What's your view on Anna's concerns here?
my guess in hindsight is that the "internal double crux" technique often led, in practice, to people confusing/overpowering less verbal parts of their mind with more-verbal reasoning, even in cases where the more-verbal reasoning was mistaken. For example, I once used the "internal double crux" technique with a person I will here call "Bob", who had been badly burnt out by his past attempts to do direct work on AI safety. After our internal double crux session, Bob happily reported that he was no longer very worried about this, proceeded to go into direct AI safety work again, and... got badly burnt out by the work a year or so later. I have a number of other stories a bit like this one (though with different people and different topics of internal disagreement) that, as a cluster, lead me to believe that "internal double crux" in practice often worked as a tool for a person to convince themselves of things they had some ulterior motive for wanting to convince themselves of.
This version of IDC has not resulted in people reporting that problem (or at least not to me) in some dozen-or-so instances of teaching it to a room of 30 people.
(What's presented above is actually an older version of IDC; my story is that Anna sort of forked away from it and kept developing and I stubbornly held with the simple straightforward thing that a middle schooler could understand and use.)
FWIW I think there's also something to the thing Anna mentions about IDC being biased towards parts that are more verbal. You can get around this with sufficient skill at focusing and introspection, but my sense is that there is indeed some number of people who use it without that sufficient skill, and therefore end up in a state where they've integrated all of the concerns of verbal part A, and only 30% of the concerns of less-verbal part B. And because they themselves aren't in touch with other channels such as kinesthetic, they think they are 100% integrated and often act from that place
A FB friend writes, in response to this post:
Just my opinion but I trust Internal Family Systems as a technique over IDC in general. Of course both have their downsides and can be taught poorly / wrongly, but the inclusion of the compassionate, curious, open Self in IFS seems critical for this sort of process to go in a positive direction, rather than an arbitrary or distorted one.
... which seems to be pointing in a similar direction. There's some convergent evolution here, in that I found that teaching "Understanding Shoulds" (which gives participants a sense of fondness and receptivity to all of their impulses and urges, even the ones they're angry at or feel are counterproductive) was a strong prereq for IDC.
If you set out to beat one of your parts into submission, or to confuse and manipulate it, you're probably gonna (locally) succeed. The solution to that problem isn't "don't use one of these techniques," because if you have the predecided mindset to begin with, it'll corrupt anything that you try to do.
This is already mentioned a bit elsewhere, but it seems perhaps worth flagging for potential readers that this technique (or perhaps, as you mentioned, a similar but different offshoot of it) is now considered... impure? flawed? risky? and is no longer part of the CFAR mainline workshop curriculum as I understand it -- I would say it's one of the more notable changes recently, though who's to say where things will evolve in the future...
Absent any kind of detail on the claims of what CFAR is now doing and why, I can't evaluate this. Not a fan of the passive voice "is considered."
All I can say is: I've taught IDC to well over a hundred individuals at this point, over multiple years in multiple contexts, and seen it be highly positively impactful for over a dozen people, and also heard zero negative reports back.
¯\_(ツ)_/¯
A couple of years ago we developed something in this direction in Epistea that seems generally better and a little less confused, called "Internal Communication Framework".
I won't describe the whole technique here, but the generative idea mostly is "drop priors used in IFS or IDC, and instead lean into a metaphor of facilitating a conversation between parts from a position of kindness and open curiosity". (Second part is getting more theoretical clarity on the whole-parts relation)
From the perspective of ICF, what seems suboptimal with the IDC algorithm described here is roughly this:
1. It doesn't do the unblending that well (IFS is better in this); this protocol seems to mostly asks whoever is the ruling party to be fair.
2. The "doublecrux-like" form nudges/pushes parts to form 2 coalitions, forming the sides of the doublecrux. This is not a natural form for many internal disagreements.
(3. Similarly, IFS errs on the side of having strong priors on the form/structure of the parts: exiles, firefighters and managers, and on the side of doing "family therapy" with them.)
4. Both IFS and IDC seem somewhat confused/vague about the vertical relations and run e.g. the risk of empowering the parts layer at the cost of the 'whole'.
From what I've seen in practice, all of these problems are often overcome by skilled facilitators who do something closer in spirit to ICF and mostly use IDC or IFS as a name for it, without really trying to follow the protocol closely.
"Instead lean into a metaphor of facilitating a conversation between parts from a position of kindness and open curiosity" is ... straightforwardly what the essay above recommends?
Similarly, IFS errs on the side of having strong priors on the form/structure of the parts: exiles, firefighters and managers, and on the side of doing "family therapy" with them.
