How exactly people learn to take things as object

a pretty concrete thing I think I learned from animal trainers, especially one animal trainer

Epistemic status: minor mashup of some Kegan stuff and a clicker training idea I got from Hannah Branigan (and less so but also Ken Ramirez). I think it’s pretty true? But it’s not like I’m that sure. I’ve gotten practical results from it a bunch of times in a row though

tldr; by practicing 3-5 different non-button-mash-y ways to relate to the thing

Preamble

I have been obsessed with behaviorism and clicker training since at least whenever I first read Don’t Shoot the Dog¹. I’m the type of obsessed with it where people ask me why I’m so interested and it’s tricky because I can’t quite relate to how other people don’t seem obsessed with it. And while a bunch of my interests (rationality, economics, psychology, adventure games, fantasy books etc.) are pretty popular in the communities I hang out in, I don’t tend to hang out with the other people who are obsessed with behaviorism. Which means that trying to learn what’s up with it for real has been pretty slow going for me, despite my longstanding intense interest.²

I will tentatively and medium confidently claim that, at least in my head, it’s finally all starting to come together. Mostly because I found Hannah Branigan, who explains things in ways that really really work for me. If anyone else wants to check her out, I recommend her podcast. By come together, I mean that when I ask myself “but how could I teach my dogs x”, I’m ~not getting stuck anymore. It’s not that my plans always work, and I know I’m clumsy at the mechanics and execution, but the ontology compiles! and not just in a piecewise way! It’s very exciting for me. I have some strong impulse to try to “write it all down”, that I hope to try to follow some day, but I don’t currently have a plan for connecting up the big picture yet, so I’m following the advice of a wise friend of mine to “start in the middle”.

I think this thing I have about how to take stuff as object is a good middle-ish place to start. It’s a chunk that I can imagine being useful on its own, and I’m also hoping it’ll be a foothold I can use to share more of the important bits I’ve learned about how behaviorism fits together.

I don’t know how to write about it with the intent to be useful though. I might know how to write about it in order to get my ideas out of my head, so that’s what I’m going to try to do. I hope anyone reading along either enjoys reading it or wanders off to do something more interesting.

Taking Stuff as Object

Taking stuff as object, or experiencing a “subject-object shift” is a Kegan thing. A friend of mine once posited that the subject-object-shift might the fundamental unit of psychological development.

I didn’t find a citation I liked better than the google overview, so I’ll quote that here:

In Robert Kegan's theory of adult development, the "subject-object shift" refers to a developmental process where individuals move from viewing things as part of their identity (Subject, or "I am") to viewing them as external objects (Object, or "I have") that can be controlled, reflected upon, and changed. This shift is crucial for personal growth and maturity, allowing individuals to develop a more complex understanding of themselves and the world.

I’ve read some Kegan, but I mostly learned this model from hanging out with a bunch of friends who use it a lot. I’m more confident that I know what “we” mean by it than that I’m using it how Kegan is, but my guess is it’s that it’s pretty close. I think it’s also closely related to the idea of being able to “unblend” in the IFS sense.

When I use the term, I mean the thing where at first I don’t know how to even notice a phenomenon as separate from “how everything always is”, and then something changes and I do. Some sort of concept gets newly allocated and I’m like yes, huh, oh, okay there’s a thing there. The fish learns that water is a thing.

Many years ago, when I used to see an IFS therapist, I said something to her about how I couldn’t be there for people if I were being emotional, she asked me about it, and I was like “yes this seems obviously true to me” until she walked me through a brief process and also pushed back some with examples until I was like “huh hmm okay yeah I guess that’s a belief I could have and relate to in all sorts of ways other than assuming it’s obviously true”. I think that was an example of a subject object shift.

A year and a bit ago when one of my kids was a newborn, she was upset about the pattern on my friend’s³ pants. My friend and I thought maybe the issue was that my baby was experiencing an uncanny valley thing about whether the pattern looked like nipples or not. We had a fun time drawing some patterns on paper and showing them to her. And then my baby seemed to be okay with the pants. And it seemed like something cool happened. My friend’s impression was that my baby had had a mini subject object shift about her nipple recognizer, or something like that. So maybe that’s another example.

