There are still two elephants in the room that I must address before concluding. Then I will discuss paths forward.

Moloch’s Army

The first elephant is Moloch’s Army. I still can’t find a way into this without sounding crazy. The result of this is that the sequence talks about maze behaviors and mazes as if their creation and operation are motivated by self-interest. That’s far from the whole picture.

There is mindset that instinctively and unselfishly opposes everything of value. This mindset is not only not doing calculations to see what it would prefer or might accomplish. It does not even believe in the concept of calculation (or numbers, or logic, or reason) at all. It cares about virtues and emotional resonances, not consequences. To do this is to have the maze nature. This mindset instinctively promotes others that share the mindset, and is much more common and impactful among the powerful than one would think. Among other things, the actions of those with this mindset are vital to the creation, support and strengthening mazes.

Until a proper description of that is finished, my job is not done. So far, it continues to elude me. I am not giving up.

Moloch’s Puzzle

The second elephant is that I opened this series with a puzzle. It is important that I come back to the beginning. I must offer my explanation of the puzzle, and end the same place the sequence began. With hope.

Thus, the following puzzle:

Every given thing is eventually doomed. Every given thing will eventually get worse. Every equilibrium is terrible. Sufficiently strong optimization pressure, whether or not it comes from competition, destroys all values not being optimized, with optimization pressure constantly increasing.

Yet all is not lost. Most of the world is better off than it has ever been and is getting better all the time. We enjoy historically outrageously wonderful bounties every day, and hold up moral standards and practical demands on many fronts that no time or place in the past could handle. How is this possible?

Here is my answer:

It is possible because old things that get sufficiently worse eventually die, and get replaced by new things that are better. It is possible because competition is never perfect, and optimization for a fixed set of metrics is never total. People need and demand slack, and care about many factors on many meta levels. No matter how many times we say it is difficult, not game theoretically sound or even impossible, coordination continues to happen all the time.

Moloch wins when all things are equal and the situation is at a strict static equilibrium. Things are not equal. The situation is not static or strict. While there are some places where Moloch has not won, exit to those areas serves as an additional safety valve.

We have spent the bulk of this sequence dealing with mazes. Mazes are, among other things, a case of what happens when all of that breaks down – when we strip away the protections. They then go on to create all sorts of strange negative feedback loops, and make life inside them, and if left unchecked life even much of life outside of them, quite bad. In many ways, even this sequence has only scratched the surface of the dynamics involved.

Good News, Everyone!

To the extent that this makes one estimate that the world is a worse place, and its people are less happy and less likely to succeed or improve, this is bad news indeed.

But, to the extent that we already knew there was a problem, to the extent that we already knew how functional our world was and how happy people are, this is not bad news. This is good news. We already knew there was a problem and had an amorphous sense of hopelessness and doom. Now that we know more about what the problem is, we have a more specific source of hopelessness and doom that we can try to do something about. We can explore further. We can limit the damage on a personal level, seek to get things done, and perhaps even improve conditions in general.

The other good news is in our estimates of the effectiveness of doing object level things. We know about how much real and useful stuff is actually created and accomplished, in the sense that we know what useful stuff we get to use and we live real lives. If we are screwing people up this much, we are wasting this much of everyone’s time, and otherwise getting in our own way, then what does that say about how powerful it is to actually do things? If only a handful of people at a major corporation are producing all the value while everyone else plays politics, then what does that say about that handful of people?

If, as my model holds, people who are allowed to actually do real things consistently output insanely great things, then imagine what would happen if we let more people do that. Then realize there are still places where this is possible. Imagine what you could accomplish if you did that.

I hope this journey has been enlightening, and that it inspires further explorations along with concrete actions. Perhaps we can create useful common knowledge. With luck, some people will avoid falling into maze traps as the result of what has been written here. With even better luck, this can improve the fate of some organizations, or even lead to broader motions towards taking down maze levels.

Conclusion

The central (Immoral) Mazes sequence is now complete.

I wrote it because someone had to, and no one else would. This was the best model I could come up with, presented as clearly as I know how to do so.

There are many things I still want to say, most of which are not mentioned above. Some of them are a matter of putting in the work to write things down. Others I don’t know how to say without sounding crazy, or I don’t understand them myself well enough to write about them. I have experience with mazes, but not with higher maze levels. I can abstractly model that mentality and mode of operation, but I can’t pretend I truly understand it.

The more others can join the conversation and move it forward from here, especially those with more direct experience but that managed to emerge intact enough to tell the tale, the more hope we have for making real progress.

To encourage the story to continue, both for myself and for others, I will now explore the question of Paths Forward. What are the best next questions?

Paths Forward

These topics are not only or even primarily for me. Others can and should take up the mantle as well. Here are some things that seem like good ideas to keep the ball rolling.

