maybe it won't. It's fine if it doesn't but that's a different dependency to deal with. Then eventually, people will be like, "Oh, okay. Well, now if I check in how I feel towards that, I feel curious, I feel open." Then at that point, we can try to get the whole story from this angry part and then go from there.
SPENCER: So you're flushing out the full emotional experience of the angry part, and trying to learn everything there is to learn from that part.
DIVIA: Yeah. I think there's a super tight analogy. I imagine when you've talked to people often they're like, "Oh, I have this problem." If you're like, "Well, okay, how about we do this." People often aren't particularly interested. I think for a variety of reasons, one of which is that they haven't really managed to fully explain what their problem is, what they see the constraints as, and what they think the stakes are. Sometimes people are receptive without all of those. But it's unreliable that people will be receptive if they don't get the empathy first and it's the same with the parts.
SPENCER: So you're saying if you try to jump into problem solving it doesn't work as well. Whereas if you really say, "Okay, what's really going on with this part? Let me listen to it really carefully. What does it really want? What does it care about?" Does that tend to work better?
DIVIA: I think the first thing can be a lot more shallow often because of, whether it's culture or just human nature. But I think a lot of the times before we can really get to the heart of things, things will sound different from how they really are, particularly filtered for reasonableness and stuff like that. Before I learned more about introspection, I had a habit of just dismissing things that came out of my mind if it didn't really make any sense to the point where I'd almost forgot [sic] that I even had ever heard them from myself. I think that's pretty common and people have different filters. It's peeling that back so we can actually just get a bunch of the stuff that's actually driving the behavior to be out in the open. I think it's 90% of what I need to do to solve these things typically.
SPENCER: Okay, let's say you let the part tell its whole story. You listen to it with curiosity. What happens next?
DIVIA: If we're talking to a protector then typically there's an exile that it's protecting. If you establish enough trust, it'll let you see the exile if you ask. Sometimes that happens spontaneously. Sometimes it can play out in different ways, like seeing a pretty memory, or maybe an archetypal memory of me in some past situation with some overwhelming amount of pain that I didn't know what to do with. Then I do the same thing there where I try to get the whole story and the emotional impact and everything from the younger version of me. There's some steps that people do like, "Okay, well, what should have happened there? How can I let go of that? How can I symbolically let go of this? How can I update that? Like this isn't really what my life is like now and there." There are different ways to do it. But it's basically the same thing where you get the whole story from the exile as well.
SPENCER: Got it. You get the whole story of exile and there's some symbolic letting go. I've heard those people like to imagine throwing something into a bonfire but I'm not sure exactly how it works.
DIVIA: I've had times where that worked for me. Sometimes it seemed meaningful and nice to visualize it that way. Sometimes it didn't seem particularly important.
SPENCER: What is that letting go process? From the exile's perspective, what's happening?
DIVIA: There's some pain and usually some belief that people will hold on to. There's common ones that people have. Like, let's say, there was one I remember. I encountered something about "there's something wrong with me" which I wouldn't say I have zero of that now. But I was working with one exile that felt a lot of shame and had a "there's something wrong with me". I think it's not a very useful perspective or belief. At least the way I was holding it. A lot of them tend to be this not very nuanced but very emotionally powerful, often in ways that I am not serving my life well. As I would let go of the thing I would find, in the ideal case, then the emotion would lift and be like, "Okay, like, I felt it." There was internal communication that this emotion was meant for and it was stuck and incomplete. But now it happened so now I don't need to keep feeling that. Ideally, the belief would shift to something more functional. Like the one with the cups — this is one I remember because it's easy to talk about — it's a belief like, "I have to permanently feel guilty." It probably wasn't exactly those words. When I thought about it, like, "Okay, what was the actual positive value driving this?" What came to mind was that it's important for me to know what's true, which is something I stand by and I'm fine having that as one of the things that comes up a lot in my psyche. It seems a lot more productive.
SPENCER: But how does that relate though to that other belief about feeling guilty?
DIVIA: Well, because they're both about the actual situation that I claimed that I hadn't knocked over this pile of cups when really I had. So that's what I feel bad about. The situation was over and it didn't seem like it made any sense to me to gather everyone and be like, "Okay, I have realized that I have lied about this." I still wouldn't really recommend that child me to do that. But I didn't know how to hold on to this idea that I cared that I had been wrong about that and I had actually knocked over the pile of cups. The only way I knew how to hold that previously was just by permanently feeling guilty about it. But now, I have a more mature ego structure and I think I've better ways to handle that.
SPENCER: So there's a revising [sic] of the belief. If you're dealing with a manager, you might revise the belief of, "Oh, whenever this exile is in pain, I need to jump in and save it." Like maybe change it to a more helpful belief. Is that the idea?
DIVIA: You could try to do that but the more canonical IFS process has to do with making it to the exile, just isn't in as much pain.
SPENCER: Oh so the manager doesn't need to jump in, and it's not the manager changing its mind?
DIVIA: Yeah, that's right. Usually the manager, if you go back to it and do some process there, it'll be like, "Oh, well, now that this exile isn't in so much pain, I can do this other thing that's more like a healthy part that's not about better not to feel this pain but is more about getting me things I want in life."
SPENCER: I see. I think I'm still not totally clear on what makes exile feel better in this model. Is it just having expressed those feelings? Or is it because exile is a bliss shift once it explicates things?
DIVIA: I think it's both. I guess the way I understand emotions, they have some purpose, like to communicate some information either internally or externally. Then once it's complete, I'm not supposed to just keep feeling the emotion indefinitely about something that happened that's already over. I don't know, I'm curious how you relate to your emotions in that way.
SPENCER: I totally think they contain information. Some people think of emotions as something you should ignore. I definitely don't think that's a good strategy. Other people take emotions as something you should always believe or trust and I also don't take that view. The way I think of emotions are like detectors for different things. They give you information. Just like any detector, they can go haywire. You want to pay attention and learn everything there is to learn but it doesn't necessarily mean that you should act in accordance with what the emotion is telling you right now. I think that is really a good life strategy. So if you're feeling sad, I ask myself, "Okay, do I feel like I've lost something here?" Because usually sadness is associated with losing something. I found a really useful question then I can start reflecting, "Hmm. What is it that I feel like I've lost?" Then I start to unpeel that. I have very very low anger — I deal great with anger — but if I do have a flash of anger. I think anger is usually caused by feeling like someone's blocking you or trying to prevent you from getting a single value or trying to remove value from you. So I think, "What do I believe someone's trying to stop me from doing? Or what kind of value do I believe someone was trying to stop me from getting?" Usually that really points to something.
DIVIA: I'm curious if and when it does point to something and you can see what it was about, does the emotion typically resolve itself?
SPENCER: Not necessarily. Because from my point of view, let's say you're feeling sad and you're not sure why. Then you ask, "Okay, what do I feel like I've lost?" Then you realize, "Oh, the thing I feel like I've lost my friend. We used to have these amazing conversations. But now, the last three times we see each other we didn't have good conversations and I really valued that part of our relationship. It seems to be gone for some reason and I don't understand." That doesn't necessarily make you feel less sad because now you just know what you feel like you've lost but you might still feel like you've lost it. Right?
DIVIA: Yeah. I think for me it does typically make me feel less sad and there's some nuance to it.
SPENCER: I feel like emotions have this funny wave property. Let's say, something really bad is happening in your life. It's not like you'll just feel consistently bad at all times because of that thing happening. Let's say someone in your family is really sick. It makes you feel constantly stressed out all day long everyday. It comes and goes in these waves. I find it's more like that. Reflecting on it won't necessarily make you feel less sad right away. In fact, it might even intensify very briefly as I'm honing on the thing that I feel like I've lost but then it will probably come down after a little while.
DIVIA: I definitely would agree with you about the initial intensifying [sic]. That's often a sign for me that I've hit on what it really is. If I'm like, "Oh, maybe it's about that and then I feel it much more strongly."
SPENCER: Yeah, exactly. You're paying attention to those little surges and then you're like, "Oh, okay. I hit on the thing that's making me sad because the moment I have that thought, that's when I really feel it."
DIVIA: Yeah. With something that is a really big deal with grief, I do think there's a process of loss that doesn't acknowledge the loss and doesn't update about that. That definitely doesn't happen immediately. In some ways, if it's a big thing, it will never be over. But for me, with little sorts of emotions, I think it's a very common experience that if I identify what it's about, then it basically goes away. Maybe not right away but within a few seconds. It's sometimes pretty dramatic but seems like that doesn't happen as often with you.
SPENCER: I think I do agree. It's just that if the thing is still very much lost in my viewpoint then I'm probably still gonna feel like the emotion is not gonna go away right away. Because I still have this perception of being lost. But I think what can happen with a lot of everyday little emotions is that when you focus on it, you're like, "You know what, okay, I'm sure that's fine." At least that's my experience. So maybe you're thinking about your friendship, and you're like, "At that moment, I felt like this thing was lost." But then you're like, "You know what, though? We've just been distracted by this other thing that's why these conversations haven't been very good." Then you're over it. I guess I would say, it's like, how much of the thing driving that emotion is still there.