For what it's worth, my experience is that while the written materials for IFS make a somewhat big deal out of the exile/firefighter/manager thing, people trained in IFS don't give it that much attention when actually doing it. In practice, the categories aren't very rigid and a part can take on properties of several different categories. I'd assume that most experienced facilitators would recognize this and just focus on understanding each part "as an individual", without worrying about which category it might happen to fit in.
I guess this is what you already say in the last paragraph, but I'd characterize it more as "the formal protocols are the simplified training wheels version of the real thing" than as "skilled facilitators stop doing the real thing described in the protocols". (Also not all IFS protocols even make all those distinctions, e.g. "the 6 Fs" only talks about a protector that has some fear.)
I think it's worthwhile to note that some things, exercise and the tackling the hard problems of Alignment are a great example, simply have Ugh Fields. The only solution to an Ugh Field is to use positive reinforcement to wade or dive into a task that you have an aversion to, in order to overcome the Pavlovian conditioning that made you associate the averred task with pain/regret/failure. Otherwise, you'll bounce off the problem in ludicrously complicated ways, often resulting in reinforcing the act of bouncing off the problem.
For exercising, for example, the solution is to wade back into it, but there's tons of tasks that are more daunting than painful, and thus the solution is to dive if positive reinforcement is probable.
Summary
Curious if this is more of a "do this when currently experiencing the conflict" or more of a "you can wait until some more convenient time"? (Ignoring that in practice the latter might often be "put off indefinitely".)
Like, do you recommend pausing netflix and delaying the run to do this? (Might be easier than pausing netflix to go for a run.) Or do you think most people will be able to access the relevant parts of themselves some time later?
For the first couple of times, you should set aside 30-90 minutes and do all the steps in full, on paper.
After that, you can do much faster five-minute or five-second versions.
I think you can usually manage to boot up the conflict even when you're not in the moment; most (i.e. literally more than half) participants at CFAR workshops had nonzero success doing IDC just ... right then, when we taught it and told them to.
Author's note: While most CFAR workshops taught double crux first and then internal double crux afterward, it's long been my opinion that both concepts would be better served in the other order. Recent experimentation has shown results consistent with this hypothesis; practicing collaborative truth-seeking inside one's own head helps one build the necessary mental muscles before moving on to doing it with a whole other separate human. [External] Double crux will be the next post in the sequence, followed by a loop back around to some of the opening session advice.
Epistemic status: Preliminary/tentative
Internal double crux (formerly propagating urges) is a technique-in-progress, with the goal of finding motivation through truth-seeking rather than through coercion or self-deception. It is currently in flux and has no formal research backing, but it follows logically from a handful of other threads about which CFAR is relatively confident (such as microhedonics, hyperbolic discounting, cognitive behavioral therapy, and useful-even-if-wrong theories like internal family systems or society of mind).
If two different people have access to the same information, and their models of the world cause them to make two different predictions, then we can confidently say that at least one of them is incorrect. We may not always be able to tell which is which, but we can be sure that one-if-not-both of them has a chance to update toward the truth.
Similarly, if a single person has simultaneously contradictory beliefs or desires, then at least one of the models behind those beliefs is wrong, miscalibrated, or incomplete.
(And usually both.)
If you both “want to get good at running” and also never want to get up off the couch and put on your running shoes, then one part of your belief set—one of your causal models of the universe—has concluded that running will help achieve your goals, and another has concluded that it doesn’t, and both of these can’t be true.
Internal double crux is a technique that seeks to resolve this conflict by helping each of these models to incorporate the information that the other has to offer. If you were to conceive of yourself as being made up of sub-agents, each of whom focuses on a different subset of your goals and has a different perspective on how the world works, then the goal is to cause those sub-agents to enter into a productive double crux conversation and correct their tunnel vision.
In particular, the hope is to have both sides of your internal disagreement update toward truth. The conflict is a result of some kind of confusion, and thus reducing confusion will also tend to reduce conflict.
CFAR's experience running internal double crux with participants is that the end result of such a process is a state of feeling (more) intrinsically motivated and internally unconflicted; of reducing one's need for duty or diligence or force-of-will and instead having one's urges aligned with one's actual goals. It's an early version of a technique for turning wanting to want into straightforward wanting.
Understanding “shoulds”
Part of the problem that internal double crux seeks to correct is our natural tendency to arbitrarily support some sub-agents or subgoals while suppressing others. Many people find it easy, for instance, to attach words like “motivated” or “goal-oriented” or “good” to the part of themselves that wants to go running or finish the project, while attaching words like “stupid” or “lazy” or “undisciplined” to the part that wants to stay on the couch.
