Chiming in because the problem of helping people level up is close to my heart.
Putting the social dynamics of the experiment aside (since there are plenty of people discussing that aspect), I’d like to offer some good-natured skepticism about the overall approach. (Good-natured meaning, I hope you actually do pursue this because I’m genuinely curious about how this will play out—assuming the safety concerns others have raised are handled well, of course).
My skepticism is: this is too meta and too complicated to lead to actual progress.
I spent a few years at company that tried to inculcate a deliberate process for getting to the right answer, including a culture of radical honesty and formal procedures for making decisions and learning from mistakes. This was a major priority at the company for a long period of time (last I checked, it’s still going on), with backing from the entire senior management team, and was enforced by firing people who couldn’t or wouldn’t skillfully participate. I.e., they took it really seriously and put a lot of effort into it. The people who conceived and implemented it were in my opinion extremely smart and competent.
That said, in my opinion the effort spent on this program did more harm than good to the functioning of the company. The values and culture became an end in itself, as opposed to a means for helping achieve goals, and endless amounts of time and energy were spent debating, elucidating, learning, and critiquing the system. Competent professionals ended up becoming ineffectual because they gave up (or were forced out of) their unreflective expertise and got stuck in endless cycles of second-guessing. Some of that self-reflection may have given rise to new levels of skill (in my case, I did in fact feel like I benefited from my time there, although I think that was largely because it was my first job out of college so I didn’t have that much to un-learn), but generally people felt disempowered by the initiative rather than improved.
In contrast, for the last few years, I’ve been running a tiny company where we have very little meta discussion and mostly just do object-level work. I feel 1000x more productive now than I did at my prior job.
My takeaway from this is that the optimal ratio of meta-level tuning to object-level practice is [small number] : [large number]. Meta-level thinking is extremely valuable and important, but I view it as the rudder on a boat: you need to be constantly making adjustments to keep pointing in the right direction, but 99% of the power generation goes into the main engine pointing forward.
If I had to generate a hypothesis as to why the concrete achievements of the rationalist community are less than might be desired, it would be that the community spends way to much of its energy on meta topics instead of on object-level progress. This is understandable, since a) meta-level discussion of rationality is what created the community in the first place, and b) object-level discussion can often be very boring compared to meta-level discussion. (I miss the intellectual stimulation of my previous job, even as I see it as basically a waste of time in terms of actually building a successful company). While understandable, I think it leads to predictable outcomes: a lot of talk happens but not much gets accomplished.
Looking at the proposed charter, I suspect there will be a very high amount of meta-level discussion, probably significantly more so than at my prior job that I thought was way too meta. That’s because a) it’s built in to the daily schedule, b) it’s built into the mission, which is expected to evolve over time with the participants, and c) it’s built into the community that the participants will be drawn from.
In addition to being too meta, I also suspect this experiment is too complex. Experimenting with a bunch of different norms, on top of the code of conduct and daily schedule, seems wildly ambitious to me. In the company I worked for, the set of norms and practices were set in stone by executive fiat, recruits to the company were presented with them prior to accepting jobs, and adherence to them were a major part of performance evaluation, and there was still a very high employee churn rate and a general agreement that the norms / practices as specified weren’t consistently well-practiced throughout the company. The Dragon charter is for a smaller group of people, which makes things easier, but the norms / practices are expected to be a moving target, which makes things harder.
In my personal experiments with self-improvement, I’ve had the most success with extremely simple plans. My most successful self-intervention to date has been to download a simple habit tracker on my phone, and add a new daily habit, moving on to the next only after successful completion of the prior one for 30 days. When I first started trying to learn new habits, I would add a bunch of new habits at once, and I would always fail. It took me a very long time to get patient enough to only try to change one thing at a time (which requires accepting that I’m going to have habits I don’t like in the interim that I don’t try to do anything about).
Similarly, I’ve been successful growing my current company by having an extremely boring strategy of: ship code, talk to customers, ship code, talk to customers.
Simplicity does not come naturally to me; I like my ideas and strategies to be convoluted, complicated, ambitious, and interesting—I get very bored with simple, straightforward approaches. So I’m a big believer in simplicity because I’ve learned the hard way against all my natural inclinations that—unlike my natural inclinations—it actually works.
So if I were trying to design a charter, I would pick one or two things that I think would be most likely to have a game-changing impact, and just focus on those things until they worked (or didn’t). In contrast, the charter as it exists now feels to me like it has way too many moving pieces. That’s just my intuition, of course, but I hope I’ve given a feel for where that intuition comes from.
