Q&A on the Internal Family Systems (IFS) therapy modelhhh

Recently my friend Galen dm'ed me the following questions about IFS, and it thought I'd share my response as a blog post so it can be useful to more people:

1. Is it risky/harmful to think of your self as being composed of parts that are in some amount of conflict with one another? (If so, in what way is it harmful? How risky is it?)
   • Imagine, here, that
     1. I'm broadly skeptical of the literal accuracy/truth of descriptions of psychology such as "human agents are composed of parts with their own needs/wants/preferences", but
     2. I also mostly believe people when they say that, by acting/thinking as if the self is 'made of parts', they've been able to resolve persistent issues, &
     3. The people making the "nah dude, it's useful" claims are similar enough to me that their experience should inform my expectations w/r/t myself.
   • My main concern: when I use wrong-but-useful 'parts' models, I may be
     1. Reifying a faulty ontology (in a way that'll be hard to undo),
     2. Naturalising/normalising the idea that "incoherence" is the default state of my mind/self, &
     3. Causing myself to have more internally conflicting "parts" than I would have had otherwise.
   • In other words:
     • If you set out to find a Thing within yourself, maybe you will find the Thing.
     • If you find the Thing, how likely is it that you 'generated' that Thing within yourself only because you were searching for it?
     • If the Thing didn't exist before the search, and the Thing also has downsides/harms that you expect will require additional work to ameliorate, should you not do the search?
2. In your experience, how complex & rich & person-like are 'parts'?
   • To what extent would you expect such-and-such a part to reappear across multiple issues?
   • To what extent would you feel that it'd be natural to name that part, or associate 'unrelated' preferences to it?
     • (eg "Part-P feels threatened in X situation, and also likes baked potatoes" vs "Part-Q dislikes all dishes containing starchy tubers, baked or otherwise")
3. To what extent has your own experience involved (roughly) "building coordination mechanisms for" or "simulating a negotiation/adjudication between" identified parts vs something much darker and stranger and more alien?
   • The default "parts work"/IFS model, as I understand it, seems to assume that one should be aiming to resolve issues by (a) identifying parts that have a stake in the issue, (b) giving a voice to the perspectives & needs of each part, and then (c) finding a way to satisfy their sometimes-competing needs.
   • It seems like some alternative approaches would start with a premise that's more, idk, Game-of-Thrones-ish or something? As in:
     • Yes, I am composed of parts (each with their own perspectives & needs); &
     • Yes, those parts often use my emotions & reflexive/unconscious behaviours in an attempt to fulfil their needs; but,
     • That should be thought of as more analogous to 'scheming'/'palace intrigue'-ish behaviour from a member of my Inner Court (rather than 'a suboptimal attempt to alter deliberation by a Member of my Inner Parliament'), and so
     • "I" should adopt whatever methods will most effectively disincentivise scheming behaviour amongst my many palace advisors/courtesans/etc ... which may include "self-coercion"-adjacent things that are analogous to "randomly executing people" and "choosing only a handful of trusted advisors, each with their own distinct capacities and networks of cutthroat agents & informants" etc.
4. Depending on your answers to (1) & (2), in particular: to what extent do you think there is (or can be) a sharp distinction between "identifying parts and resolving apparent disagreements/coordination-failures amongst them" and "creating/simulating novel Other Minds/Agents and using the outputs of simulated dialogue with those agents to inform behaviour"?
   • Or: in practice, how close is this to inventing a pantheon of private gods and then devoting yourself to them?
5. What have you found to be the easiest ways to do concrete+useful Parts Work without relying on therapy-ish conversation with another human?
   • Also: what do you think are the prerequisites for such methods?
6. Are you at all sympathetic to a perspective/argument-against "doing Parts Work" (and "doing 'therapy' in general") which takes the form: "it is unhealthy to shine a light on all parts of one's self"? (or: "the thing that is better, at least for some kinds of people (& plausibly me), is to instead retain some secrets from one's self; to learn to manage the side-effects of illegible-to-yourself internal 'conflicts'/'competing perspectives & needs'; to focus, instead, on external goals; to adjust your own contexts & systems to account for anticipated paradoxical behaviour from one's self, but to decline to ever attempt to directly address or resolve whatever is 'causing' it")

First off I'll state what I think is literally true about the mind that's most relevant to IFS. I'm largely onboard with the model Kaj Sotala outlines in his Multi-Agent Models of the Mind Sequence, which if you drop the focus on the "multi-agent" part, is basically a mashup of Global Workspace Theory and Predictive Processing. Basically it's a models the minds architecture as naturally shaped for parallel/distributed computation, with key places (the global workspace) where distributed computations get reconciled into cohesive results that are then available to all sub-processes. Two specific posts I'd recommend are Subagents, neural Turing machines, thought selection, and blindspots and System 2 as working-memory augmented System 1 reasoning.