Kegan talks about broad developmental patterns like how below a certain age kids basically don’t relate to their impulses as things they could be aware of and choose whether to act on, and then above a certain age they do. (I’m sure I’m not doing his theory justice, and I encourage anyone who is actually interested in it to look it up.)

What I learned from animal trainers about how taking stuff as object works

Hannah Branigan uses the catch phrases: “different but not harder” “different every day” “make lateral moves”. She most often says these things when people are describing places where they and their dog are getting stuck.

Like maybe my dog can sit perfectly well when I’m 3 feet away, but at 3.1 feet away he acts like he has no idea what I could possibly be asking of him. Or my dog lies down on his mat just fine when I go outside myself and ring the doorbell, but I have no idea how to get my dog to do that when I have a real guest, especially given that I don’t have any real guest who is willing to do a bunch of dog training experiments with me. Or my dog will happily chew his chew (for treats) when I’m sitting next to him, but I stand up and he looks at his chew like why would I ever chew that.⁴

That’s when Hannah Branigan says to change something else that isn’t the sticking point. Ask the dog to sit at 2 feet but wear a hat. Or move the furniture around. Hold your hands in an unusual position while you ask. Turn off the lights. Hold the treats in your hand instead of in your treat pouch. The changes can be almost anything, but try 3-5 different things. And then, she says, you’ll have a much better shot getting the dog to do it at 3.1 feet.

Ditto with preparing for the guests to come through the door. The theory is that it should help a lot to have the dog lie down on the mat when a few different human family members come in, when my other dog comes in, and when I come in holding an umbrella. I haven’t personally tried this (I usually keep my dogs outside when I know people will be coming through the door), but the dog with the chew story is a real one about my dog Argos.

I was trying to get him to be able to chew his medium value chews in more situations. I reshaped the behavior from scratch using treats, but the thing where it only worked when I was sitting next to him was a big sticking point⁵. So, following Hannah Branigan’s advice, I tried a bunch of variations of sitting on the ground next to him. I tried facing him, facing away from him, having my knees up, having my legs crossed, having my toddler right next to me trying to get into everything, talking to my kids while I was cuing him, etc. Some of this stuff I did on purpose, like changing my body position, and some of it was more about going with the flow and resisting my “natural”⁶ tendency to try to keep my cue picture as consistent as possible. And indeed, when I let my body position etc. vary, often Argos would stop offering the full version of the behavior (where he was putting the chew between his paws and chewing it “for real”), but, unlike when I stood up, I was still getting something. Usually he would at least pick the chew in his mouth. And from there, it was ~straightforward to reshape a fuller version of the behavior.⁷ Once I did all those variations of sitting next to him, it sometimes worked to get Argos to pick up the chew when I was standing up, and from there, I could gradually shape the full version of the behavior not only the in scenario where I was standing up next to him, but a in bunch of other ones too⁸.

Here’s how I think of what happened in terms of taking stuff as object, and some stuff about why it seems natural for me to frame it like that. I think before my training project, Argos’s deal with the chews was ~that when the stars aligned he would sometimes he would “find himself chewing”. When he was sort of stressed and the chew was pretty nearby, he often would. It didn’t seem to ever occur to him when he was calm, and when he was too upset he had other priorities, like barking at me. His chewing behavior wasn’t very flexible, and he couldn’t slot it into a bunch of situations where it might be helpful. Afaict after I worked through my training project with him, it became much more the sort of thing where he could chew in “various” situations, where “various” was like a new node that got inserted into his menu of when he could try which behaviors. My interpretation is that Argos substantially learned how to take chewing as object once he had a few more ways to relate to it.