DIVIA: So if what you've done is you've uncovered that there's some distortion and there's some distorted belief that was creating the emotion, then when you look at it, it's like, "Okay, well, that's not even true." Then the emotion will go away.
SPENCER: Yeah, is that true? Or it's like, "Oh, you know what, that's just actually not that important." If it's not a big deal, I know that I can deal with that. Whereas if you're like, "Oh, now that I'm starting to think, actually, that's really important and it's really valuable to me. It still feels like it's lost but my emotion is not gonna go away right away."
DIVIA: It's funny. I think we're probably describing very similar things. Can I give a different example though, to maybe make things clear? I definitely discovered about this IFS framework, and I'm like, "Alright, so what can I use this to relate to?" IFS is pretty broad in terms of which things that it's saying is helpful to relate to his parts. Being confused, for example, could be a part, or feeling tired is one that I worked with a lot.
SPENCER: Interesting. That's fascinating to me that you can work with feeling tired.
DIVIA: This was actually a big thing for me for a while. I think it doesn't happen to me anymore. But I used to have a pretty strong pattern which it took me... Obviously, we do need sleep and sometimes, if the deal with me being tired is just that I want to go to sleep, the tired part is telling me to go to sleep and it all makes sense.
SPENCER: What I think tiredness and sleepiness is different [sic] so I'm not sure what you're talking about.
DIVIA: That's fair. I think that I tend to use tired also to refer to sleep. Like I just got a lot of exercise and now I want to sit down but I don't want to go to sleep.
SPENCER: I like to break it into four pieces. There's mental tiredness which is like imagining you just spent two hours taking a SAT or math test and you're just like, "Oh, my brain." Then there's physically tired, like you've just worked out with really heavy weights and your whole body just feels really tired. Then there's sleepiness which is a real tendency to want to go to sleep or the ability to fall asleep quickly. Then the final one is, I would call it slowness, which is the feeling you get right after you wake up, or if you just drank a bunch of alcohol or things like that. They all are related feelings but I distinguish them.
DIVIA: Okay, that makes sense. Also, I used to feel sleepy when I felt angry.
SPENCER: Oh really? Interesting.
DIVIA: I think I basically resolved this one. But it was a pretty clear pattern that I had missed until I started poking around. Liike, "Why do I feel sleepy in these different situations?" Because I think ultimately, what are these things for? So if I'm sleepy then maybe I'm supposed to go to sleep, then that just makes sense. I'll do it and that's fine. But sometimes I feel persistence less so. I used to feel more frequently persistent sleepiness [sic] in situations where I wasn't going to go to sleep. I think that there's something a little bit dysfunctional about that. Not that it's the worst thing in the world. But when I'm like, "Okay, well, why is this happening?" Sometimes it could be the parts like, "Well, okay, you can't go to sleep now. But you've not been prioritizing sleep enough in general. So I'm going to keep reminding you that you should get more sleep until you deal with it." And in that situation, I would find that if I make an authentic deal with myself that I really mean, where I'm going to prioritize sleep more, the sleepiness would usually go away immediately.
SPENCER: Like it was a bid to get you to care more about sleep and then once you made the deal, it's like, "Okay, I don't need to keep reminding you anymore."
DIVIA: It can be. Then I also had this thing where sometimes it would happen because I was uncomfortable with feeling angry. So I'd feel angry, and then I would feel sleepy.
SPENCER: What do you think it was doing there?
DIVIA: I think it was probably adaptive when I first started doing it. I really actually didn't have good strategies for dealing with anger. I would say this is like a firefighter thing. Like, "Okay, this is some amount of anger that it's maybe going to be socially destructive, it's maybe going to be something other people will judge you for, maybe you just don't know what to do with it. So how about you just feel sleepy and avoid all those problems?"
SPENCER: It's like someone taking a sleeping pill in order to prevent them from yelling at their spouse. Or it's just like an internal knock-you-out thing.
DIVIA: Yeah. It has some costs. There's a reason when I worked with it it was not stable while trying to figure out what was going on. But I don't think it was the worst strategy in the world either.
SPENCER: We ran a randomized controlled trial where we found an interesting link between tiredness and anxiety. Firstly, we actually did a survey and a research where we asked people how tired they feel right now. Then we also asked them how anxious they feel right now. We found a bizarrely high correlation between the two which was surprising. So when we did a randomized control trial where we actually had some people in the control group and we asked them how tired they felt. Then wait a little while asking how tired they felt again. The other people there in the intervention group got a calming exercise but it was designed to be one that would not get you energized. If anything would probably make you more sleepy because it's relaxing. But in fact, the people felt less tired at the end which was just testing our hypothesis that for some reason, sometimes when people are anxious, it seems to actually make them tired. Then removing the anxiety actually can lift the tiredness. So it led me to think that actually people can often be confused and be like, "I'm tired," And not attribute it and not realize it's actually caused by being anxious.
DIVIA: Yeah, that sounds pretty similar.
SPENCER: I want to step back and think about IFS as a system, internal family systems as a system. Well, the first thing I want to ask about that actually is, how do you view it? Do you view it as a simplified model of psychology that gets some things right and some things wrong? What's just the way you look at it?
DIVIA: I guess I would agree with that. It does seem like a simplified model of psychology. I don't think it accounts for everything in psychology but it seems to me like a pretty good one. It seems like a lot of stuff I read about psychology had elaborate theories about why I might be doing things that I didn't want to be doing or what I might be worried about that was driving my behavior. Another example would be, I, like many people, had a hard time focusing on homework and schoolwork and stuff like that. I would typically get it done. I would do it at the last minute. But it wasn't like, "There was something." I'd been exposed to a lot of guesses from different psychological sources about why this might be going on for me. But ultimately I found that part of what IFS is, is a set of tools for introspection that once I managed to introspect I think it was just pretty different from the guesses that I'd read elsewhere. So part of what the appeal is to me is that there's less presupposed about what's going on with people. It's more, "Here's a way to relate to yourself where maybe you can introspect correctly." People could also get false positives. If I'm working with someone or myself, I always say, "I don't really have any opinion about whether this worked unless you go away and then it's different." In which case it did. I don't know if I answered your question very well but that's some of how I relate to it.
SPENCER: The false positive thing worries me because, just to be totally frank about it, I've experimented with IFS and I don't think of it as True (the capital T). But I'm like, "Oh, it's a psychological system. It gives you a very specific set of steps that you can do and you can try. If they're useful to you, great. If they're not useful, fine." Personally, I have not found it particularly useful. I also know I haven't given it a huge amount of time but I've put a few hours into trying it. I have not found if I got benefits but it doesn't mean very much. I do worry about false positives from it. I do worry that I am creating stories about myself. Then suddenly I feel like I don't know how to verify them. How do you deal with that? Do you just say, "Well, what really matters is I feel differently after, if I feel better, and if it improves me in the way I want to be improved."?
DIVIA: I do have heuristics that I would use if I'm working with myself or someone else to best guess whether we're on the right track. But ultimately, I hold them pretty lightly because I think it's possible to start optimizing for my heuristics of whether it's working. I don't want to do that. So the real proof is in the pudding like, "Does it improve the situation?" I've just seen enough times where people got really dramatic improvements really quickly. There just seems to be something there, and it's certainly been for me. I've had that experience many, many times where I've had a relatively short hour of introspection about something and a problem that I've had for years. It's just gone.
SPENCER: Well, and then it's just permanently gone at that point?
DIVIA: Not always. I think some problems are just complicated and some things are inherently hard to resolve.
SPENCER: But you've had times that were really clear and then it just was gone and didn't come back?
DIVIA: Yeah, absolutely.
SPENCER: On the false positive thing, I think a lot of people underestimate how good we humans are with coming up with things and how there can be a subtle distinction of like, "Did I just make that up or is it just a story that I'm telling myself versus that's a real thing?" I think I'm just very acutely aware that my brain can just tell a story and then boom, "Is that true?" For example, one thing I noticed, I don't know why my brain does this, but sometimes it wants to make things good that are not good. Just as a silly example, I'll drop a glass, it will shatter everywhere, like all over the floor, and my brain will be like, "That's great. Because you don't really like that glass and now you don't have to drink from it." And I'm like, "Brain, what are you doing? That's not good. Why are you telling me a story about why this is good? It's just trying to make me feel good about it." I could have just not used that glass. I don't want it to shatter over the floor. I just notice my brain do things like that a lot [sic].
DIVIA: Of course, though me [sic] as somebody who likes IFS, I could be like, "Well, it sounds like there's a part of you that's interested in making things good. You could dig into it if you want, and maybe if you did, then it wouldn't do it as much."
SPENCER: Yeah, that's a perspective. When you think about these different parts, do you give them a visual image in your mind? Do you imagine them as entities? Or do you just think of them as copies of yourself in the past?
DIVIA: Well the exiles, I would typically think of it as copies of myself in the past. I would think my metaphors for the protectors are more all over the place. I would say that this is a very important part of how I relate to myself and in some ways to other people. But I don't tend to do it formally as much anymore. These are all outdated answers because mostly my impression was that I picked a lot of the low hanging fruit for me. I'm not sure if that's right or I just got distracted with it. But when I was doing it a bunch, I would say the protectors were (who knows) sometimes I did have some visuals, sometimes I didn't and the exiles were invariably a version of me.
