Create a Full Alternative Stack

My interpretation of the previous several posts is: alignment of organizations is hard, and if you're even a little bit misaligned, the mazeys will exploit that misalignment to the hilt. Allow any divergence between measures of performance and actual performance, and a whole bureaucracy will soon arise, living off of that divergence and expanding it whenever possible.

My interpretation of this post is: let's solve it by making a fund which pays people to always be aligned! The only hard part is figuring out how to verify that they are, in fact, aligned.

... Which was the whole problem to begin with.

The underlying problem is that alignment is hard. If we had a better way to align organizations, then organizations which use that method would already be outperforming everyone else. The technique would already be used. Invent that technology (possibly a social technology), and it will spread. The mazes will fight it, and the mazes will die. But absent some sort of alignment technology, there is not much else which will help.

This is a problem which fundamentally cannot be fixed by throwing money at it. Setting up a fund to pay people for being aligned will result in people trying to look aligned. Without some way of making one's measure of alignment match actual alignment, this will not do any good at all.

One needs an order of magnitude more than that over a lifetime to be secure while not compromising one’s interactions with society. 

I'm confused by this. Isn't compromising one's interactions with society the point?

That is, suppose most costs are rent-seeking. If you want your startup to do something in the Bay, you have to pay Bay Area landlords about half of your investment capital (through rent and increased salaries). But why does your startup have to do something in the Bay? Because marketing, both to customers and future investors, and employees who want to be able to jump between companies, and various other things.

If you instead want to deliberately avoid marketing / mazes, why not do it in rural Pennsylvania? The rents are low, there.

Like, it seems to me the thing you're suggesting is something like an Amish community, but with something more healthy than God at the center. And that suggests you should do something more like what the Amish do, and less like staying in NYC, where if you don't have ~$10M in the bank or in expected future compensation you're not going to be financially secure while purchasing all the markers of being in the professional class. Like, why pay more for a house in a 'good' school district when you're just going to unschool your kids anyway?

Freeing a select group to do things without regard to such concerns, and people knowing they have the option to join this group, would be a major cultural change. Ideally this would then become the ‘city on a hill’ that shows what is possible, and gets emulated elsewhere.

I think the history of the Thiel Fellowship should be interesting, in this regard. My sense is that it tried to do this, couldn't find the people for it, and then pivoted to just be another way to perform well in the broader maze.

Create a Full Alternative Stack [LW · GW] is probably in the top 15 ideas I got from LW in 2020. Thinking through this as an option has helped me decide when and where to engage with "the establishment" in many areas (e.g. academia). Some parts of my life I work with the mazes whilst trying not getting too much of it on me, and some parts of my life I try to build alternative stacks. (Not the full version, I don't have the time to fix all of civilization.) I give it +4.

Broader comment on the Mazes sequence as a whole:

The sequence is an extended meditation on a theme, exploring it from lots of perspective, about how large projects and large coordination efforts end up being eaten by Moloch. The specific perspective reminds me a bit of The Screwtape Letters. In The Screwtape Letters, the two devils are focused on causing people to be immoral. The explicit optimization for vices and personal flaws helps highlight (to me) what it looks like when I'm doing something really stupid or harmful within myself.

Similarly, this sequence explores the perspective of large groups of people who live to game a large company, not to actually achieve the goals of the company. What that culture looks like, what is rewarded, what it feels like to be in it. 

I've executed some of these strategies in my life. I don't think I've ever lived the life of the soulless middle-manager stereotyped by the sequence, but I see elements of it in myself, and I'm grateful to the sequence for helping me identify those cognitive patterns. 

Something the sequence really conveys, is not just that individuals can try to game a company, but that a whole company's culture can change such that gaming-behavior is expected and rewarded. It contains a lot of detail about what that culture looks and feels like.

The sequence (including the essay "Motive Ambiguity") has led me see how in such an environment groups of people can end up optimizing for the opposite of their stated purpose.

The sequence doesn't hold together as a whole to me. I don't get the perfect or superperfect competition idea at the top. Some of the claims seem like a stretch or not really argued for, just completing the pattern when riffing on a theme. But I'm not going to review the weaknesses here, my goal is mostly to advocate for the best parts of it that I'd like to see score more highly in the book.

This post is one of my three picks from the sequence, along with The Road to Mazedom, and Protecting Large Projects Against Mazedom. (Also Motive Ambiguity which is not technically part of the sequence.)

(This review is taken from my post Ben Pace's Controversial Picks for the 2020 Review [LW · GW].)

What would such a full-stack organization do?

