Rationality Habits I Learned at the CFAR Workshop


37


elharo

Recently Leah Libresco asked attendees at the January CFAR Workshop, "What habits have people installed after workshops?" and that got me thinking that now was a good time to write up and review what I learned (or learned and already forgot). I thought that might be of some interest to folks here, and this is what follows.

What I Learned and Implemented

The most immediately useful thing I learned was the Pomodoro Technique, as I've written about here before. In addition to that, there were a number of small items that I'm continuing to work on.

First, I've become quite fond of the question "Does future me have a comparative advantage?" Especially for small items, if the answer is "No" (and it's no far more often than it's yes) then just do it right now. The more trivial the task, the more useful it is. For instance, today I asked myself that while standing in the bedroom wondering whether to take 30 seconds to move my ExOfficio Bugproof socks from the dresser to the correct box in the closet. (Answer from a few minutes ago:  if I don't take my dog for a walk right now, he's going to pee all over the floor. Future me does have a comparative advantage of not having to clean up pee on the floor. The socks can wait.)

I've begun to notice my confusion and call it to conscious attention more often, though I suspect I learned this first from HpMOR and the sequences before the workshop. Example: when Leonard Susskind states that conservation of information is a fundamental principle of quantum mechanics, I notice that I am confused because A) I have never heard of any such fundamental law of physics as information conservation B) Every definition of information I have ever heard indicates that information most certainly can be destroyed. So just what the heck is he talking about anyway? I am now making a conscious effort to research this topic rather than letting it slide by.

The workshop introduced me to the concepts of System 1 and System 2. System 1 is the faster, reactive, intuitive mind that uses heuristics and experience to react quickly. System 2 is the slower, analytical, logical, mathematical mind. I didn't immediately grok this or see how to apply it. However the workshop did convince me to read Daniel Kahneman's Thinking Fast and Slow, and I'm beginning to follow this. It could be useful going forward. I particularly like the examples given at the end of each chapter.

Similarly I completely did not understand the concepts of inside view vs. outside view at the workshop; and worse yet I don't think that I even realized that I didn't understand these. However now that I've read Thinking Fast and Slow, the lightbulb has gone on. Inside view is simply me deciding how likely I (or my team) is likely to accomplish something based on my judgement of the problem and our capabilities. Outside view is a statistical question about how people and teams like us have done when confronted with similar problems in the past. As long as there are similar teams and similar problems to compare with, the outside view is likely to be much more accurate.

During conversation, Julia Galef and I came up with the idea of *********.  It turned out it already exists, and I'm planning to start attending these events locally soon. I've also joined my local LessWrong meetup group.

Stare into Ugh fields. Difficult conversations are an Ugh field for me. Recognizing this and bringing it to conscious attention has made it somewhat easier to manage these conversations. Example: when I went to the workshop I had been putting off contacting my dentist for months, not because of the usual reasons people don't like going to the dentist, but simply because I was uncomfortable telling her that the second (and third) opinion I had gotten on a dental issue disagreed with her about the proper course of treatment. Post-workshop, I finally called her (though it still took me two more weeks to do this. Clearly I have a lot of work left to do here.)

Consider whether the sources of my information may be correlated and by how much. I.e. Evaluating Advice. For instance, if two dentists who share an office give me the same advice, even assuming no prior disposition to agree with each other simply out of friendship, how likely is it that they share the same background and information that dentists in a different office do not?

COZE (Comfort Zone Expansion) exercises have pushed me to talk more to "strangers" and be intentionally more extroverted. On a recent trip to Latin America, I even made an effort to use what little Spanish I possess. I've had some small success, though this has led to no obvious major improvements in my life yet.

Thought experiments conducted at the workshop were very helpful in untangling some of my goals and plans. Going forward though this hasn't made a huge difference in my day-to-day life. That is, it hasn't led me to seek different paths than what I'm on right now.

What I Learned and Forgot

Going over my notes now, there was a lot of material; some of it potentially useful, that has fallen by the wayside; and may be worth a second look. This includes:

  • Geoff Anders introduced us to yEd, a nice open source diagram editor. I still prefer StencilIt or Omnigraffle though. He also used it to show us a really neat way of graphing, well, something. Goals maybe? I remember it seemed really useful and significant at the time, but for the life of me I can't remember exactly what it was or what it was supposed to show us. I'll have to go back to my notes. This is why we write things down. (Update: I suspect this was about Goal Factoring.)
  • Anticipation vs. Profession (though from time to time I do find myself asking what odds I'd be willing to bet on certain beliefs)
  • The Planning Kata.