This is an effective shortcut for some people, but it comes at a cost—you’re ignoring signals from part of your belief set, and expending energy on internal conflict and executive overrides that could otherwise be allotted to the things you actually want to do. Instead of containing, suppressing, or drowning out your conflicting urges, IDC encourages you to update and integrate them, or at the very least to give them an actual, impartial hearing before deciding that they’re inappropriate.
To return to the running example: you may have a belief that it’s good to exercise, and furthermore that running is the best and most efficient way to exercise, and furthermore that doing so is a better way to spend your afternoon than, say, Netflix.
If you happen to be watching Netflix at the time, this belief is likely to ruin your fun. Perhaps you get up, put on your shoes, and begin to run—yet as soon as you do, you find yourself longing to stop, and continue only with effort and some minor degree of suffering.
Rather than summarizing this situation as “I’m just lazy” or “I struggle to stay motivated,” it’s instead productive to think “in addition to my belief that it’s good to run, I apparently also have a belief that it’s good to watch Netflix.” This isn’t just a cute, permissive reframe; it’s what’s actually going on. Some part of you believes that Netflix is exactly the Thing To Be Doing.
And this part of you believes this for some causal reason. Beliefs don't come from nowhere; they're essentially always a response to some kind of past experience. The part of you that is generating pressure-toward-Netflix is doing so because it thinks that staying on the couch will make for a better life, and bring you closer to your goals. It's not lazy or stupid, it's tunnel visioned, failing to take into account things like long-term health, or the value of following through on your self-commitments.
(Just as the part of you that's clamoring to get off the couch also has tunnel vision, and is discounting the value of relaxation or hedonism.)
At CFAR, we often characterize these internal disagreements as “shoulds.” Given any default action, a should is an urge or a pressure to do something else instead:
If, on the other hand, you went ahead and grumped at your coworker, you might feel that you should have stuck to your communication goals; if you buckled down for a writing sprint, you might feel that you should have taken time out to spend with your significant other, or conserved resources for work the next day; if you start taking classes, you might feel that you should have saved the money, or spent it on something else instead.
Many people default to one side or the other when they notice a should—they have a deontological policy of defending their inner emotional selves, or of conforming to social expectations, or of sticking to the plan, or of being flexible and changing the plan. The problem is, any one-size-fits-all solution is going to miss a large percentage of the time, and writing the bottom line without actually considering the arguments is a recipe for inaccurate beliefs.
At their core, shoulds are data, and data is something an aspiring rationalist almost always wants more of. Just as regular double crux encourages us to remain open to the idea that others might have better information than we do, so too does internal double crux encourage us to listen to the input of every aspect of our motivational structure. Different parts of your psyche are better equipped to pay attention to different swaths of the available evidence, and they process that evidence in different ways. Given the complexity of the world, it makes sense to start from the assumption that a synthesis of conclusions will be more accurate than any one conclusion on its own.
The agent at the top mistakenly believes that the correct move is to head to the left, since that seems to be the most direct path toward the goal. The agent on the right can see that this is a mistake, but it would never have been able to navigate to that particular node of the maze on its own.
The part of you that wants to run is good at paying attention to your long-term goals, your social standing, your health, and your sense of yourself as a strong and capable person. The part of you that wants to watch Netflix is good at paying attention to your short-term urges, your energy levels, your sense of comfort, and whether or not the new Stranger Things episode seems likely to be good.
You can ignore one side or the other indefinitely, but the result is often feeling halfhearted or torn, ruminating or struggling with decisions, burning willpower, suffering from your decisions, and endorsing one part of your psyche beating up on another part. In order to build a maximally detailed understanding of the world and correctly strategize across all of your needs and goals, you’ve got to bring all of your models to the table—implicit, explicit, S1, S2, endorsed, embarrassing, vague, and exact.
An Example IDC
Moreso than with most techniques, we have found that participants learning IDC get substantial value out of holding themselves to a very specific format for at least their first couple of attempts. There's some amount of magic that often requires experience and is less transmissible up front, so simply giving it a shot as-written (as opposed to making your own tweaks and adjustments on the fly) is something we recommend more strongly than usual.
Step 0: Find an internal disagreement.
Look for any sort of "should" that's counter to your current default action—something you feel you aren't supposed to think or believe (though on some level you do), or a step toward your goal that feels useless or excessively unpleasant.
Step 1: Take a piece of paper, and, at the top, draw two dots, representing the two perspectives/viewpoints/sub-agents. Name them.
Step 1(b): Check the names for resonance and fairness.