SPENCER: Sometimes people give them names too, right? They try to name them something that doesn't degrade them but just tries to fit them in some natural way.
DIVIA: Yeah, and some people are also like, "I have these enumerated persistent parts with these names that I like to use." That can be interesting. Sometimes I notice ones that pop up more often for me but I never put a lot of effort into trying to remember them over time. Because for me, it was like, "Okay, I did some stuff. I messed around in my head. Is it better or not?" And then try to go back to it next time with fresh eyes.
SPENCER: How seriously do you take the parts? What do you think of them as being?
DIVIA: Better like different firing patterns of neurons in my brain? I don't know. Maybe, maybe that's not what you're asking?
SPENCER: Well, I guess to what extent do you actually think you have sub agents? Like, those classic studies where they would take someone who had the corpus callosum that connects the two halves of the brain separated. Sometimes people when they have really bad seizures, they get the corpus callosum cut. Now, the two halves of the brain have much less ability to communicate with each other. Then there's interesting ways, in certain experimental setups, you can get that person to almost act like two people like the two halves of the brain connect independently. So it suggests that, "Hmm... maybe we really have these different agents inside of us." But normally, they're like cooperating so much that maybe it's hidden from us. So I'm just wondering, how agenti do you think of these different parts of being?
DIVIA: I guess it really depends. And also when you say, "Normally we cooperate with ourselves so much," I sort of want to push back. I don't know. I feel like I do okay. But when I talk to a lot of people out there in the world, it seems like it's very salient how people do not necessarily believe they're cooperating with themselves. So I will flag that.
SPENCER: Fair enough. Fair point. I guess I mean, the integration of information is usually really high between the left and right hemisphere...
DIVIA: [interrupts] I don't know.
SPENCER: You don't think so?
DIVIA: I really don't know if I would agree with that statement overall. But insofar as what I think parts are, there's a cluster of beliefs, emotions, and (sort of) affordances for doing things that all come online together that probably have patterns of vocal tension, vocal patterns, or muscle tension. Maybe this has happened to you. Maybe it's not. But when I sometimes talk to people, I don't really focus on this stuff super much because I think it's more normal to just interface with people as though they're like some coherent entity.
SPENCER: [laughs] You don't just talk to people's parts.
DIVIA: But sometimes I will see people really switching. They'll be talking about one thing and they'll say something. Then their whole demeanor and everything will shift to a pretty different perspective. My conceptualization of that is like, "Okay, well, it seems like a different part of them is now active."
SPENCER: I want to contrast this IFS model with a different model. I'm curious, whether you think they are just two ways of describing the same thing or they are really two different models of the way the mind works. The other model I want to contrast it with is: What if we just think of a human as one entity or one agent? But that agent has some parameters, like their emotional state, their energy level at that moment, and what is loaded into the working memory, and things like that. So the reason that we can suddenly act differently is because we can be in an elevated anxiety state and have certain memories that are really salient at that moment. Then suddenly we're going to act a certain way versus being in a different emotional state with different memories activated and a different energy level and so on. How would you like to contrast that?
DIVIA: I guess I'm not sure what distinction you're pointing at. Is the distinction something like, to what extent are the different parts of us optimizing for different goals? To me, those sound like two descriptions that are basically saying the same thing but I'm sure you're getting at some distinction.
SPENCER: Well, it's like, should we think of ourselves as different agents that can come online at different times? Or should we think of ourselves as one agent that has this vector of information, and this vector of parameters? Like, how anxious are we at a given moment? How sad are we at a given moment? How tired are we at a given moment? Etc. So one is like a single agent with this continuous set of information that's causing different behaviors and different situations. The other is just like different agents. I don't know. It seems to me that these actually are different models. Like if you're going to program a human mind, I feel like the code that you'd write would be different for these two models.
DIVIA: Yeah. Are you up for saying more about how it would be different?
SPENCER: Yeah. In one case, you'd have a categorical variable that's like, which agent are they in right now or which agent is activated right now. Maybe there are 27 agents and you're one of the 27 agents activated at a given time. The other model wouldn't have a variable for what agent they are. It would have a vector of information about, let's say, how anxious they are? It would be one number or it could be anything from zero to 100%, or whatever. Then on Saturday that's another number from zero to 100% and so on. The behavior patterns they would have, the vocal tension, the level of avoidance, and the things that they might think about would be a function of this vector of information, so it changes based on the kind of the state they're in.
DIVIA: Ultimately, if I had to pick between the two models, I pick the second one because I don't think most people have persistent 27 different agents. However, I think that they're probably something a little more parsimonious than making all of those. I would not expect those parameters to move independently at all. In particular, I would like to...
SPENCER: [Interrupts] But there's high correlations between that right?
DIVIA: I would expect there to be some cluster of memories that are particularly salient with the different emotional responses with the different words being used that tend to be really tightly correlated whether those things are all active or none of them are active.
SPENCER: It was like an attractor state in the vector space where if you get near it, all these different variables are gonna cling around a certain distribution. Is that the idea?
DIVIA: Yeah. Though I wouldn't expect this to be stable over time. That's the point of the IFS process. You can integrate this stuff. Then it makes sense to not think of it as this— the point of the exiles is to not have to be exiles anymore. Then I would say the healthy parts tend to look more like you're just doing different strategies at different times for different things. But not this sticky, like, all these things are moving together.
SPENCER: Do you think the IFS could be falsified?
DIVIA: Yeah. For example, if people did the process and it never yielded useful results, I would consider that to be falsified. Or maybe you mean something different?
SPENCER: Well, that's viewing it as just a useful tool. Like a technique you can use. But I guess I mean, more as a theory.
DIVIA: I don't know. I think I do relate to it as a useful tool more than as a theory.
SPENCER: I really liked that way of looking at it. Like, "Hey, there's many different things you can do with your mind. Some of these are useful and some are not. Let's figure out the useful ones." It's very cool to try to understand why they work or the theory behind them. But you could also just use them as a tool. You don't have to understand how a plunger works in order to use a plunger. If a plunger is useful, use it.
DIVIA: I do want to understand how it works. For me, falsifying it seems more like, when you ask people whether their parts can step aside, does anything interesting ever happen? And I claim, "Yes." But maybe I'm fooling myself. Maybe someone could run a bunch of studies where I'd be like, "Alright, interesting." I'm still confident that I've observed what I've observed, but maybe it was something that had nothing to do with those questions that I was asking, and I should relate to this whole paradigm differently.
SPENCER: I touched on this a little bit before but to me it seems like a lot of the IFS stuff is stuff that I do but in other ways like with other metaphors or framings of it. Like when I'm in a serious emotion I try to pay attention to it and say, "Why do I feel this way?" And I try to gently coax out my feelings on that thing and my motivations. Then I try to note that the motivations are conflicting and that I have certain behaviors that are trying to change my emotions because maybe I don't want to feel that way right now. It seems to me, this is just, in some sense, not that different from what IFS is doing. It's just in a different language. What do you think about that?
DIVIA: Yeah, that sounds right. It sounds like you probably are doing something that has a lot of overlap. Parts work isn't just IFS and I've known multiple people that, on their own, came up with this idea that it could be very useful to relate to themselves as parts and to take that metaphor pretty far. Again, it came out of this guy noticing that this is how people already talked and ran with it. So I definitely think that all of the IFS techniques are on the spectrum of how people tend to relate to themselves anyway.
SPENCER: You mentioned to me before we started recording the podcast that you see a connection between IFS and operant conditioning. Do you want to talk a little bit about what operant conditioning is and then how you see them as connected?
DIVIA: Mostly, the short answer about how they're connected is, I think they're different theories that both make interesting predictions and produce interesting magic tricks about what you can do with the human mind. I have a decent amount of experience with clicker training. I've done this with my two cats and two dogs. I experiment in general. The basic theory behind clicker training is that there's an animal exhibiting a bunch of different behaviors. Let's say, a really basic one, I want my dog to sit. There are different ways I can do it but one of them is I wait around until the dog does sit. Then I use some event markers like I click the clicker when the dog sits and then I feed the dog some food. The idea of this, of course, is that the dog wants the food. It won't work if the dog doesn't want the food. The click is meant as an event marker. At first, the animal doesn't understand necessarily what's going on but eventually they'll realize, "Okay. I always get the food right after I hear the click. So this click is important somehow." Then they'll realize, "Okay. I see, she keeps clicking when I do the same behavior." Then they'll start offering the behavior." That's the basic idea. You click when the animal does something you want and you give them the food then they start doing that more. You can also then put things on cue and stuff like that. There's also the process of shaping by successive approximations. So let's say, with sitting, dogs do it fairly often. So maybe you wouldn't want something like that one. You can maybe just wait around till they did it. But let's say it's something like going to lie on a mat. I've trained my dogs to go lie on a mat. At first I'll click if they just go near the mat, or if they're on the mat, or then if they sit down. Until finally, they're lying down. So I'll click for successive approximations. This seems to work pretty well. It's very popular. Modern dog training is largely this and people in zoos are using this all the time. That's not a full theory of operant conditioning but it is most of what clicker training is. I could say more if you want me to.