They couldn't try to optimize for any thing.

So... don't optimize? Not even for longevity. But also don't signal anti-optimization, or whatever the opposite of optimizing is. Don't reward signalling, and don't do any of the opposites of rewarding signalling.

I try to imagine such an entity ex nihilo, and I keep getting a novel experience that I'm going to call 'paradigm error'. But when I try to apply those qualities to various organizations, there is simply a mismatch- and different organizations mismatch different things.

Don't try to optimize anything, including meta to this description. Instead, try to do a good job of that thing. Including doing a good job at this meta-thing. My intuition says that doing a good job at avoiding Goodhart while doing a good job of doing a good job things is often going to mean using fuzzy metrics. I can't describe what I actually mean by 'fuzzy metrics', but using fuzzy metrics to evaluate a thing is adjacent to, and not, having someone observe the thing and then rate how good they think it was on a scale (that method is a badly done hard metric). It might look like a narrative evaluation of an expert observer, but I think a core feature of what I'm calling a 'fuzzy metric' is there is no way to generate a fuzzy metric by following a written or unwritten formal procedure.

When checking to see if you're doing a good job, look at some things that can be measured objectively, and combine those measurements with a thing that it is impossible for me to tell you how to get. Maybe "Don't not go with your gut." (double negative intended) might be a good job of explaining the non-measurement part of evaluation.

Such a full-stack organization can of course not optimize for maintaining an ideal culture, because that would be sacrificing literally all value. But they can try to do a good job of maintaining a good culture, identifying people who make the culture worse and humanely moving them to locations where they stop influencing the organizational culture.

In large organizations, hierarchy is impossible to avoid. I think a good tool to reduce that is to say that each level of hierarchy should have a unique object-level thing that they do, beyond bookkeeping or managerial tasks for the other levels. If in a corporate context, anyone who successfully replaces their own job with a few spreadsheet fomulae should not by default be punished for/with Redundancy.

Hmm. This seems to ignore the underlying actor-agent problem that is a partial cause of mazes: most (perhaps all, perhaps including you, certainly including me) people aren't willing to truly dedicate their entire life to a thing. There just aren't enough people who are willing/able/whatever to ignore all interpersonal competition for some of the slack (whether that be money or time or other non-shared-goal-directed value).

I think I am missing one possible solution in your list of ten solutions though the tenth one seems closest to what I called "General Public Contract" in this LW question from 2020 [LW(p) · GW(p)] (though the idea is older):

General Public Contract - A growing cooperative society - by some called a cult - that applies a positive-sum mechanism to build a better society on top of - or embedded in the existing society at large. It does so by using a valid legal contract among all parties. A contract like the "General Public Virus", that requires participants to contribute to it to gain its benefits. One extreme example would be a contract that requires you to effectively give up your private property except for your immediate belongings in exchange for access to the net utility of the property managed under the contract. 

Milestones are set. If they are met, funding is unlocked at pre-negotiated levels, locked in for both sides in advance.

It seems to me like the pressure to Goodhart are higher when you agree to be only funded by a single entity with specific pre-negotiated milestones then if you are in a state where you are going to seek capital from a bunch of people with at least slightly different evaluation criteria.

I may be oversimplifying here, but if I wanted to sum up what being a maze is about in a few words, I would say it has to do with a kind of Goodhart law problem: in large organizations, it is hard for the top to get information, so everything is reduced to simple metrics, and we end up optimizing for them, eventually destroying everything else. If that problem of information really is the bulk of it, the fact that your plan does not really remove the need for indirect measurements of things is the big issue, and I am not sure of what could be done to solve that. In fact, I’m not sure we could do it completely in any context: You want to reward people who "Disengage entirely with mazes and traditional distortionary incentives, competitions and signals of all kinds", for mazes, something might be tried, but disengaging from status-seeking, attractive as it sounds, looks like saying "just disengage with cognitive biases", good goal if you know you won’t achieve it.

More broadly, the most reliable thing I can think of to reduce these communication problems — indeed the one thing we replaced with mazes — is a lot of social capital, which probably implies limiting oneself to small communities and small businesses. Also, much more social pressure, with all its problems. To an extent, that’s the point, but it’s also something we really would want to avoid doing too much. Or it may even be that we are mainly complaining about mazes because that’s what they are already doing. I wonder to what extent the need for social capital if we don’t want to have mazes might be compensated by the fact that, compared to an hypothetical pre-mazes era, communication costs are down by a massive amount nowadays.

 

I read this and thought of organized religion. Unable to figure out why though.