Anyway, I admire the ambition in doing a project like this, so I hope my criticism is constructive and useful.
Thanks for the long and detailed response. I enjoyed reading it.
It’s interesting that you highlight meta as being a dangerous failure mode—I actually strongly agree, which is why the aesthetic is tuned toward stuff like “just exercise” and “housemates should produce visible work.” My sense is that a strategy of just doing stuff outstrips in practice a strategy of think really hard until you find the ideal move, especially when you take into account how many iterations you can get in if you’re churning hard.
Hilariously, though, I’m further inside the rationalist bubble than I thought, because I accept your overall summation even though the intent was to be THE OBJECT LEVEL HOUSE (or at least, the house that does stuff even if it goes meta on norms). I still think we’re set up to be relatively ahead, but as you point out, that’s not necessarily a sufficient bar.
However, I’m much more concerned with:
In addition to being too meta, I also suspect this experiment is too complex. Experimenting with a bunch of different norms, on top of the code of conduct and daily schedule, seems wildly ambitious to me.
That rings very true to me, and has been an active concern of mine for the past couple of weeks. It seems like there are something like a hundred activities/experiments/norms/projects that are worthy of including in this, and something like 1.3 slots per week (and thus not even room for half), and I’m not at all certain how to best pick and choose and prioritize and optimize for success. In part, I’m hoping that if we just throw ourselves in and iterate (see above) we’ll do better than if we agonize, but yeah, there are a lot of moving parts, and I wouldn’t be surprised if we ended up trying to drastically simplify in like our fifth week house meeting.
If I had to really zero in on basics, I think they are:
Never give up on an experiment until its predetermined end date
Spend ~20 hours a week actually interacting in the same physical space as housemates (at least a subset)
… those, I think, are the iron core of the project.
Spend ~20 hours a week actually interacting in the same physical space as housemates (at least a subset)
I’m curious why this is so important to you, unless that it’s just something to try out. I currently live alone and I like it that way, and I see no reason why spending more time with other people would be such a great thing.
You seem really rigid about excuses though. I think the tendency will be that people will come up with an excuse which one finds it unpleasant or difficult to dispute. For example, when I was in the data science bootcamp in Berkeley, people would very frequently say, “I’m sick and I will be working from home today.” Now a lot of people were in fact sick precisely because of so much physical proximity. But it was very obvious in many cases that the basic reason they were staying home was that they were tired of all the company and felt the need to get away. They did not however feel comfortable saying, “I just feel the need to get away.”
The same thing was true when I lived in a monastery. You could not say “I just feel like sleeping in this morning,” so people said “I didn’t come this morning because I didn’t feel well.” We all knew that this simply meant they were tired and felt like sleeping in. But no one is comfortable confronting someone with the fact that they’re not really sick if they say they are.
The focus on physical presence is a combination of research showing that it matters (there’s some stuff I’ve collected from Dunbar, for example) and strong personal intuition from past experience. In many ways, it’s the core of the thing being tested out, but I have a lot of weight on “it turns out to matter more than just about anything else.”
re: excuses, the intention of the house is Not To Do The Stupid Thing.
Clearly, “mental health” days are a real phenomenon—I’ve taken some myself. And on a larger scale, psych blockers/motivational issues are also real. So it’d be stupid to a) pretend they don’t happen, and b) push directly against them all the time, and never look at undercutting them or working around them. This plan pushes directly against them some, with commitments to just show up anyway, but that’s not the only tool—one of the things I hope to do is increase the candor of all housemates, at least within the context of the house. This will take some practice and reinforcement, but I much prefer a norm of “Huh. I notice I just really didn’t want to show up today” --> figure out what’s going on and address it systematically, to a norm of “little white lie that nobody calls out.”
It’s also worth noting that the house has a pretty high introvert quotient, so there will be a lot of us (myself included) who are motivated to safeguard systems giving one the ability to get away from people for a while.
Thank you for writing that! It’s great to see the “too meta” problem spelled out so clearly. It’s similar to the situation in programming that has long puzzled me. Many people and companies have accumulated processes that they swear by (code review, type systems, continuous integration, agile and whatnot) but at the same time lots of people do amazing work with very little process.
It seems like meta stuff has a way of self-justifying and growing, like a bureaucracy. It’s useful if you’re stuck and nothing works, but if you’re making any progress at all, it’s better to steer with the engine so to speak. Radical meta proposals sound attractive to people who have fought their minds to a standstill, but even for such people I think a better idea is starting one small object-level thing on a strict schedule (gym is a good choice), making the mind more mobile for other things in turn.