SPENCER: That's helpful. To unpack that a little bit, the dog wants the food. The dog, at first, views the click as neutral. The dog doesn't care one way or another about the click. But the dog comes to see that the click is perceived as food. That actually changes the way the dog views the click. Because over time, whereas initially, you had to always give the food as a reward to get the dog to do the behavior you wanted. Over time, the dog starts to view the click as a reward in itself. Is that right?
DIVIA: Yes. The dog does. The idea is that clicking would be very rewarding. However, the trainers, at least that I follow, say you should always give the food afterwards because the click can also lose its value.
SPENCER: But you could use the click as an intermediate reward before the final reward, right?
DIVIA: Well, temporarily, yes. The idea is that the good thing about the click is you can time it very precisely.
SPENCER: Got it. So the moment the dog sits, like the second he puts his foot on the ground, you can click and the dog's like, "Nice. I know I did something right." Then it gets the food. The click lets you move this psychological reward right to the moment of behavior even though you're about to give the food a few seconds later.
DIVIA: Yeah, totally. I don't know if you've read the book "Tiny Habits''.
SPENCER: No, I haven't.
DIVIA: I recommend it. It's basically a book about clicker training yourself. I think that's almost precisely what it is. It talks about how to figure out what is the mental motion that's your internal upvote button that is like you giving yourself a psychological reward and to do it after you do things that you like that you did.
SPENCER: Oh, that's cool, and I like that. Because it can be tricky to reward yourself. But it's the idea that you have some psychological thing you do that's an internal reward.
DIVIA: Yeah, totally. He has a list of them. He also talks about shaping. That you should start your habits small and grow them and that you need the reward piece. He basically says there's— People will say it takes a month to form a habit and his contention is, "No, no, if you haven't noticed something different after the first few times you've done it. If it doesn't seem to be catching on then you should debug it now and not wait until longer. Because there's probably some much better way that you could be doing it."
SPENCER: That's BJ Fogg, right?
DIVIA: Yeah, that's right. For me, it's this internal like "Yes! I did that!" kind of thing. I think it's going to be different across people and it's not the sort of thing people tend to be super articulate about. But insofar as it's something that I actually want to be doing for my own reasons. If I want to give myself material reward, I think that it's always weird with me. Because the thing with me and the dog is that I'm holding the bag of food and he's not. Whereas this is me giving myself some material reward, I could have done this at any time. To me, it tends to feel a little artificial. Whereas this moment of like, "Yes, I did it. That's what I wanted. I've noticed it and I'm internally upvoting that." seems like a very coherent thing. And yes the click is supposed to be that moment for the dog.
SPENCER: Right. Because you're not gonna just be like, "Yes! I'm awesome!" just at a random moment. You actually want to do that only when you've done the good thing, right?
DIVIA: Yeah. In the book, he says that that's how you should figure out what your thing is. Most people have some circumstances where they did something awesome where they naturally have some response like this and to figure out what that is and see if you can use it outside the contexts where you normally would.
SPENCER: That's nice. Let's talk for a second about the relationship between clicker training and operant conditioning. I understand that operant conditioning is a broader theory. And in operant conditioning, you have four quadrants. You have positive rewards, negative rewards, positive punishments, and negative punishments where I think positive here means adding something and negative means removing something. So a positive reward is like giving a dog a treat. A negative reward would be to take away something the dog doesn't like. Like opening a dog's cage if it's locked in the cage. Was that right?
DIVIA: Yeah. Usually, they call that negative reinforcement. People confuse negative reinforcement and punishment understandably because both are almost certainly happening if there's an aversive event. For example, one way you can train a dog to sit is if you put them in a shock collar and you have a low level shock that you remove when the dog sits down. Negative reinforcement is a pretty reliable way to train behaviors. One issue with it is that whenever there's something aversive that I'm adding, there's probably going to be some fallout from that. Maybe the animal will want to avoid me in general. Certainly, whenever I start this shock that then I'm going to remove when the animal sits. It's like I punished when I started the shock.
SPENCER: Right. If we think about the reward side, those two quadrants versus the punishment side. The reward side has a lot of advantages. One is that people and animals like being rewarded. So if you're rewarding your dog a lot, your dog actually wants to be involved in training, presumably. Whereas on the punishment side, people don't like being punished and the animals don't like being punished. The animal or person is going to want to avoid those kinds of interactions with you. Also, if you care about your pet, you probably don't want your pet to feel bad or be in a negative state. So just right there that gives an advantage to reward.
DIVIA: Yeah. I have a lot to say about this. I think the thing with punishment is actually — and my understanding of the literature in this certainly squares with what I've seen in my anecdotes is that — it doesn't work very well. If it is going to work, it usually works in one of the first few times. But beyond that, it often doesn't work.
SPENCER: One thing I've noticed about punishment is that, let's say a dog gets punished if it pees on the floor. It seems like in my experience what the dog will learn is, "Oh, pee on the floor when nobody is around." Then if you punish it later when you eventually find the pee on the floor, it will not work because it might be two hours later. The dog has no idea what it's being punished for. The dog learns to avoid the punishment and then you have no way to tie a future punishment with that earlier action.
DIVIA: One of the predictions that operant conditioning makes is the theory that timing matters a lot. I think that's true. I think this is empirically well validated and also seems to match with my experience. I feel like many of these theories of learning have blackbox of like, "Okay, but which generalization will the animal make?" And as you point out, the animal could very easily figure out that there's some discriminatory stimulus which is you. You're there when the dog pees then there's a possibility of having a punishment. If you're not there, the punishment never seems to happen. That's a pretty intuitive thing for the animal to learn.
SPENCER: Whereas with reward, let's say you're trying to get the dog to rollover. Maybe initially you're doing shaping. So if the dog lies down you reward it at first. Then if it lies down and slightly rolls, you reward it. Over time, eventually you only reward it for a full roll. It's much less gameable because you're just there watching. So if the dogs are scheming situations and doing things that are not really on the trajectory towards rolling, you're just not gonna give it a reward, right?
DIVIA: They're pretty different situations. One of them is you're trying to get rid of a behavior and the other one is you're trying to construct a behavior. I think clicker training is good at creating behaviors and then you can even put them on cue so you can teach the dog to rollover. I think it is a different shape of problem if there's something that the dog is doing that you want the dog to stop doing. There's some list of humane ways to do that but I think it's a different challenge.
SPENCER: Can you use clicker training and reward to get a behavior to stop?
DIVIA: Yeah, absolutely. It's just different. For example, you could take the dog out regularly to pee outside then click and give the dog a treat when they pee outside. Because they only need to pee so much.
SPENCER: They're like, "Oh, might as well get peeing when I get a reward."
DIVIA: I think that's basically how that tends to work with the peeing [sic]. But other behaviors, for example, you could train on alternate behaviors. This is one where I could bring up different paradigms of how to relate to it but many people don't like. I have a dog who sometimes likes to jump up to greet people but people don't typically like this. I think the most classic clicker training answer is when you train an alternate behavior which is you train the dog to sit for a greeting, or to go to lie down on a mat and wait to be greeted. I think that can work super well to train an alternate behavior. That's one way to use clicker training to eliminate a behavior.
SPENCER: By the way, I don't know if I told you this but you really helped me with my cat by just giving me like one minute of advice.
DIVIA: Oh, no, I didn't know and I wanna to hear about it.
SPENCER: Yeah, so I have a large and very wild personality cat. He's just a house cat but his personality is a bit like a wild cat. He used to sneak up on me while I was working, very stealthily, and bite the back of my leg and he thought this was like the most fun thing ever. So my first attempt to stop this behavior was I got a bunch of little water guns and I hid them around the house. The idea was whenever he did this to me, I would immediately leap to grab the nearest water gun and try to squirt him. But he's very clever and he figured out that if he bites me and then immediately starts running as fast as he can, he can usually get out of there before I can hit him with the water gun. So this was a total failure basically. This is the punishment strategy. I'm trying to punish this behavior. When I was talking to you about this — I don't even know if you remember this because it was just an offhand comment and it was super insightful — you basically said, "Well figure out why he's biting you. Like, what's he trying to communicate while biting you?" When I actually did figure it out, I figured out that he basically means he wants to play. So I actually taught him that if he— so I have a weighing scale next to my desk and I taught him that if he sits on it, then I'll play with him basically. I might not have time to play with him very long but I'll play with him for at least a little bit, like 30 seconds at minimum and maybe a lot longer. Now, when he wants to play, he comes and sits on the scale, and I'll play with him and he almost entirely stopped biting me like that. So it's really cool.
DIVIA: I'm so glad! That's really cool.
SPENCER: Basically, it was just like he was trying to communicate something to me and then he didn't have a certain acceptable way to do that.