What you are doing during the internal double crux technique is, essentially, moderating a debate between two different parts of yourself/two different perspectives you're capable of adopting.
It's often useful to visualize these two different perspectives as something like distressed, angry kindergarteners, each of which is focused mainly on its own priorities and doing whatever it takes to get what it wants.
You, the moderator, want to make sure that you aren't partisan in this disagreement. Remember, you are taking as given that each side is in possession of some non-negligible pieces of the truth—you want to convince both sides to bring their map-fragments to the table, so that you can incorporate all of their information into your larger world model. That won't happen if you're secretly allied with one side, and helping it beat up on the other.
It's fine to have more sympathy for one side than the other, to be clear. And indeed, if this is the case, you definitely want to notice this fact!
But when you choose to play the game of internal double crux, you want to correct for that default sympathy, and make sure to offer additional, compensatory support to your inner underdog.
Looking at the names above, it's clear that the moderator is "on Team Go Running." They've given the alternative the epithet "laze around" rather than a phrase that viewpoint might have chosen to describe itself:
Step 2: Decide who speaks first.
Which side, if either, feels more urgency? Which side is clamoring more loudly to be heard? If you have no clear sense, feel free to just flip a coin.
Step 2(b): Embody that perspective, and, from that perspective, say one thing.
This is where moderation comes into play. Often, the kindergarteners will want to unleash a flood of words, and it's your job to ease that flow into something productive and comprehensible.
Step into the mindset of one of the sides. Get in touch with what that side wants. Feel into what it's like to hold [that value], and speak from that point of view.
The usual prompt here is "what's one important thing that the other side doesn't understand?" One crucial piece missing from their model of the world; one consideration that perspective is failing to take into account, or failing to weight properly.
It's important that you try to actually embody the viewpoint, rather than doing something more aloof or insulated like imagining what it "would" say. It isn't a performance piece—the idea is to connect with what-it's-like-to-be-the-version-of-yourself-who-really-wants-to-go-running (or whatever), and try to produce authentic sentences from that perspective. The skills you sharpened in the previous section on Focusing will come in handy for making sure that the words you write down actually resonate with that side.
Step 3: Get the other side to acknowledge truth.
The overall aim of the exercise is to cause each perspective to absorb the truth/wisdom/ experience of the other. In order for that to happen, you-the-moderator will encourage each side to start off its turn by first finding some grain of truth in what the other side just said:
... it doesn't have to be complete, and sometimes it won't even be something the other side directly said, so much as a logical consequence or an underlying assumption. It just needs to be something that draws the two sides a tiny bit closer, however begrudgingly.
Step 4: That side gets to add its own "one thing."
Having acknowledged that the other side has some kind of point somewhere, this side now gets to lodge its own objection.
... in this case, that it's been a really rough week, and so perhaps it makes sense that we've skipped running three times in a row.
Notice that this whole point is not necessarily spelled out—you don't have to force yourself to speak in full, coherent, justifiable sentences. Sometimes one side might not even speak in words at all—might draw a picture, or leave a scribble, or just write AAAAAAAAAA.
It's important to allow these things to happen rather than to impose order from above. You-as-moderator are there to nudge the conversation back on track, as necessary; you don't have to pre-restrict the dialogue to things you've already thought of, or sentences that already pass some kind of filter.
Step 5: Repeat.
Back and forth, each side should a) acknowledge truth contained within the previous entry, and b) add one new bit of information from its own perspective.
Again, sometimes things do not go according to plan:
... in this case, the discussion started going too quickly. Orange's acknowledgement was perfunctory, and blue didn't acknowledge at all. Noticing this, you-as-moderator might pause, and look back at the previous orange statement, and see if you can nudge your blue side to make some form of genuine acknowledgement before proceeding.
It's fine for off-script or against-the-rules things to happen, as long as you bring the conversation back toward productive discourse.
... here, for instance, the blue side had something of a minor meltdown.
But that meltdown did get written on the page. It's not the job of the moderator to pre-censor, but rather to correct after-the-fact. The moderator gives the blue side a chance to blow up and blow off some steam—to register the magnitude of its disagreement, rather than forcing down its reaction—and then gently requires that it nevertheless find some grain of truth in orange's previous point.
In this case, once it was orange's turn, orange broke the rules a couple of times—first by skipping straight ahead to objection, and then by responding with several points instead of one single point.
(Some CFAR participants have benefitted from actually writing down moderator interjections, sometimes in another color. For instance, after "Years are made of—" you might write down "Wait—can we do acknowledgement first?" and after "gained fifteen pounds" you might write "One thing at a time.")