I think that on top of this architecture you can implement both very coherent agent like things and also things that kinda seem more multi-agenty. The most extreme multi-agent-like end state that seems plausible to me is something like Julian Jaynes's bicameral mind "ancient people used experienced their right -hemisphere 'thoughts' as the Gods speaking to them and issuing commands" type stuff.

To the point about "does this reify incoherence as the default?", it a very specific sense I think "incoherence that naturally gets cohered in the course of things" is the default, though there's defs versions of the "incoherence is default!" claim that I think are stupid and obviously wrong. The type of default incoherence I expect is nothing like "there's good inside you and evil inside you and they fight", but the stuff you can observe in babies as they're trying to figure out how to move their limbs at all. There's "disagreement" between different sub-processes making bids for different motor commands that cause conflict and trouble. These processes aren't agent like at all, they just haven't learned to syncronize yet becaue you're 2 months old. You get a shit ton more coherent by age 2, a shit ton more coherent by age 5, etc etc.

So I think the mind starts in relative incoherence and gradually coheres itself. But I think the process of cohereing is pretty natural, and it's definitely not a "default state of humans" to get stuck in incoherence. I think anyone claiming that is full of shit. It's common for sure, but it requires additional energy or force to gum up the natural process of integrating and cohering. I think when therapy inclined folk talk about "integrating", it almost always refers to "integrating aspects of me that are stuck in a double bind, or some Knot of some kind".

K, so that's a lot of what I think is literally true. Back to the framework. The main way that I think IFS isn't true is that the miscellaneous subprocesses in that are stuck in some recurring conflict aren't necessarily that agent-like. To your question about richness and continuity, when I've done parts work in the past I had very little continuity between parts. Like, there'd be recognizable things that would keep popping up, like "of there's a sense of adventure here that wants something" but it would flow into other related beliefs and desires and aesthetic bundles with no clear boundaries, the "boundaries" were only ever as determined by some ad-hoc internal conflict I'd notice. There was defs a lot of richness, but again it doesn't feel fair to say the "parts" where rich because they didn't seem to "own" the details.

In some sense for me doing parts work was just paying attention to internal conflicts and making each side in a given conflict a character and trying to talk to them. I never learned that much about the whole IFS frame of protector, firefighter, exile, etc.

Oh, aside on your comment on "is this another way to procedurally generate conflicts for yourself?" which I presume is similar to Tyler's more general point here. Basically to the degree that's a concern for anyone, I don't think IFS is more or less prone to that. If you were gonna have this problem, you'd have it if you dropped all talk of parts and sub-agents and just had the directive "notice internal conflict and try and interact with it". Though on a related point, I think you can totally train your mind to be more sub-agent like. Tulpas seem to be a thing, etc. Taking the Game of Thrones IFS model is something I'd expect to artificially inflate conflict, because to feel like you're doing Game of Thrones you'd probs need all the parts to more more agent like than they might be on a given day (in order for the plot to be riveting and worthy of being called Game of Thrones).

Okay so if I think "parts" are only kinda actually "parts", what does the frame afford, separate from being compatible with the literal "parallel architecture" thing I mentioned earlier?

There seem to be really two things that IFS is likely to give someone. First is that the cues and instructions for what to do upon encountering internal conflict involve listening. You're acting as if there is some truth somewhere you are asking to reveal itself. With any kind of introspection, you're hoping that you're finding something real as opposed to getting on the spot confabulation or prior confabulations that have been cached and indexed into your self-concept. I think a lot of people model confabulation as like.... there's a "place" in your mind you go to for "beliefs and motives", and in that spot you find things, which can be real, or can be confubalted forgeries, and the trick of veritical introspection is distinguishing forgeries from the real deal. But I think it's more like there's a different "place" you go to in order to confabulate. It actually feels like a different move, looking in a different place, etc etc, and you can practive reaching for the confabulation generator or reaching for something else.

All that to say, cues that prompt you to do something that's more like "listening" seem like they could help someone not reach for confabulation. Defs not infallably, but if you're new to this and don't have a feel for the difference yet, this might be the thing that gets you to notice it. This is also a benefit that I think doesn't actually need the parts frame, you could easily generate some introspection instructions that cue you towards listening for what is, that don't rely on sub-agent like notions.

The second main benefit I think people get from IFS, and this is what I got from it, is that it taps in intuitions of other-directed fairness to compensate for lack of skill at being fair with yourself. For me, for a lot of my life I thought of discipline as one of my superpowers that gave me an edge on everyone around me. And the way it was common for people around me to talk about not being in control of themselves felt honestly kinda disgusting, so I kinda of just decided to Be Decisive and Commit To Things. Up till university or so my general life competancy was way greater than all the problems I encountered, so this largely just worked out for me, but it meant I was practicing ignoring inner conflict and it felt like "inner conflict can never really be deep, and I can always just have my self-concept Cut The Gordian Knot and it works fine", and when I started facing challenges that required more than a minute of deliberation, I ended up steam-rolling a bunch of shit.