I’m partway through a similar process where I’m trying to get Krypto, my younger dog, to be more flexible around balls. He really loves balls. He’s the sort of dog where sometimes I’ve put all the balls in the house away so that he’ll stop repeatedly bugging people to throw them for him. Nowadays he mostly doesn’t do that with our family, but he still does it a bunch with sympathetic guests. He has long had his own repertoire for how to get people to throw them for him including: dropping the balls on people’s laps, pawing them, backing up and staring, and getting more balls to make a pile on people’s laps. I’m not totally sure what it means for my model that he already has 3-5 behaviors, but I think maybe it matters that all these behaviors are pretty directly hill climb-y button mash-y sorts of things that involve intensely orienting towards the balls. Historically, it’s been borderline impossible for me to get him to do behaviors that are normally easy for him but have nothing to do with balls, such as touching my hand with his nose, or sitting, when a ball is anywhere in the picture. Even when I try to reinforce him by throwing the ball after he does something I was looking for, he hasn’t repeated the thing much. But he has a much easier time doing all that stuff when I stuck the ball in my pocket.

One practical limitation to training by throwing balls is that once I throw the ball, I don’t have it anymore. Often Krypto will bring them back and give them to me, and I can do multiple reps by having multiple balls, but even so I find the process frustratingly slow and janky.⁹ It finally occurred to me that me presenting the ball was quite reinforcing on its own, and was probably the critical contingency anyway¹⁰. Indeed, once I started paying attention to what exactly he was doing when I presenting the ball, it became much easier to shape different behaviors, including ones where he briefly oriented away from the ball.¹¹ This training project of mine isn’t finished—maybe I’ll both make progress on it before too long and remember to post an update once I do—but even with the little I’ve done with getting him to do a few different things to get me to present the ball, I’m already getting the, very reinforcing for me, sense when I look at him that his gears are turning in a new way. And my previous felt sense of “I suppose in theory all sorts of things are possible and I trust people who claim to get these sorts of results (in part bc I’ve seen video etc.)” has shifted to “there’s a glimmer of him being able to take balls as object, so even though it would take a bunch of steps to generalize it, I can see the path to him getting it—and having ‘various’ option for how to relate to balls—and the path doesn’t have any magical steps in it”.

My current strategy (when I have the bandwidth and inclination) to help my young toddler when she’s fixated on something that she can’t have (like running into the street) is moving her back so that we aren’t that close to it, blocking her a little, getting her to do an easy thing that she knows how to do and generally enjoys (like high fiving me), and then reinforcing that by unblocking her and letting her get a little closer to the thing. Then repeating this process with her doing a few more different things (e.g. fist bumps, clapping her hands, and taking a few steps backwards, saying some words about the thing). I’ve never once heard anyone else describe this strategy, and it has worked better for me than everything else I’ve tried. (It’s not perfect or anything. I’m sure I could refine the technique. I’m agnostic as to whether perfect exists with that type of thing at her age.)

(A different framing I’ve played around with to describe the shift I have seen with this sort of training process is “object permanence”. A couple of years ago, I was really into trying to process what was up with constructional affection. I watched a bunch of video examples where people started out petting dogs pretty continuously, and then inserted gradual pauses until they were fully leaving the room, closing the door, waiting for about ten seconds, and then coming back in, and it seemed like this process caused some sort of conceptual shift in how the dogs related to people walking by their enclosures that I wanted to describe as the dogs having object permanence around the people coming and going. And the dogs stopped barking and jumping to get the people. Having seen more examples since then, and played around with more training projects since then, subject-object shift now seems to me like a better description of the interesting thing that happens, but I wanted to give at least a brief nod to my past process of trying to get my head around what was happening. Oh, and I think there’s a thing here too about sign tracking vs. goal tracking.)

Why 3-5 things?

I don’t have a ton to say about why 3-5 seems to be the number where then the mind will do something like allocate a new node. Ken Ramirez brought up these numbers too when he came on Hannah Branigan’s podcast to talk about concept training. My guess is that that number is enough to triangulate a thing. It seems intuitive to me why two data points wouldn’t be enough, and why three often would be. But sometimes it would take a couple more.

Some first person experience with trying 3-5 things and subject-object shifts

I mentioned earlier that my old IFS therapist did a brief process that got me to unbend from my belief that I couldn’t be there for other people when I was emotional. Here’s a little more about how I remember it happening. IIRC, first my therapist asked me how I felt towards that belief. I remember not getting any sort of real answer to that question, as is typical in IFS when the person is “blended” with a thing. Instead I said that the belief seemed true to me. As I remember it, my therapist asked me to do some exercise about imagining it as a thing, and imagining physically moving away from it. She also seeded a potential new way for me to relate to this belief of mine by saying that needing to be unemotional to be there for people wasn’t how she thought of it at all, and that the opposite seemed pretty true to her—that she needed to be in touch with her emotions in order to be there for people.