I still don’t have (1a) Moloch’s Army in a place where I am ready to post it, but I promise to keep trying. At a minimum, I hope that trying leads to finding more missing concepts and ideas that can help bridge the gap.

I previously tried to bridge into this with The Darwin Game sequence, but that foundation wasn’t enough. If I resumed that line now then I would next talk about (1b) The Rise of Cliquebot. I am still confused on whether that would (eventually) get me where I would need to be.

One of the words I know I will likely need is (2a) Fnord. The word fnord comes from the Principia Discordia. In its original introduction, we are all conditioned when we see the word fnord to not notice it but to become stressed and irritated. Thus, it is peppered into places that people with power want us not to look, but it is entirely absent from things like advertising. Having a name for things that make you not want to notice them, and understanding how that dynamic works, seems important. Mazes, and many aspects of them, are fnords. Another good closely maze-aligned Discordian concept is (2b) The Snafu Principle, whereby communication is only fully possible between equals, leading to Situation Normal All F***ed Up. It should be mostly already covered by implication but is worth a special focus slash better place to link to.

Another key idea I need are (2c) Basilisks, as such things have much more important rolls to play than the example that gave them their name. Needless to say one must proceed with caution. Prior restraint will be a thing, here.

I introduced terms for things related to mazes, including maze behaviors, leading Wei Dei to ask the logical question (3a) What Are Maze Behaviors? It would be good to have a compact link-to-it answer. But I’m also coming around, after having started a draft of this, to the perspective that this is likely asking the wrong question. We could be better served to ask about (3b) Maze Levels and especially about (3c) The Maze Nature, which I also introduced, instead. The Maze Nature is closely related to Moloch’s Army and might be the right way in. Or perhaps it will be the other way around.

One worry is that mazes are the wrong central concept on which to build these metrics and understandings. It might be more helpful, either in general or in some places, to talk about (4) Simulacrum Levels. Ben’s post here is the best reference created so far, but it definitely needs to be improved upon. I also do not properly understand simulacrum levels, or at least my understanding and Ben’s are importantly different. These things need to be explored carefully, as the dynamics seem central to what is happening to the world.

Many of the proposed solutions, and hypothesized causes, would benefit from more careful treatment. People could gather numbers, make profiles and case studies especially where they have their own experience. Many great comments were (5a) Maze Examples, people talking about the place they work and saying how that fits into the bigger picture. Some claimed they worked at mazes. Many claimed that where they worked should have been a maze, based on size and what not, but that it totally wasn’t one or at least wasn’t as much of one as one would expect.

The question of (5b) How Did They Avoid Mazedom? is a good one for people who claim it was avoided. And of course someone should try to answer the question (5c) How Maze-Intensive Are Our Corporations? A few things to consider along the way,

The first is that these are the people reading DWATV and/or LessWrong. That is not a random sample. The whole point of such websites is to promote a method of thinking that is incompatible with a maze. If you care enough about good methods of thinking to read these websites, you likely have a strong aversion to maze behaviors compared to others with similar levels of human capital, both not tolerating them in others and not being willing to do them yourself. You likely chose your place and class of employment in part on this basis, regardless of how you thought about that decision when making it.

If someone was under full sacrifice-everything pressure from anything, they’re not going to read a book length sequence about it, because they can’t.

The second is that the best way to get comments on the internet is to get someone to tell you that you are wrong. People who think the model is failing in their case are much more likely to comment (or so my model says) than people for whom the model is accurate. That is how commenting typically works. So the sample does not provide much evidence, in that sense.

The third is that mazes are things people do not want to see, and people will select for choosing exactly the mazes they do not notice. If they did notice them, and were of the type to be reading this, they would probably avoid them. My estimate of the maze levels of several of my jobs jumped dramatically in the months after I left, because I had the chance to get perspective and to experience life without that aspect of things. Even if you are not yourself being eaten by the maze in question, it is instinctive to not notice it, or to justify what you are experiencing as normal and healthy and reasonable, or at least not so bad, when it is nothing of the kind.

It is also inevitable that some people have already self-modified to adapt to the mazes and that this makes it harder for them to notice what is going on. Not only does it mask this thing especially hard, it masks noticing things in general. Eventually, this would drive people away from reading such things, but inertia in such matters can last a while.

I am sure reading this sequence helps (in a probabilistic sense) but it is a hard problem. It is clear to me from the comments that I am not yet getting the full situation across successfully to many readers.

One thing someone will need to at some point write (6a) Mazes That Are Not Within Organizations, discussing dynamics that produce similar results without people strictly being bosses and subordinates. And generally (6b) What Types of Things are How Maze-Like, (6c) To What Extent do People At Large Have the Maze Nature, (6d) Close Examination of Maze Interactions, and so on.