DIVIA: My dog used to, when he was a puppy, would just randomly pick things (probably not that random), but he would pick things that I didn't want him to have, and he would run away with them. That's similar, he definitely wanted to play and he didn't care about having the thing. Occasionally he did but usually what he wanted was for me to chase him. This is another clicker training thing, to put it on cue. So I taught him a cue where like, I would say, "I'm gonna get you." Then he picks up something then I chase him around. It's really fun. I think he did want to do this, but he didn't want to do it infinitely much. I think it'd be even cooler if I taught him some way to request it from me. I think I did try to be more attentive like, "Does he want to play signals?" and then play the game with him. Maybe there was some of that going on even not explicitly. But also me having a way to initiate it made it reframed it in his head like this game that we sometimes play that usually I initiate.
SPENCER: One of the things I think is really cool about the operant conditioning and clicker training more specifically, it's a way of communicating with a creature that doesn't have a language in common with you. Through the training, you learn to talk to each other in a sense. So now, I have a situation where my cat sitting on the scale can be one of three different things: it's how he communicates with me. If he sits on a scale when I'm eating food, it means I want some of the food you're eating so I always give him a little tiny bit. Often, he really likes chickpeas for some reason and when I'm eating chickpeas, he'll sit on the scale to get one. The second thing it can mean is, "I wanna to play." I know it means that because if I go to pet him when he's on the scale, he'll dodge me and that means he wants to play. The third thing might mean he wants head scratches, so if I go reach his head, and he leans into it, that means he wants a head scratch. It's really cool because now we have a different kind of language we built up.
DIVIA: That's awesome. I think you put it super well that it's a way to communicate in a pretty high bandwidth way without language.
SPENCER: With regard to that, I'm curious to hear your thoughts on clicker training and operant condition in the parenting realm, with humans instead of animals. I'm wondering, to what extent you have tried it and to what extent do you feel like it's a good strategy?
DIVIA: I have a lot to say here. I told the story about many people's dogs don't jump up. They don't want them to jump up but they do jump up then you can try to teach them alternative behaviors and stuff like that. So I read a different book. I forgot what it's called. It might be called "Love Is All You Need", about a woman who developed what she called bond-based choice training. She trained service dogs. She does actually get results with what she's doing. That is more like attachment theory for dogs. Maybe I shouldn't even say attachment theory for dogs because my understanding is that attachment theory, insofar as they've studied attachment theory, dogs do behave like toddlers. So what she says about dogs jumping up is if a dog jumps up on you — it's like your same thing — "Why are they doing it"? It's because usually they want to be close to your face and they want some greeting. Her thing is just to actually connect with the dog and pet the dog until the dog voluntarily gets down. Do it as many times as you need to. Her claim — I forget, if she made a prediction about the number of times whether like after you do this a dozen times, or 20 times or whatever, but she was like — if you do this consistently then the dog will feel secure and reassured and not as stressed about like, "I need to be near your face and figure out how you're feeling towards me right now" and they will naturally relax. I really like clicker training as a paradigm. In my head, I have it separated into, there's a matter of teaching dog guilt, or not even dogs, teaching animals skills. I talked to an animal trainer once and she was like, "If you've ever seen a routine at a circus with some seals or an aquarium..." She's like, "That thing where you teach the animal to perform in a certain way, that's pretty easy and you could probably do 90% of it in an afternoon. However, the part where there's a line of five animals behind that animal patiently waiting their turn, that's the hard part." I think clicker training does have a lot to say about that because it's still good to use the rewards instead of trying to use negative reinforcement or punishment which doesn't even really work. It is good to go gradually and shape it. But I think that those things — how to be calm and relax in different situations — then gets me into: Well, I think clicker training is one good paradigm for that. Attachment theory and things more like internal family systems, they certainly can't do the process with animals, are other ways that I think have a lot of explanatory power for what allows people and animals to be calm and relaxed in different types of situations.
SPENCER: The one just mentioned. What is attachment theory?
DIVIA: I'm probably not going to give the best answer ever, but it's the idea that — I think at least with [inaudible]. I'm going to talk about it with humans so — that a kid will have a bond with a parent and in what is considered the ideal case (according to the theory), it's pretty secure and that is associated with certain behaviors. There's some classic experiment people do which is called I believe this "Strange Situation Tests" where the person, the caregiver leaves. So, if my daughter's there, then I leave then some other person comes in and then that person leaves and then I go back to my daughter, how does my daughter react? There's certain behaviors that are characteristic of what they call secure attachment, which is basically, it's safe to be vulnerable, it's safe to get close to this person and it's safe to express your feelings. This is supposed to be the good thing that allows people to self-actualize more. Then there are some other attachment styles like anxious attachment, avoidant attachment, disorganized attachment. I think this is super clear in the worst case where people get diagnosed with reactive attachment disorder which I think is pretty common in kids who did not have any stable caregiver situation when they were young. And it can cause a lot of problems.
SPENCER: So anxious attachment, that's the idea of being really uncertain of whether you are getting love, or you're going to get love and you want to constantly be told that you're good or that you're cared about. That kind of thing?
DIVIA: Yeah, totally.
SPENCER: Then avoidant is like you just don't even want to get involved emotionally because you're worried that you'll be abandoned or the person will stop loving you. It's like people who are less likely to even create stronger passwords in the first place to protect themselves.
DIVIA: That's right. Classically, you'll often see relationships where, even if it's not to some clinical level, one person is more the one who's playing the role of, "Oh, I'm independent. I don't need anybody." Then somebody else who's playing the role in the relationship of, "Come on care about me. You deem me. I'm important." I think a lot of relationship conflicts come down to those types of dynamics.
SPENCER: Got it. So you're saying, if you have a situation where maybe what's going on is an anxious attachment or avoidant attachment, then maybe the clicker training or operant conditioning theory is not the right way of viewing that situation. It's not about rewards, or rewards and punishments. It's about, "Can you create a more secure attachment?" Is that the idea?
DIVIA: Yeah. I think it's interesting. Again, I think the clicker training paradigm actually offers fairly useful advice for this situation. But I don't know if it's always the best advice and it's certainly not the advice that I would personally always reach for. So when I think with kids for example, since you asked about that, I think that most of the things that people object to in terms of kid behavior is downstream of the kids being stressed in some way.
SPENCER: Oh, really. Interesting.
DIVIA: Yeah. A lot of people describe it as attention-seeking. Maybe that doesn't fit your experience but that is something that I think about. Most of the conflicts that I tend to have with my kids about things that I don't want them to be doing, I do think it's because they're stressed or upset typically.
SPENCER: Got it. So it's not just that they want something or they want attention. They're actually trying to relieve some negative state.
DIVIA: Well, I think that they're hoping the attention will relieve the negative state. So I do think that they do legitimately want the attention. And that if they were really calm and relaxed and happy, then if they did want the attention they would probably be asking for it in ways that were easier for the parents to deal with. If they expected that to work.
SPENCER: Where do you see the clicker training ideas, even if you're not actually using a clicker (I assume), as being useful in parenting situations?
DIVIA: The situations where I think it has paid off probably the most for me, this is an ironic example, but I remember teaching my daughter how to use the baby gate at one point. Because we had a dog and I did want the gate there for the dog but she was old enough where I wanted her to be able to open the gate. She was not super enthusiastic about learning about this. Some of my kids have more naturally figured out baby gates than other ones. But in this case, I remember being very focused on what exact physical motions she was making and very precise about being like "Yes", "That", "Exactly" when she was doing the ones that were right for opening the baby gate and breaking it down so that she was learning one physical component of the skill at a time. I thought that that had a high payoff. I feel like it didn't take that long for her to figure out how to use the baby gate. It was because I was teaching her, she'd been fiddling with it herself and it hadn't been that successful. It was because I was paying attention to those principles.
SPENCER: Those rewards in this case were just saying, "Oh, great job!" or communicating positivity?
DIVIA: In my memory is that I'm not opposed to paying my kids to do things and they know this. However, I think it's super interesting that often they're not very interested in me paying them to do things. It seems complicated. Another thing I remember I did ask a dog trainer that I spoke to for advice about different parenting things. It seems obvious when she said it but she was like, "You have to notice even just the really small things and draw their attention to them and be grateful for the really small things." That has been really true. For example, let's say my kids are fighting with each other and I see one of them take an unusually deep breath. You should point it out to them and be like, "That's great. I'm so glad you did that. That seems like a really helpful thing that you did." Or just really little things. Because I think it's easy as a parent to have some idea in my head of, "Oh, I wish my kid were doing something that's 180 degrees different from what they're doing right now," and to miss when they make a one degree shift towards something that I think would be more functional.
SPENCER: That makes a lot of sense and that seems like it's just an application of shaping from operant conditioning theory. The way you get a dolphin to go through a hoop is not to just teach it to go through the hoop but to teach it to first bump the hoop with its nose and then to put its nose slightly more through the hoop and so on.
DIVIA: Shaping and the idea of focusing on the behavior that I do want instead of I'm trying to get rid of some behavior that I don't want.
SPENCER: Well, that's what I wanted to ask about here. It just seems to me that there'd be a lot of applications and things like, "Oh, I wish my child did this behavior more." Let me just try to notice — exactly you're saying — that they're taking a little step towards more of that behavior and just rewarding them probably with attention or praise. It could be other things. It could be money, or treat or whatever. That just seems like it would have a lot of applications but maybe it doesn't have as many as I expect. What do you think?