Her suggestions worked, I unblended from the belief, and then I related to it more as a working assumption that I had some evidence for, and some evidence against. I could see it as separate from me. I think it’s pretty accurate to say that I could newly take the belief “I can’t be there for other people when I’m being emotional” as object.¹² (Is “unblending” in the IFS sense the same thing as having a subject-object shift about a part? I think it must be, at least in the way I’m using the term subject-object shift.)

On an intuitive level, I’m a still little confused about whether this story is an example of getting a subject-object shift by relating to my belief in 3-5 different (non-button-mash-y) ways, but I think it ~is. I think the new way to relate to it were roughly:

• as a thing I could feel a way about

• as a thing I could visualize and imagine moving away from

• as I thing I could compare to roughly the opposite belief

I think what I’m saying is a lot like the monad tutorial fallacy, a classic blog post about how people frequently give too much credit to the final insight they have that makes a concept click, when the click couldn’t have happened without the person having understood a bunch of concrete details first.

Joe Haskeller is trying to learn about monads. After struggling to understand them for a week, looking at examples, writing code, reading things other people have written, he finally has an “aha!” moment: everything is suddenly clear, and Joe Understands Monads! What has really happened, of course, is that Joe’s brain has fit all the details together into a higher-level abstraction, a metaphor which Joe can use to get an intuitive grasp of monads; let us suppose that Joe’s metaphor is that Monads are Like Burritos. Here is where Joe badly misinterprets his own thought process: “Of course!” Joe thinks. “It’s all so simple now. The key to understanding monads is that they are Like Burritos. If only I had thought of this before!” The problem, of course, is that if Joe HAD thought of this before, it wouldn’t have helped: the week of struggling through details was a necessary and integral part of forming Joe’s Burrito intuition, not a sad consequence of his failure to hit upon the idea sooner.

But I claim the pedagogical principle is more specific than “struggling through details”—it’s more like: for each new abstraction I need 3-5 ways to relate to it. When abstractions are built on top of each other, as they often are, then of course I may have to work through quite a lot of things.

Just the other day, a friend of mine couldn’t really follow what I was saying and asked me to please say it a totally different way. I did, and then I said it in yet a totally different way again, and it seemed like that worked pretty well.

One of my favorite pieces of parenting (but not just parenting) advice is to, in difficult situations, think of two choices and go with the better one. Two isn’t 3-5, obviously, but two is faster, the advice is meant to be used in realtime, and most difficult situations will come up multiple times, so I think it works out to the same thing.

What’s my point

I don’t have a point, but I wanted to write about this anyway. Or rather, my point is that I care about these ideas in the way where I wanted them on the internet, and I thought if I wrote all this down then I might be able to write some different stuff down afterwards. It’s all more complicated than I’m saying. People are people, and animals are usually people too, and sometimes people don’t want to take things as object. Sometimes they explicitly don’t want to in the way where they have taken the idea of taking things as object as as object, but usually not, IME.

Taking stuff as object isn’t binary either, at least as far as I can tell. Or if it is binary, then the binary thing doesn’t often have a short description length. “The thing where I freeze up and start getting tunnel vision button-mash-y when people do thing x—I can now fully take it as object when I’ve had enough water and gotten decent sleep at least one of the past two nights. Or almost all the time when I’ve explicitly remembered about the possibility of trying within the last 5 minutes. But in that case it’ll take me an extra 10 seconds to get there. It doesn’t work on Tuesdays or in July, and sometimes it works never.”

Please leave a comment if anything I said sparked a goes-on-the-internet-shaped thought in you :-).

1. I’m also perpetually upset that this book has no kindle version. I check every few months. If one ever comes out, please tell me. Yes, I do have a pdf copy of it, and they aren’t hard to find. I’m still mad.

