I think there’s a preformal / formal / post-formal thing going on with Double Crux.
My impression is the CFAR folk who created the doublecrux framework see it less as a formal process you should stick to, and more as a general set of guiding principles. The formal process is mostly there to keep you oriented in the right direction.
But I see people (sometimes me) trying to use it as a rough set of guiding principles, and then easily slipping back into all the usual failure modes of not understanding each other, or not really taking seriously the possibility that they might be the wrong one.
Right now in some contexts I’ve come across as a bit anal about sticking to “Formal Doublecrux rules”. Model share. Check for cruxes. Recurse until you find common cruxes. Look for experiments you can actually run to gain new evidence. Aim to converge on truth.
And it does clearly seem that these steps aren’t always the best approach for a given conversation. But I often perceive what feel like basic errors, which would have been caught if you were following the formal rules.
So I’m currently, like, on a crusade to make sure the people around me that I end up in the most disagreements with are able to nail the Formal Doublecrux Framework, and once we’re all roughly on that page I’ll trust us to do a post-formal version of it where we trust each other to get the basics right, so we can relax about the rules.
I believe I’m one of the people who commented on your strong focus on using the Double Framework recently, but on reflection I think can clarify my thoughts. I think generally there’s a lot to be said for sticking to the framework as explicitly formulated until you learn how to do the thing reliably and there’s a big failure mode of thinking you can skip to the post-formal stage. I think you’re right to push on this.
The complication is that I think the Double-Crux framework is still nascent (at least in common knowledge; I believe Eli has advanced models and instincts, but those are hard to communicate and absorb), which means I see us being in a phase of “figuring out how to do Double-Crux right” where the details of the framework are fuzzy and you might be missing pieces, parts of the algorithm, etc.
The danger is then that if you’re too rigid in sticking to your current conception of what the formal framework of Double-Crux, you might lack the flexibility to see where you’re theory is failing in practice, and you need to update what you think Double-Crux even should be.
I perceive something a shift (could be wrong here) where after some conversations you started paying more attention to the necessity of model-sharing as a component of Double-Crux as maybe a preliminary stage to find cruxes, and this wasn’t emphasized before. That’s the kind of flexibility I think is need to realize when the current formalization is insufficient and deviation from it is warranted as part of the experimentation/discovery/development/learning/testing/etc.
I think there’s a preformal / formal / post-formal thing going on with Double Crux.
My impression is the CFAR folk who created the doublecrux framework see it less as a formal process you should stick to, and more as a general set of guiding principles. The formal process is mostly there to keep you oriented in the right direction.
But I see people (sometimes me) trying to use it as a rough set of guiding principles, and then easily slipping back into all the usual failure modes of not understanding each other, or not really taking seriously the possibility that they might be the wrong one.
Right now in some contexts I’ve come across as a bit anal about sticking to “Formal Doublecrux rules”. Model share. Check for cruxes. Recurse until you find common cruxes. Look for experiments you can actually run to gain new evidence. Aim to converge on truth.
And it does clearly seem that these steps aren’t always the best approach for a given conversation. But I often perceive what feel like basic errors, which would have been caught if you were following the formal rules.
So I’m currently, like, on a crusade to make sure the people around me that I end up in the most disagreements with are able to nail the Formal Doublecrux Framework, and once we’re all roughly on that page I’ll trust us to do a post-formal version of it where we trust each other to get the basics right, so we can relax about the rules.
I believe I’m one of the people who commented on your strong focus on using the Double Framework recently, but on reflection I think can clarify my thoughts. I think generally there’s a lot to be said for sticking to the framework as explicitly formulated until you learn how to do the thing reliably and there’s a big failure mode of thinking you can skip to the post-formal stage. I think you’re right to push on this.
The complication is that I think the Double-Crux framework is still nascent (at least in common knowledge; I believe Eli has advanced models and instincts, but those are hard to communicate and absorb), which means I see us being in a phase of “figuring out how to do Double-Crux right” where the details of the framework are fuzzy and you might be missing pieces, parts of the algorithm, etc.
The danger is then that if you’re too rigid in sticking to your current conception of what the formal framework of Double-Crux, you might lack the flexibility to see where you’re theory is failing in practice, and you need to update what you think Double-Crux even should be.
I perceive something a shift (could be wrong here) where after some conversations you started paying more attention to the necessity of model-sharing as a component of Double-Crux as maybe a preliminary stage to find cruxes, and this wasn’t emphasized before. That’s the kind of flexibility I think is need to realize when the current formalization is insufficient and deviation from it is warranted as part of the experimentation/discovery/development/learning/testing/etc.