DIVIA: No. I think it does and this is where some stuff about decision theory seems to come in. I would not say this is some ideal example of my parenting but I remember once with one of my kids. I really wanted that kid to clean up some toys and the kid did not want to do it. I remember I was offering money, and I was like, "Okay, well, if I paid you..." (I forgot what I offered but to me it seemed like a few dollars.) I thought it would be a lot of money for somebody spending a few minutes picking up some toys. My kid was like, "No." I tried to dig into why, and the kid was like — I don't remember the exact words but it seemed like — a policy decision to the kid. Because it's like "What if you ask in the future and then I always say yes then I spend my whole life cleaning." I don't know how much dogs think of it that way. I think it's harder to tell because they would never say it in words. But with my kids, I think it's true that I do have some influence in terms of offering upside even if it's just attention and me saying "thank you" and appreciation. I can use shaping and I think they are also tracking how much influence do they want me to be having, or how much do they want to have autonomy basically, and how much do they want to be making the decisions about what they want to do with their life.
SPENCER: Hmm, that makes sense. So sometimes they just don't want to be shaped.
DIVIA: Yeah. I think if I'm pretty relaxed about it and I'm using positive methods, they're not going to necessarily mind. But I do feel, unlike how my dogs seem to be, my dogs seem endlessly interested in doing things that I will feel like doing which I guess is probably because we bred them for that. Whereas my kids, it seems like I pretty quickly come up against this desire for autonomy.
SPENCER: That makes a lot of sense. Some people have reactance as a personality trait when someone tries to get them to do something too hard and they'll actually want to do the opposite.
DIVIA: Yeah. We've talked about the four quadrants for operant conditioning. I listened to a podcast with a dog trainer a while back, where she talked about how she had been training her dog to let her cut the dog's nails using treats which is pretty common. I think it's a basically good idea. I do it with my dog because the grooming behaviors are for the health and happiness of the dog as well as me, but it's something dogs typically start out feeling uncomfortable with. But I'd go slowly and shape them and give them treats, so I can cut their nails. So great. But she didn't have any treats so she thought, "Okay, well, my dog also likes tugging and maybe I can use tugging to do the training and work on..."
SPENCER: What's tugging?
DIVIA: Just like, you take a tug toy and play with the dog. Many dogs love this. In fact, what her impression was that the dog liked it enough that it was not working. Because the dog was getting too stressed about like, "Okay, I have to put up with this nail trimming in order to get to the tagging which I really really wanted." I think her encapsulation of what's going on is that you tend to get fallout if you're offering a really high stakes reward for something that the animal is conflicted about or has trouble doing.
SPENCER: Is it so stressful because it wants the reward so badly?
DIVIA: Even though it is, I don't think it makes sense to think of it as negative reinforcement or punishment. I think it basically is, before the animal didn't have the opportunity and then you're offering the opportunity. But I think this aligns fairly well with people's intuitions about what seems coercive or bad or unhealthy even when the thing that's being offered is purely upside.
SPENCER: I see. If you're offering too big a reward for something, it feels like it's coercion because the person can't say no to it. Is that the idea?
DIVIA: Yeah. Let's say I was like, "Oh, hey, Spencer, what if I give you $100 million to cut off your pinky finger?" I think it's possible to be the person where you're like, "Okay, well, that's fine. Because I can just say no to that if I don't want to cut off my pinky finger." But I think it does create some tricky response [sic]. Like a response that many people find tricky to handle in themselves if faced with this question.
SPENCER: Right. That's really interesting.
SPENCER: Another application of operant conditioning related to parenting that I wanted to ask you about comes up with bad behaviors that might be reinforced by parents. For example, I'm curious what your reaction is, but let's say you have a situation where the kid learns that if they cry the parent will cave into their request. So the kid really wants this toy and the parents don't wanna give it to them. But then the kid starts crying and the parents feel embarrassed because it's in public and so then they buy the toy. I suspect that happens a lot. What do you think?
DIVIA: This is near the limit to me of where I would even want to use the operant conditioning paradigm. Because I think the attachment theory paradigm has a lot to say too. Because I think ultimately expressing distress is an attachment promoting behavior. I also share the intuition that in some of these situations, I have some sense of, "There's something bad going on here and the parent shouldn't necessarily be getting the toy for the kid just cause the kid's crying." However, if I try to formalize that, I mean people used to say, "Don't pick up your babies when they cry or they'll cry more" which I think that's why it seems important to me to have some conceptual distinction. Because that's really messed up and not how people should relate to their kids. I think if kids feel safe expressing their distress to their parents, that's a good thing. I don't think it makes sense to see crying as a bad thing. I think it's also true that in your example, it wasn't that the parent was like, "Oh, wow. Now I get how much my kid wants this. But now I feel embarrassed that my kid is crying." Right? To me, that distinction does matter. I think it's how the parent, the behavior is reinforced by the parent. But to me, if I imagine some counterfactual situation where let's say my kid wants to have a piece of toast, and I'm like, "Oh, well, we don't have any." And my kid's sobbing. My first response is, "Wait a minute. This is way more important than I realized." I think it may end up with me being, "Okay. Well, let me just place a grocery delivery order right now, or let me go to the store because this obviously mattered to you more than I thought it did." So I think there's some interesting question of what is the distinction between a kid feeling comfortable expressing distress to the parents and feeling safe and the parents updating on the importance of whatever it is to the kid, versus something that can look more like a manipulative dynamic that people suspect there's something wrong with.
SPENCER: It seems so tricky. Because on the one hand, insofar as your child actually cares about the thing and you want to satisfy their desires. On the other hand, you don't want to create a dynamic where they essentially have an incentive to manipulate you. How do you navigate that? Do you think that that is a significant risk that kids are essentially trying to manipulate you by rewarding them, essentially, when they do express that they really want something, or they start crying, or they throw a tantrum?
DIVIA: I think that the bigger risk is me trying to manipulate them into relating to me in some way that seems less embarrassing or more acceptable to me. I think I'm more concerned about that risk than the other one.
SPENCER: Can you elaborate on that risk?
DIVIA: Sure. Again, I think the most extreme example is the people not picking up their babies. I definitely do track both of these things but I have sometimes tried to notice my behavior in the moment and being, "Okay, it seems like I don't want to do this for my kid" because I want them to be acting differently in some way that has felt manipulative on my end. I do worry about that.
SPENCER: I see. Well, just to give you an example of my cat because I don't have any children. When we go to bed, my cat really wants to get in and hang out with us. Basically, I've just instituted a policy. After bedtime we never open the door for the cat no matter how much he scratches the door or whines at the door. That seems to work really well because it seems like if we were to open the door in that scenario because it's actually annoying to have them scratching there and it's disruptive. Then suddenly, he learns, "Okay, so I can get in the bedroom if I just scratch hard enough." Then now we're training him to scratch in a way that's really annoying and that's gonna be disruptive enough that we open the door.
DIVIA: Yeah, definitely. I think it's good to be aware of those dynamics of which behaviors the cat is using to get you to do the thing that the cat wants. But to me, more of the interesting part is getting clear on how much do I care if the cat's in the bedroom? What is my actual proposed policy about that? I've personally find [sic] that when I get clear on these things for myself then most of the problems tend to resolve themselves without me trying to explicitly focus on what I'm rewarding or not. I don't know if that's been your experience at all.
SPENCER: Wow. Just to give it a side, the cat example. One time he was hiding and so he actually ended up in the bedroom without us realizing it. The first time I'm woken [sic] up at 3 am with him biting our hair and then at 5 am I'm woken [sic] up again with a very strange sound. It was like a repeated clicking sound like "clu-clu-clu-clu". And I ask, "What is going on?" I opened my eyes and the cat is sitting onto the keyboard just typing endlessly. The monitor of my computer is on and just blaring, filling the bedroom with light. I'm like, "Oh my gosh." So yeah, that didn't go so well.
DIVIA: Yeah, that's really funny. If you're clear that it really absolutely does not work for you to have the cat in the bedroom at night and then that's non negotiable, then that makes sense. Not responding to the cat trying to get into the bedroom also makes sense. I do, as with the other thing, I think it's always interesting to ask, "Why does the cat want to be in the bedroom? Is there some other thing I can get for the cat that will resolve this?" But let's say there isn't. Whereas for me, there feels like something very important on a relational level to the first time the cat is scratching and scratching and scratching that my first priority is, "Okay. I should make some relational update that the cat cares about this more than I thought. Maybe I should be more willing to do it because the cat cares more than I thought." And actually consider that. To me, it feels like skipping a step if I'm like, "Okay. Well, I can't get into that." To me, it seems like potentially throwing out a lot of information about strength of preferences. It's something that I think can be bad for relationships.
SPENCER: I think that's a really nice way to put it. If someone's engaging in a behavior, whether it is your child or your cat, that seems disruptive or harmful, the first question is like "Why is this behavior happening? What are they trying to get? Does this mean they care about the thing more than I thought? What's going on here?" Then there is eventually a point where you retreat to, "Okay. We just need to not incentivize this behavior because we understand why it's happening and there's not really a good alternative." But in other cases, maybe there's actually just a way to resolve it like can you give them a thing they want in a way that isn't disruptive.