The model of perfect competition presented early in the sequence proved very non-intuitive to many people and got a lot of push back. A lot of that was my using technical economic terms that trigger strong intuitions about what the answers are. I think a lot of this is that these terms are usually taught in a school context, where there are right answers and right principles that these things are supposed to illustrate, and the models always work a certain way. I was using them a different way, changing the assumptions and what things would be implied versus not implied, in order to provide justification for the transition to the later part of the sequence.

If I had to do that over again, I’d look for a way to take a different approach entirely, because it looks like it was more confusing than I expected, and the objection I felt I needed to overcome was much weaker than I expected it to be slash to the extent it existed the people who had the objection didn’t feel satisfied by the explanation. So it didn’t really work the way I wanted it to. I do stand by what I wrote, and think there’s important stuff there, but kill your darlings and all that.

However, there’s a whole series of posts I could write (7a) On Competition, going deeper into what I was getting at there. Or even (7b) On Ultimate Human Value and all that, which is kind of a big deal, but again, super hard and I’m constantly terrified of writing to advocate ethical positions because I assume others with better rhetoric and academic chops in the area would just blow anything I write to shreds and nothing would be accomplished. That doesn’t mean I think I’m wrong or anything, but it’s a problem, and it’s why the old “(7c) Can I Interest You In Some Virtue Ethics” post never got written after the last paths forward. Still, (7d) How Consequentialism Is Ruining Things And What To Do About It might be more tractable. I thought our decision theory would be better than this by now, but alas.

Another valuable thing to do would be to give people practical advice. Advice on how to choose fields of study and careers and jobs and where to live, and other major life choices, to avoid the dangers of mazes, while taking into account the many other things that matter. We have things like 80,000 Hours, but that leads to a consequentialist calculation with many of the key terms missing because they haven’t been quantified, and many of the best options never considered because they can’t be standardized, which is playing right into the hands of mazes. It’s the same as the usual social pressure to go work for mazes, if hopefully a little more efficient about collecting at least something in return. So, (8) Practical Career Advice to Avoid Mazes, A Continuing Series.

In particular, we need (9a) How to Start a Small Business. Not a start-up – I don’t think creating or joining one of those is a bad idea, but the how of that is something we are much more familiar with already – but a small business that actually does business. A way of life that involves buying things and then making or transporting things and selling things and having the things you sell generate revenue that pays expenses plus the rent. The kind of thing immigrants often do to great effect, except you can do even better if you are already integrated into the culture and have better connections and seed funding.

More fundamental would be the simple (9b) How to Do Business. There is specialized knowledge of how to start a business, but the most important doing-business related skill is simply how to do business. Start-ups backed by VCs are different because they fundamentally do not (yet) do business. They might do some facsimile of business by getting customers, or even unit economically profitable customers. But that’s a completely different model of how to succeed than trying to find customers and make money from them and use that money to pay the rent and the surplus to grow the business. What such start-ups are actually doing is the performance art of “doing business” aimed at the VCs that are their bosses. Doing business is a different thing entirely.

One hint you might be in a maze is that you are “doing the thing” in quotation marks rather than doing the thing.

I could also perhaps do (10a) Some Businesses Worth Starting, (10b) Some Start-Ups Worth Starting, (10c) Some Ways To Make Money I Made Work Before, and/or (10d) Some Ways To Make Money I Didn’t Do But I’m Confident Will Work. I’m sure good versions of 10c and 10d would be appreciated, but it’s a truism that no one cares about your start-up idea, whether or not they should, so perhaps not 10a and 10b.

Looking at the potential causes in more detail, and attempting to better understand their dynamics, would also be valuable. (11a) To What Extent Do We Need Large Organizations?, (11b) The Demand for Blamelessness, (11c) The Illusion of Security, (11d) The Dynamics of Rent Seeking, (11e) On Atomization, (11f) Education as a Maze, (11g) Education as Maze Indoctrination, and so on.

We could also look further at the solutions. One valuable place to go if we could do it well would be (12) How to Explain Mazes. We could also look at (13a) The Full Alternative Stack in much more detail. The biggest problem starts with (13b) How to Tell if Someone has the Maze Nature slash (13c) How to Tell if Someone Has Edge (or just (13d) Edge). That will need a better name because edge is a massively overloaded term. Here edge means the tendency to ‘angle shoot‘ or otherwise take every opportunity to take local advantage in underhanded ways. Do you always need to be ‘on guard’ against someone, or can you relax?