What usually happens over the course of this back-and-forth is that the problem reveals itself to be some other problem, usually one that's a layer deeper and more interesting. It's like the couples' therapy truism that "it's never really about the dishes." The question of "should I go running or keep watching Netflix?" is a stand-in for, or an instantiation of, a more complicated dynamic.
Once you realize that—once the underlying disagreement makes itself known—it often helps to draw out a new sheet of paper and draw two new dots with two new names:
That in itself often has tremendous value. Getting a clearer understanding of the deeper generators of various dissatisfactions and internal conflict gives you much better odds of actually solving them (as opposed to flailing around in the dark, patching symptom after symptom).
Why so many little rules?
In part because most other ways of resolving internal disagreement seem, to us, to fail to strike at the root of the problem.
Continuing with the metaphor of the kindergarteners—an adult can fairly easily force two kindergarteners to stop fighting. You can separate them, and admonish them, and demand that they behave civilly toward each other, and cow them into compliance.
But until you cause them to actually be cool with one another, any cease-fire is going to be fragile, and dependent on the continued presence of an enforcing authority. As soon as the grownup is out of the room for long enough, they'll be right back at each other's throats.
Correspondingly, if one part of your value system is having to repeatedly brute-force overwhelm another part, any cease-fire based on your active attention or conscious consideration is going to be patchy at best. Even if you continue successfully engaging in the "right" behavior every time, you'll still be burning energy and willpower on costly self-control.
By following a process that causes your inner models to actually understand each other, you imbue each with some fraction of the wisdom and virtue of the other, leading them to be fundamentally less-in-conflict and better able to support each other (and the higher strategic you).
As always, you should in fact tinker with and iterate on this technique, or abandon it entirely if you find some other method to achieve the goal. But the rigid, rules-based approach has been surprisingly useful to a surprisingly large fraction of participants, so we do honestly recommend actually trying it before moving on to your own personal IDC'.
The IDC algorithm
0. Find an internal disagreement.
1. Find a charitable handle for each side.
2. Embody one perspective and, from that perspective, write down one thing that the other perspective is failing to properly take into account.
3. Embody the other perspective and, from that perspective, write down an acknowledgement of one grain of truth in what the previous side had to say.
4. Still embodying the second perspective, offer back one counterpoint for the first side to consider.
5. Repeat steps 3 and 4 until the disagreement dissolves (or transforms). If useful, start over with step 1 with new names.
IDC—Further Resources
Psychologists Carver and Scheier (2002) use the theory of control systems to model goal pursuit, where feedback about one’s progress towards a goal is translated into pleasant or unpleasant feelings. These feelings then mo- tivate the person to continue an effective approach or change an ineffective approach. In order for the system to function smoothly, it is necessary for the relevant part of the system to recognize the connection between the goal and one’s current behavior.
Carver, C. S., & Scheier, M. F. (2002). Control processes and self-organization as complementary principles underlying behavior. Personality and Social Psychology Review, 6, 304-315. http://goo.gl/U5WJY
People tend to be more open to information inconsistent with their existing beliefs when they are in a frame of mind where it seems like a success to be able to think objectively and update on evidence, rather than a frame of mind where it is a success to be a strong defender of one’s existing stance.
Cohen, G.L., Sherman, D.K., Bastardi, A., McGoey, M., Hsu, A., & Ross, L. (2007). Bridging the partisan divide: Self-affirmation reduces ideological closed-mindedness and inflexibility in negotiation. Journal of Personality and Social Psychology, 93, 415-430. http://goo.gl/ibpGf
“Focusing” is a practice of introspection systematized by psychotherapist Eu- gene Gendlin which seeks to build a pathway of communication and feedback between a person’s “felt sense” of what is going on (an internal awareness which is often difficult to articulate) and their verbal explanations. It can be understood as a method of querying one’s inner simulator (and related parts of System 1). Gendlin’s (1982) book Focusing provides a guide to this technique, which can be used either individually or with others (in therapy or other debugging conversations).
Gendlin, E. (1982). Focusing. Second edition, Bantam Books. http://en.wikipedia.org/wiki/Focusing
“IFS,” or Internal Family Systems is a form of psychotherapy developed by Richard C. Schwartz in which the mind is conceptualized as a set of parts or subpersonalities, each with its own perspectives, interests, memories, and viewpoint, and each with positive intent for the overall person. IFS uses family systems theory (a separate branch of therapy) in a metaphorical way to understand how those subpersonalities are organized and how they interact with one another.
Schwartz, R. (1997). Internal Family Systems Therapy. Guilford Publications. https://en.wikipedia.org/wiki/Internal_Family_Systems_Model