DIVIA: Well, it's interesting that you said incentivize because I think that this game theory paradigm and the reinforcement learning paradigm makes somewhat different predictions about how to relate to this. For example, the reinforcement learning paradigm will say that if you wait until it's been however many seconds of the cat not scratching at the door, it's fine to let the cat in. Whereas, I think a game theory perspective would be admitting the possibility that maybe the cat will figure out, "Okay, well, so I've scratched on the door for a while and I have to be quiet for a while. Then maybe that gets me what I want and maybe it's worth it."
SPENCER: That's funny. Well I actually do that. If for some reason I do need to go out, I'll specifically wait for a little while until after the cat scratches to not create an incentive. But from the game theory perspective, I guess game theory models agents as perfectly rational. They have the ability to do things like figure out patterns. Like, "Oh, if I scratch and then wait five minutes then that will actually open the door."
DIVIA: Yeah. Whereas the reinforcement learning paradigm is more like a model that the animal maybe could learn about a delay but it's very counterintuitive and the timing does matter quite a bit.
SPENCER: My sense is, in that paradigm, doing something one second after a behavior, like giving a reward one second after the behavior is better than giving it three seconds later. Giving it three minutes is dramatically worse than three seconds to the point where it might even be hard to teach it. Giving it three hours is just hopeless. There will be no learning essentially, three hours later.
DIVIA: Yes. That's right. About the things with the cat, I could see that helping a lot in the case with the cat and it seems like in your case it did. But I think this is not a clear example of where this always works out well. I hear plenty of stories about my neighbor's dog is out in the backyard and is just constantly barking. I think that making sure never to externally reward this behavior in a way that anybody notices is probably not going to solve the barking.
SPENCER: It sounds like the dog will just keep doing it regardless of whether you reward it or not. There's something else going on.
DIVIA: I think that's right. In particular, many of these behaviors are where we start to hit the edges of, "Is this a useful paradigm?" Because what people will often say is, "I believe in Don't Shoot the Dog, Karen Pryor even says that she thinks barking is often a self-rewarding behavior."
SPENCER: Hmm, interesting. By the way, I love that book, Don't Shoot the Dog. Definitely recommend that for people interested in operant conditioning, clicker training, stuff. What's the self-reward going on there with barking?
DIVIA: Again. That's a weird way to put it if it doesn't make that many predictions. But I think it relieves tension. Like certainly, I might cry for social reasons. Crying is not the same as barking but as an analogy. Sometimes if I'm by myself, I might cry too. If I introspect on getting upset about something including maybe expressing it and getting loud. I think sometimes it has a social component that is responsive to whether people are actually doing it. Sometimes it doesn't and sometimes it's more like self-signaling. Another paradigm that I think about a lot, I got this from some Personal Growth MP3 I used to listen to where this guy talks about the [inaudible] theory of chaos and reorganization. I got out of it what I got out of it. But that's a different paradigm where it's like, "Okay. Well, systems can only process so much information. Then once they get past that point, they either try to offload information for external processing or they try to shutdown and get less information from the environment." That theory does not predict that the offload for external processing will be super responsive to whether it's actually working.
SPENCER: Do you have some examples of that?
DIVIA: Well, the dog barking, to me, fits into this. The dog is somehow upset with the situation and can't solve it on its own so it's generically getting loud. That puts it out there to the universe. Something else changed which could mean, one way this could get resolved is maybe the human is like, "Okay. Well, actually the dog wants to be inside." And brings the dog inside. Or maybe a thing that could happen is other dogs — I don't think this is very realistic but... No, no it could be. — Maybe it's stressful for the dogs to have other animals walking past. Maybe if the dog's barking all the time, the animals will stay away. But it's a generic, if I anthropomorphize it, request from the universe to change the situation.
SPENCER: With that chaos and order thing, what would an example be for a person?
DIVIA: If somebody's screaming in the grocery store for some reason, I feel like I tend to assume that something has gone really wrong and they can't handle their life at this moment. So they're being expressive and maybe something will change now that they've done that.
SPENCER: I see. What would it look like? You said that there are two different strategies, one of those is offloading. What's the other strategy?
DIVIA: Taking in less information.
SPENCER: And what would that look like?
DIVIA: If I'm really upset, I'd go lie down in the corner and close my eyes. Rollercoaster is another good example. Many people, when they're on a roller coaster and it is overwhelming, they scream. I don't know that they really...
SPENCER: [interrupts] Or other people shut their eyes.
DIVIA: Yes and that's more me. I'm more like, "Okay. Let me shut my eyes and make sure to tense all my muscles" which I think does make me get a little bit less input from the outside environment when my eyes are closed and my muscles are tense. It doesn't work super well. But I absolutely have this instinct on rollercoasters.
SPENCER: I do this with horror movies which is not something I enjoy. But if I happen to be watching one, I like squinting my eyes to lower the amount of visual information during some scenes.
DIVIA: Yeah. And some people will scream instead, right?
SPENCER: Yeah, absolutely. That's interesting. I never really thought about it that way. That's cool. So we've talked about these different theories today. We talked about internal family systems. We talked about operant conditioning and then more specifically, clicker training. We very briefly talked about attachment theory, and also just a little bit about game theory. How would you tie these systems together?
DIVIA: Another thing that I think about a lot that to me feels like it has a lot of promise for tying these things together is decision theory.
SPENCER: So what is the decision theory?
DIVIA: It's meant to be an upgrade for game theory that produces better results. For example, one formalization people talk about a lot is Newcomb's problem where you have some agent that's supposed to be infinitely good at predicting your behavior that has given you two boxes. Box A is clear and always contains a visible $1,000. Box B is opaque and its contents have already been set by the predictor. If the predictor thinks that the player will take both boxes, then the box contains nothing. If the predictor predicts that the player will only take Box B, Box B contains $1 million. The deal is he's like, "Okay. Well, if you're the agent that would pick only one box then that box has a lot of money in it. If you're the agent, and I know you really well, that would take both boxes, then you're not going to get any money." It messes with people's intuitions a little bit because there's causal decision theory as opposed to the timeless decision theory that we're looking for. There's causal decision theory intuition that's like, "Well. Okay but he's already put the things in the boxes so maybe he thinks he knows me so well. But at this point, shouldn't I just take both boxes? Because then I'll get more money."
SPENCER: Got it. Basically, you're standing in front of these two boxes. You know that the boxes either have money in them or they don't. So there's nothing at this point that can change, and you think, "Well, I might as well take both boxes because everything's set at this point. It's either there or it's not there. So might as well take them both." But then if you're the person that would pick both boxes, the problem is that the predictor will have not put a million dollars in one of the boxes. So you'd rather be the person that would just take one box even though given that things are set right now, it seems like you might as well pick two.
DIVIA: Yeah. I think it's a pretty artificial problem, of course, and it's not possible in real life to have a perfect predictor. But I do think it has some relevance to actual real world situations. Andrew Critch has a blog post about this where he's like, if we accept the premise, then people think, "Okay. Well, I guess I should get one box because I'll get more money that way. That way I'll get the million dollars." But he says, "Can you point to the exact flaw in the two boxers' argument?" Then what he says is the exact flaw in the two boxers' argument. Assume this. If we accept the premise that this thing is a perfect predictor. Then when you're deciding "Which box should I take?" It's silly to think that you're deciding only for the real world after the fact he's already put the money in the boxes and the phenomenology corresponds to that and not the other thing where he's sitting there imagining, "Okay. Should I put the money in the box or not? What is that person going to do?"
SPENCER: So if you get to choose whether you're the person that picks both boxes or one box, then you're actually determining the reasoning process going on in the mind of the predictor as well. Because by choosing to be the person that picks just one box you're also determining the result of the algorithm that the predictor's running to predict you. Is that the idea?
DIVIA: Yeah. I think there's something pretty deep about it. About not identifying with the calculation that's going into deciding things in my own head but identifying as being a distributed algorithm that runs across different people. And that when I have some phenomenology of, "Okay, I'm making a choice." I should assume that that applies for both the me that gets to control my body and the me that is the simulation living inside other people's heads.
SPENCER: How would you then take this idea of decision theory which is this general theory of how we should make decisions and connect it to the different psychological theories we were talking about before, like IFS and clicker training, and so on?
DIVIA: I think it's more obvious to see with clicker training, how there's something about it that can break down around the edges. I think that with an animal that's not very smart that I have a limited relationship with, that's basically fine. But as we pointed out, it's possible to imagine a cat who's gaming the system more and that's in the more adversarial case. But it's also too impossible to imagine. Sometimes people have really good relationships with their animals and I don't have personal experience with this. But people will tell their border collie to go over somewhere else and do something with some sheep and trust that their animal will do it. I think the problem with some of these paradigms, like operant conditioning, only interfaces with the reinforcement learning layer of somebody. And game theory assumes you're interacting with something more conscious. But I still think it doesn't allow for the fact that people are actually super good at reading each other. Like in a prisoner's dilemma situation, where both people are right there, I think people can often look at each other and tell what they're going to do, and that most interactions that people have aren't as cleanly as something like, "I'm in some place where I can't communicate with you and we're independently making choices." And I think Attachment theory gets at that better. It's like, "Okay. If you open up to me then I'm not going to hurt you." And it's trying to make it incentive compatible, to share information and be vulnerable with people.