The other half is a post or series (13e) On Being a Source of Money. This is a huge unsolved problem. The moment anyone sees you (where you can be a person, or a group or organization, or anything else) as a source of funds, even a possible future source of funds, it corrupts all interactions even when everyone involved has the best of intentions. You always have to worry someone is after your money. The people around you are likely to be there largely because of your money. Others have to worry about whether you think they’re after your money even when they totally aren’t, and you can never rule such a thing out entirely. There is a reason a lot of rich people keep their wealth (or at least their giving/funding) a secret. Consider the chapter of Skin in the Game titled Only the Rich are Poisoned. Money can be shockingly useless or backfire, so what to do?

The other solutions to mazedom are largely political actions, where I am loathe to get more into the weeds for mostly-obvious reasons, or at least so far I have chosen to strive to be maximally apolitical throughout not only this sequence but on the entire blog. That’s another thing mazes do. They encourage us to stay out of such questions because they create an amorphous feeling that getting involved might backfire against us at some point in some way. Talking more about (14) How Mazes Scare You in such ways might be important. Of course, a lot of it is also that I do not need or want the trouble of arguing or advocating politics on the internet, nor do I expect much good to come of such a thing, and all that. And once I start down that road it is very easy to make oneself only seen from that perspective. It’s mostly a massively over-determined decision. But perhaps that decision is wrong, or will become wrong?

One more-related-to-this-than-you-would-first-think thing I’ve wanted to do for a while but that would require a lot of work and which might not come together, and which is motivated by this post, is to tell (15) The Journey of the Sensitive One. It would look over the story of the artist Jewel, as told, in explicit content in chronological order across her first five albums.

It might be most valuable of all to simply get case studies. I would love to see someone take this perspective and in particular examine (16a) What Happened to Boeing? Something seems to have happened that caused them to rapidly have higher maze levels, to the extent that they were unable to produce a plane that did not crash, ignoring huge quantities of written warnings that they were in fact producing a dangerous plane, and despite this being a very very bad thing for such a business to do. Steve Jobs was mentioned as a potentially interesting special case, so someone looking into (16b) Apple and Mazes could be interesting, especially to see if maze levels declined there when he came back and if so how he did that. And so on.

If one wanted to do a full extension of the project, (17) On The Gervais Principle could be anything up to and including its own sequence. As I’ve noted elsewhere, I consider Gervais Principle and Moral Mazes to be fully compatible, and Gervais Principle has a bunch of stuff that expands upon the model.

I’m sure I’m forgetting a lot of other stuff, as well. And of course, I have other things I need to and/or want to write, both about gaming and about other stuff. The game I’m creating is almost ready to get rolling, so I’ll probably want to focus on the gaming side even in my writing more and more over the coming months.

I seem to have already written close to a book’s worth of stuff. The main sequence (excluding this post, Quotes from Moral Mazes and Moral Mazes and Short Termism) is about 40,000 words long, there are a lot of places I could expand upon, and a quick Google says that the average business book is 50,000 to 60,000 words long. If anyone is seriously interested in publishing were I to turn this into a book, or has the know-how to make that happen and thinks it would be a good idea, please contact me with your thoughts and what that would look like. If it would be worthwhile to rewrite this in proper full book form I am open to that idea. If not, it might still be a good idea to print some copies mostly as-is to help reach more people.

And of course, thanks to you for reading, and hopefully thinking about and building upon all of this.

Acknowledgements

I would like to thank Ben Hoffman, Sarah Constantin and Michael Vassar for providing the impetus to read Moral Mazes and take it seriously. It takes a lot of motivation to get through a book that heavy.

I would like to thank Ben Hoffman, Michael Vassar, Raymond Arnold, Ben Pace, Oliver Habyrka, Zack Davis, Jessica Taylor and my wife Laura Baur for helping me edit the sequence. Without strong editorial feedback, this sequence would likely not have come to be, and if it did it would have been much weaker.

Ben Hoffman in particular was vital early on in helping me wrap my head around the problem, Michael Vassar later on for making sure I didn’t miss the important points, and Raymond Arnold for making sure what I wrote would have a fighting chance of being understood. Many thanks.

I’d also like to thank the commentators, including the ones who made comments I found frustrating and wrong, but especially the ones who challenged me in ways that improved my thinking.

And also those that helped keep up morale so I could finish. Engagement matters, and knowing you have people you respect interested matters too. Little notes of appreciation can go a long way. It’s important to give praise. This includes Scott Alexander, and also Robin Hanson, who said he would read Moral Mazes on the strength of the quotes I provided. I look forward to seeing both their takes.

I’d also like to thank the person who got How to Identify an Immoral Maze to be featured on Hacker News. The majority of hits I get come from being featured somewhere, in one form of another. I do my best not to let that change what I write, other than being happy to edit posts to make them stand on their own better if this is standing between them and being linked in this way. If that is ever the case and the modifications seem reasonable, please do let me know.

Finally, thanks again for reading. I hope the time you have invested has proved worthwhile.