2. I think there are many senses in which I’ve been “really trying”, and I’m sure it’s not all of them. I’ve read a fuckton of books, joined mailing lists and Facebook groups, tried many of the exercises in them, gone to ClickerExpo once in person and once virtual, hired professional dog trainers a little and tried to learn from them, gone on and on about all this to my friends who are willing to listen to me etc.

3. I’m so sorry but I ~never know when to cite people by name when the thing in question isn’t already on the internet. If you are ever the friend in my stories and want attribution, lmk and I’ll do it.

4. Often, with my animals or my young kids, I look at them and think “they could never do that. I can’t even picture them doing that,” even when it wouldn’t be a physically hard action. My goto example of this is that when I look at most 1 year olds, I can’t at all imagine them standing around waiting for a bus at a bus stop. (I can think of one exception. I think she’s truly excellent at standing still for someone her age.)

Sometimes, of course, I’m simply wrong about what they are capable of. But I think I’m usually pretty well calibrated, and when that’s the case I think it’s often because I can tell that the person in question can’t take something load-bearing to doing the action as object.

5. I’m pretty sure the root of the issue was that Argos didn’t want Krypto, our other dog, taking his chew away from him, which is a real thing that happens, especially when I’m not right there. I have also worked with Krypto on leaving Argos alone when he is chewing, but in this case I didn’t have to solve the root cause to get Argos to chew in more situations. Probably especially because at this time in my training project, Argos was in his crate with the door closed, and Krypto couldn’t actually take his chew regardless.

6. By which I mean that I think I’ve been (positively and negatively) reinforced into it, since sometimes small deviations in how I ask my dogs to do stuff will break the behavior.

7. I had previously heard the advice to teach dog things in lots of little shaping steps, even when it might work to capture the entire behavior. I remember highlighting this passage from When Pigs Fly!: Training Success with Impossible Dogs:

I have found that, the smaller the approximations you use to shape a behavior, the more durable they will be. What do I mean by durable? Under stress or in a new context, dogs will often fail to perform behaviors that we have taught them. If you have taught the behavior in one big piece then, under enough stress or distraction, the whole behavior disappears. If you have taught the behavior by reinforcing small approximations of the behavior, the behavior will usually only break down to the last or near to the last approximation that you reinforced.

and the equine clicker training blog says:

Because the behavior is captured in a larger chunk, it’s harder to go back and rebuild the behavior if you lose it. With a shaped behavior, I can always go back to the original shaping process if the animal stops doing the behavior. If the behavior is captured, I don’t have any way to remind the animal by reviewing the original training process.

I think I’ve now heard trainers say “captured behavior falls apart” as an adage, though I searched and failed to find a source for it.

It hasn’t been until pretty recently that I have had training experiences of my own that led me to take this advice to heart.

8. I feel compelled to note that a load-bearing part of process was also following Hannah Branigan’s “easy hard easy” advice. So right after I did a rep of him chewing while I was standing (which was initially hard) I did a rep where I was sitting next to him (which was easy) etc. until the thing that used to be hard stopped being hard.

It also used to trip me up that I would seem to “lose progress” so much between sessions, but once I framed it as “behaviors are easier to do when the antecedent is that someone has done them very recently”, it seemed like part of a normal progression, not a problem to be solved.

9. I think probably the more real issue isn’t so much the slowness as the discreteness. I’m planning to write a whole post about how important I think this is, but tldr; I currently think conjugate reinforcement schedules are fundamental and “real” than episodic ones.

10. Maybe there’s some sort of general principle here about how when people can’t take things as object, something like traction tends to be more functionally reinforcing than actually getting the object-level thing? I think sign tracking and goal tracking are related to this.

11. This approaching of using the ball presentation as the reinforcer was heavily inspired by the training plan Hannah Branigan used to get her border collie to stop being so obsessive about chasing shadows. I’m pretty sure she talks about it in the Q&A section of this video if anyone is interested.

12. Fwiw, I think I still largely have this belief as a working assumption, and I’m still not sure how true it is. My best guess is that it’s medium true; it exists in a misleading ontology, and it’s a little bit false.