SPENCER: To expand a little bit on what you said. You have this idea of a prisoner's dilemma, where in essence it's a situation where two parties are strictly better off defecting on each other. Yet, the scenario where they both defect on each other is actually not a good scenario. It's not the best outcome. And it'd be better if they somehow could both pre-commit to cooperating because then instead of getting defect-defect where they both screw each other over, you would get cooperate-cooperate. So even though each of them has an incentive to defect and regardless of what the other one does, they actually better agree to cooperate. You're saying that in the real life scenarios because we have this ability to read each other and we can learn to trust each other, if we can develop a secure attachment then we can actually avoid prisoner's dilemmas a lot of times or at least both cooperate in them. Is that what you're getting at?
DIVIA: Yeah. It is what I'm getting at. When you say pre-commit, I think that's the causal decision theory type intuition whereas if I take a more timeless view that I'm going to just use this abstraction when I have a high trust relationship with someone, I'm going to assume that the decisions I'm making are for me, and for the copy of me in the person's head. I do think about it this way in every life. I remember one time I was at a party and I told Will, my husband, "Okay. Will, I'm going to go but I'm not going to stay that late." And he's like, "Okay, I'm fine with that." Then I go and I'm having a really good conversation and I'm like, "Oh, maybe I'll just stay later but I told him I'm not staying that late." We didn't have an exact agreement about the time and it clicked with me, and I was like... One scenario is that I'm here at this party and if I decide that I'm going to keep talking to this person then I get to talk to the person. And another thing that's also the same amount going on here is that I am Will, imagining me, being like, "I wonder what would she do if she had a really good conversation? Because if she's gonna stay really late if she has a really good conversation, I don't know if I'm actually going to say I'm okay with her going to this thing." So it felt like a weight off my mind that I'm not making a decision for one of those things. I'm making a decision for both of those things. Because Will and I know each other really well. We interact all the time and we're going to iterate this tons of times. I think it's basically the right abstraction to just collapse those into being the same thing.
SPENCER: I guess one way I've heard it, it's said with timeliness decision theories. When you're making a decision, your choice is choosing for all sorts of sufficiently similar algorithms to what you're running. So whatever decision process you're using, you're choosing for all such decision processes including Will's model of you in his mind, assuming that that actually is an accurate model of you or reasonably accurate.
DIVIA: I don't think I have some perfect theoretical model but this is some of how I tried to tie this together in my own head. Both with the IFS parts and the attachments to other people, and when it seems right to use the clicker training. It's like, "If I want to have a trusting relationship, would this work? If what I were doing is also.... They have some simulation of what I'm doing and what I care about and they're shared? Does it work to assume that sort of trust? Or am I breaking some sort of trust?"
SPENCER: Just to unpack that a little further, when you're making that decision of, "Do I stay later at this party?" You're also thinking, "Well, this is going to essentially affect the model that Will has of me in the future. And it may actually affect him saying he's comfortable with me going to certain parties." So you're deciding those two things simultaneously. You're deciding on behalf of whether you want to be the priority now but also on the behalf of how Will will think about you going to parties in the future. I think you're saying you just collapse those two things together and don't view them as separate choices. Is that right?
DIVIA: Yeah. It feels different. It really feels hard to talk about. But it feels different to think, "Okay. If I decide that I stay, then I get to stay this time but maybe not next time." I think that's the wrong way to think of it and it's relating to time in the wrong way. Because what's closer in my head is, either I decide that I want to stay longer and talk to this person and then what happens is I get to talk to the person or I decide that I want to stay longer and talk to this person and what happens next is Will's like, "Yeah. I don't think I'm okay with you going to the party." It's not about next time. I should have some uncertainty about what causal thing the phenomenology is hooked up to.
SPENCER: So you think decision theory as a framework can help bring together these different psychological theories. Is that right?
DIVIA: I think it does get it, at some piece of it. Because I think that there's some way that they're all the wrong abstraction and there is something that is more timeless that is a more parsimonious way to model some of it. And that when they appear to conflict, like, "Okay. What if the kid's crying for the toy, do I give the toy or not?" For me, the decision theory helps because I think of it as, what's going on is that we're in this toy store and maybe the kid wants the toy. If they cry more, they're gonna get the toy and what's going on is maybe the kid's sitting around thinking, "I wonder if mom will get me that toy. I wonder if mom would get me that toy if I really wanted that toy and if I cried about that toy." Those are both happening and it makes the whole thing seem important in the right way but less weighty in the way that I think is ultimately a confusion.
SPENCER: What is that way where you think there's a confusion?
DIVIA: I think the way that it's a confusion is to think that I get to have those things be separate from each other.
SPENCER: That's a really interesting way to put it.
DIVIA: Of course I could try to optimize for those things being separate from each other by limiting the amount of information that was flowing back and forth between me and my kid. But I don't want to and that's not the relationship that I would want to have.
SPENCER: We go back to the different theories, just for a moment. I just want to point out how they model things. Internal family system says we're a bunch of agents and these agents have different motivations and goals. They all have positive intent but they're all trying to achieve different things that can sometimes create problems for us as they can get in conflict with each other or can cause us to behave in ways that we're confused about or we regret. Then we have operant conditioning which says we just respond to rewards and punishments. So we will take an action more and more if we tend to get rewarded for it. We'll take it less and less when we can get punished for it. Then we have this attachment theory model which says we want affection and love. So if we feel like we may not get it and we're gonna keep making bids for it, or if we feel like we just can't get it all, we're going to avoid situations where others will reject us. Then we have this game theory model that says we're all rational agents and we're playing these games or we're trying to get certain outcomes. We're just trying to model, "What if I do this and the other person does that? Then I'll do this other thing." Return to it and reason out far into the future by modeling everyone else's behavior and maximizing expected value. So these are four different frameworks for thinking about the way humans or animals interact with each other. I think what you're saying is that decision theory is the more of the general case, maybe? Whereas these are special cases.
DIVIA: I don't know if it fully is, but I do think it makes a prediction which is something like trying to... Insofar as things are going to be adversarial and I'm going to try to make it so that I want one thing to be my real decision, but I want some other thing to be what you think I would do, then the way to get that is by limiting information flow. I think this is very relevant because the clicker training people, according to me, they don't have a super mature framework for talking about this. But one of the things they say is that the behaviors that are learned with positive reinforcement tend to be better integrated and accessible to the animal in different scenarios. Whereas behaviors learned through punishment tend not to be like that as much. They can generalize a lot in terms of, "Okay. Now the animal is generally shut down and doesn't want to do things because it might be punished." But if I teach a behavior that way, they're less likely to experiment with it and integrate it. There's echo of the internal family systems thing where you're like, "Okay. Well, but at what point does it make sense to model them as separate agents?" I'm like, "Well, more so when there's trauma that causes something to split off." Because I think there's some sense of, "Okay. This is not safe for the system. I have to limit information flow." Then something starts to look more like a different agent.
SPENCER: Maybe what you're saying is that in this high information flow, where everything is really integrated, behavior becomes more like decision theory. Whereas as we limit information from different parts of ourselves, it starts looking more like IFS?
DIVIA: I think ultimately decision theory should cover all of these scenarios. It's just that the formalizations that I've seen mostly talk about the ones where people can predict each other's behavior well. At least I haven't read much stuff about the case where people can't. I think decision theory to me seems like the broader paradigm but there are parts of it that I don't really know what they say. But I have some intuition that it's going to tie these things together better than what I have now.
SPENCER: I guess one way I think of these different theories is you can have lots of different models for things that are all accurate in some domains. Even though they may seem really different, they all can have their use cases. In situations where we're following a gradient of reward and punishment, operant conditioning can work really well. But in some situations, it's not rewards and punishments that are driving behavior. For example, maybe it's more about some long term goal we're trying to achieve and then I think operant conditioning might not fit the situation very well. Or similarly, if we're in a situation where, let's say we're really craving affection and care. Then maybe attachment theory is a good model for that. But in other cases, we're just seeking pleasure or food, maybe operant conditioning is going to outperform at predicting what we're doing. What do you think about that? Do you think that's a good way of looking at it?
DIVIA: Well, so I think there probably is reinforcement learning going on inside your brain even when it's those other things. I think of it more like physics, biology, and chemistry. I think that physics is always driving all of these things. But if I don't understand whether I'm gonna get sick with the Coronavirus, it doesn't make sense to try to model that out using physics. However, sometimes physics is useful at a higher layer of abstraction, like if I'm trying to throw a ball and see how far i'll go. Then I want to use physics again, not biology.
SPENCER: Got it. So it's how zoomed out or how much detail is there in the model?
DIVIA: Yeah. And at what point is the reinforcement learning going on in some complicated self-referential way that's pretty hard to model and at what point is it going on in a pretty legible way that's easy to model.
SPENCER: Divia, this was super interesting and thanks so much for coming on!
DIVIA: Thanks so much for having me.
Click here to return to the list of all episodes.
Sign up to receive one helpful idea and one brand-new podcast episode each week!
Subscribe via RSS or through one of these platforms:
Host / Director