Chiming in because the problem of helping people level up is close to my heart.
Putting the social dynamics of the experiment aside (since there are plenty of people discussing that aspect), I’d like to offer some good-natured skepticism about the overall approach. (Good-natured meaning, I hope you actually do pursue this because I’m genuinely curious about how this will play out—assuming the safety concerns others have raised are handled well, of course).
My skepticism is: this is too meta and too complicated to lead to actual progress.
I spent a few years at company that tried to inculcate a deliberate process for getting to the right answer, including a culture of radical honesty and formal procedures for making decisions and learning from mistakes. This was a major priority at the company for a long period of time (last I checked, it’s still going on), with backing from the entire senior management team, and was enforced by firing people who couldn’t or wouldn’t skillfully participate. I.e., they took it really seriously and put a lot of effort into it. The people who conceived and implemented it were in my opinion extremely smart and competent.
That said, in my opinion the effort spent on this program did more harm than good to the functioning of the company. The values and culture became an end in itself, as opposed to a means for helping achieve goals, and endless amounts of time and energy were spent debating, elucidating, learning, and critiquing the system. Competent professionals ended up becoming ineffectual because they gave up (or were forced out of) their unreflective expertise and got stuck in endless cycles of second-guessing. Some of that self-reflection may have given rise to new levels of skill (in my case, I did in fact feel like I benefited from my time there, although I think that was largely because it was my first job out of college so I didn’t have that much to un-learn), but generally people felt disempowered by the initiative rather than improved.
In contrast, for the last few years, I’ve been running a tiny company where we have very little meta discussion and mostly just do object-level work. I feel 1000x more productive now than I did at my prior job.
My takeaway from this is that the optimal ratio of meta-level tuning to object-level practice is [small number] : [large number]. Meta-level thinking is extremely valuable and important, but I view it as the rudder on a boat: you need to be constantly making adjustments to keep pointing in the right direction, but 99% of the power generation goes into the main engine pointing forward.
If I had to generate a hypothesis as to why the concrete achievements of the rationalist community are less than might be desired, it would be that the community spends way to much of its energy on meta topics instead of on object-level progress. This is understandable, since a) meta-level discussion of rationality is what created the community in the first place, and b) object-level discussion can often be very boring compared to meta-level discussion. (I miss the intellectual stimulation of my previous job, even as I see it as basically a waste of time in terms of actually building a successful company). While understandable, I think it leads to predictable outcomes: a lot of talk happens but not much gets accomplished.
Looking at the proposed charter, I suspect there will be a very high amount of meta-level discussion, probably significantly more so than at my prior job that I thought was way too meta. That’s because a) it’s built in to the daily schedule, b) it’s built into the mission, which is expected to evolve over time with the participants, and c) it’s built into the community that the participants will be drawn from.
In addition to being too meta, I also suspect this experiment is too complex. Experimenting with a bunch of different norms, on top of the code of conduct and daily schedule, seems wildly ambitious to me. In the company I worked for, the set of norms and practices were set in stone by executive fiat, recruits to the company were presented with them prior to accepting jobs, and adherence to them were a major part of performance evaluation, and there was still a very high employee churn rate and a general agreement that the norms / practices as specified weren’t consistently well-practiced throughout the company. The Dragon charter is for a smaller group of people, which makes things easier, but the norms / practices are expected to be a moving target, which makes things harder.
In my personal experiments with self-improvement, I’ve had the most success with extremely simple plans. My most successful self-intervention to date has been to download a simple habit tracker on my phone, and add a new daily habit, moving on to the next only after successful completion of the prior one for 30 days. When I first started trying to learn new habits, I would add a bunch of new habits at once, and I would always fail. It took me a very long time to get patient enough to only try to change one thing at a time (which requires accepting that I’m going to have habits I don’t like in the interim that I don’t try to do anything about).
Similarly, I’ve been successful growing my current company by having an extremely boring strategy of: ship code, talk to customers, ship code, talk to customers.
Simplicity does not come naturally to me; I like my ideas and strategies to be convoluted, complicated, ambitious, and interesting—I get very bored with simple, straightforward approaches. So I’m a big believer in simplicity because I’ve learned the hard way against all my natural inclinations that—unlike my natural inclinations—it actually works.