When I tried IFS like stuff (first thing I did was Internal Double Crux at a CFAR workshop), I felt able to spend time working thru the inner conflict and eliciting much more rich information from parts of me that were noticing conflicting things, largely because I had strong intuitions about that being obviously what you need to do when working out disagreements between people. "Obviously I can't just unilaterlayy force a group to go along with something if there's real dissent" was something I was versed in, but not for myself, so treating parts of myself as "others" helped redirect some useful intuitions. But also, as I generally chilled out, had more iterations of "oh, useful information hiding in some internal conflict that took a bit of time and care to surface, that was great" parts framing became less salient/compelling. There's also a way it helped defuse certain shame knots. Things that I wasn't up to the task of going with "i have some desire I'm scared to have, what do I do?" became "part A wants this thing, and while I don't identify with that at all, you can't actually murder parts and so it's in here with me so I got to figure out how to work with it".

That second benefit does seem like you really need a parts like frame to get it. It's also easy for me to imagine someone never dissolving the higher order shame that makes them only able to interact with certain things as "othered but linked-fates" parts. I don't think I have any general stance on using or not using "crutches" like this, though I can defs imagine encountering different people at different points in their life who I might say either "yeah, seems totally chill you're putting that off" or "wtf dude you've been stuck in this pit for years and this is obviously what it centers around".

The last thing I want to respond to is your thoughts on the non-IFS specific, "maybe it's not a good idea to shine light on all parts of yourself".

I feel like I'm quite qualified to comment on this one specifically because I've both had tremendous gains from having more and more of myself "enter the light", and I also at one point, thru aggressive pusruit of some naive variants of "shine light on everything", gave myself an existential breakdown that felt like it snapped something in my brain and permenantly destroyed my ability to enjoy anything in life (it wasn't permenant, 3 days accute, 4 months low-key). So I'm quite aware of the risks and rewards.

(I've tweeted about that a bit here and here. For me the damage was less "toppled load-bearing lie that led to painful life destabilization" and more that naive and angry attempts to purge myself of everything that "wasn't me" led me to conclude I wasn't real and to emotionally attack myself whenever I enjoyed something till I stopped enjoying things.)

Two things I think, one is that it's always possible for it to be contingently the wrong moment to try and unravel some knot. Some people handle scheduling/timing/prioritization by treating themselves as "bad" for not having done something, and that's a way you could try and manage timing, but you could also just get gud and figure out other ways to relate to things that are hard to plan. The second thing is that there really truly aren't any "bodhisattva lies", no knots or internal conflicts that categorically can never / should never be attempted to unravel because the lie is actually perfectly calculated to actually produce the optimal results for you, and/or it hides such a devestating truth that there's categorically no way you'll ever be ready for it.

I'm with Eliezer on the claim that in order to actually be confident that a given self-deception/knot/inner-conflict is robustly beneficial to you, you'd have to have a richer detailed understanding of the situation that you're actually capable of getting when you start ignoring/hiding/fudging internal accounting. Now, that doesn't mean a given knot can be locally load-bearing such that your life will get quite destabilized if you tried to unravel it at that time. And destabilization isn't categorically bad but it's often painful and if shit's destabilized enough you can die. It depends a lot on how much slack you have.

So I defs don't support "anytime you notice internal shenanigans you to drop things and attend to it" nor do I support "XYZ knot is something I'm commited to never engaging with". Some examples of in betweens I do. I'm mostly fine with ignoring strange knots when in various social situations, from a place of "yep, I simply do not have the spaciousness to try and interact with this at all with other people, so I will simply be tightened up for the course of this interaction and if need be I can unpack it later when i'm chilling at home".

So in terms of actual advice for anyone, I guess I'd recommend like progressive poking, and the more you notice stuff that's screaming HEEEY WE'RE NOT OPENING THIS the more I'd devote some resources to increasing the slack in my life. Do I have enough slack to have 15m in the morning to journal about this? What about to get off all digital comms and be alone for a whole day? Do I have any social settings or any friends that feel like I could bring this up to? Generally the more intense knots will be the ones most related to "ways I understand myself to be committed to some life-path that I'm understanding a lot of downsides to" and feeling like you don't have other options. And you're that much more likely to actually never have other options if you never scheme about it, so it seems worthwhile to bit by bit make space to integrate the info about downsides and figure out what to do.

idk if that got at everything, but that's what I got for now.