So if I were trying to design a charter, I would pick one or two things that I think would be most likely to have a game-changing impact, and just focus on those things until they worked (or didn’t). In contrast, the charter as it exists now feels to me like it has way too many moving pieces. That’s just my intuition, of course, but I hope I’ve given a feel for where that intuition comes from.
Anyway, I admire the ambition in doing a project like this, so I hope my criticism is constructive and useful.
Thanks for the long and detailed response. I enjoyed reading it.
It’s interesting that you highlight meta as being a dangerous failure mode—I actually strongly agree, which is why the aesthetic is tuned toward stuff like “just exercise” and “housemates should produce visible work.” My sense is that a strategy of just doing stuff outstrips in practice a strategy of think really hard until you find the ideal move, especially when you take into account how many iterations you can get in if you’re churning hard.
Hilariously, though, I’m further inside the rationalist bubble than I thought, because I accept your overall summation even though the intent was to be THE OBJECT LEVEL HOUSE (or at least, the house that does stuff even if it goes meta on norms). I still think we’re set up to be relatively ahead, but as you point out, that’s not necessarily a sufficient bar.
However, I’m much more concerned with:
That rings very true to me, and has been an active concern of mine for the past couple of weeks. It seems like there are something like a hundred activities/experiments/norms/projects that are worthy of including in this, and something like 1.3 slots per week (and thus not even room for half), and I’m not at all certain how to best pick and choose and prioritize and optimize for success. In part, I’m hoping that if we just throw ourselves in and iterate (see above) we’ll do better than if we agonize, but yeah, there are a lot of moving parts, and I wouldn’t be surprised if we ended up trying to drastically simplify in like our fifth week house meeting.
If I had to really zero in on basics, I think they are:
Never give up on an experiment until its predetermined end date
Spend ~20 hours a week actually interacting in the same physical space as housemates (at least a subset)
… those, I think, are the iron core of the project.
I’m curious why this is so important to you, unless that it’s just something to try out. I currently live alone and I like it that way, and I see no reason why spending more time with other people would be such a great thing.
You seem really rigid about excuses though. I think the tendency will be that people will come up with an excuse which one finds it unpleasant or difficult to dispute. For example, when I was in the data science bootcamp in Berkeley, people would very frequently say, “I’m sick and I will be working from home today.” Now a lot of people were in fact sick precisely because of so much physical proximity. But it was very obvious in many cases that the basic reason they were staying home was that they were tired of all the company and felt the need to get away. They did not however feel comfortable saying, “I just feel the need to get away.”
The same thing was true when I lived in a monastery. You could not say “I just feel like sleeping in this morning,” so people said “I didn’t come this morning because I didn’t feel well.” We all knew that this simply meant they were tired and felt like sleeping in. But no one is comfortable confronting someone with the fact that they’re not really sick if they say they are.
The focus on physical presence is a combination of research showing that it matters (there’s some stuff I’ve collected from Dunbar, for example) and strong personal intuition from past experience. In many ways, it’s the core of the thing being tested out, but I have a lot of weight on “it turns out to matter more than just about anything else.”
re: excuses, the intention of the house is Not To Do The Stupid Thing.
Clearly, “mental health” days are a real phenomenon—I’ve taken some myself. And on a larger scale, psych blockers/motivational issues are also real. So it’d be stupid to a) pretend they don’t happen, and b) push directly against them all the time, and never look at undercutting them or working around them. This plan pushes directly against them some, with commitments to just show up anyway, but that’s not the only tool—one of the things I hope to do is increase the candor of all housemates, at least within the context of the house. This will take some practice and reinforcement, but I much prefer a norm of “Huh. I notice I just really didn’t want to show up today” --> figure out what’s going on and address it systematically, to a norm of “little white lie that nobody calls out.”
It’s also worth noting that the house has a pretty high introvert quotient, so there will be a lot of us (myself included) who are motivated to safeguard systems giving one the ability to get away from people for a while.
Thank you for writing that! It’s great to see the “too meta” problem spelled out so clearly. It’s similar to the situation in programming that has long puzzled me. Many people and companies have accumulated processes that they swear by (code review, type systems, continuous integration, agile and whatnot) but at the same time lots of people do amazing work with very little process.
It seems like meta stuff has a way of self-justifying and growing, like a bureaucracy. It’s useful if you’re stuck and nothing works, but if you’re making any progress at all, it’s better to steer with the engine so to speak. Radical meta proposals sound attractive to people who have fought their minds to a standstill, but even for such people I think a better idea is starting one small object-level thing on a strict schedule (gym is a good choice), making the mind more mobile for other things in turn.