What I Learned But Didn't Implement

Value of Information calculations seem too meta and too wishy-washy to be of much use. They attempt to put quantitative numbers based on information that's far too imprecise to allow even order of magnitude accuracy. I'm better off just keeping things I need to consider in my GTD system, and periodically reviewing it.

Similarly opportunities for Bayesian Strength of Evidence calculations, just don't seem to come up in my day-to-day life. The question for me is more commonly "Given that the situation is what it is, what actions should I take to accomplish my goals?" The outside view is useful for this. Figuring out why the situation is what it is rarely seems to be especially helpful.

Turbocharging Training may be helpful but the evidence seems to me to be lacking. I'd like to see some strong proof that this works in particular areas; e.g. foreign languages, sports, or mathematics.  Furthermore, it's not clear that it's applicable to anything I'm working on learning at this time. It seems very System 1 focused, and not especially helpful with the sort of fundamentally System 2 tasks I take on.

I have begun to declare "Victory!" at the end of a meeting/discussion. it's a bit of fun, but has limited effect. Beyond that I don't seem to reward myself for noticing things, or as a means of installing habits.

What I Didn't Learn

Getting Things Done (GTD), Remember the Milk, BeeMinder, Anki, Cultivating Curiosity, Overcoming Procrastination, and Winning at Arguments.

GTD I didn't learn because I've used it for years now or at least the parts of it that really work for me (lists and calendars mostly, and to a lesser extent filing).

Remember the Milk because my employer's security policy prohibits us from using it, and too much of my life happens at my day job to make maintaining two separate systems worthwhile.

BeeMinder and Anki because I just don't have anything that seems it could benefit from being stored in those systems right now. All of these might be more beneficial to someone in different circumstances.

Cultivating Curiosity because I am already a very naturally curious person, and have been for as long as I can remember. I don't need help with this. Indeed if anything I need to tamp down on this tendency and focus more on accomplishing things rather than merely learning them.

Similarly, Overcoming Procrastination didn't help a lot because I don't have a big procrastination problem, at least not compared to what I had when I was younger. Of course, I do say that in full knowledge that right this minute writing this article is a form of structured procrastination to avoid doing my taxes. :-)

Winning at Arguments, I am already very, very good at when I want to be, which is rare these days. It took me many years too realize that even though I "won" almost every argument I cared about, winning the argument wasn't usually all that useful. Winning an argument is the wrong goal to have for almost any purpose, and rarely leads to the outcomes I desire.

Unofficial ideas from fellow attendees:

Polyphasic sleep: I'm going to let the younger, more pioneering attendees experiment with this one. Even if it does work (which seems far from obvious) I don't see how one could integrate it into a conventional day job and family.

At breakfast one morning, a fellow attendee (Hunter?) suggested putting unsalted butter in my coffee to add more fat to my diet. It's not as crazy as it sounds. After all butter is little more than clarified cream, which I do like in my coffee. I tried this once and I still prefer cream, but I may give it another shot.

Finally, I've referred two workshop attendees to my employer as potential hires. If anyone else from the workshop is looking for a job, especially in tech, sales, or legal, drop me a line privately. For that matter if any Less Wronger is looking for a job, drop me a line privately. We have hundreds of open positions in major cities around the world. Quite a few LessWrongers already work there, and there's room for many more.

What the workshop didn't teach

There were a few techniques that were conspicuous by their absence. In particular I think the CFAR/LessWrong and Agile/XP communities have a lot to teach each other. I was surprised that no one at the workshop seemed to have heard of Kanban or Scrum, much less practice it. Burndown charts and point-based estimation are a really interesting modification of the outside view by comparing your team to your team in the past, rather than to other teams.

Pairing is also a useful technique beyond programming as at least Eliezer (not present at the workshop) has discovered. Pairing is an incredibly effective way to overcome akrasia and procrastination.

In reverse, I am considering what the craft of software development has to learn from CFAR style rationality, more specifically epistemic rationality. I have begun to notice my confusion during conversations with users, product managers, and tech leads and call it to conscious attention. I less frequently let unclear specs and goals pass without comment. Rather, I ask for examples and drill down into them until I feel my confusion has been conquered.

So far these techniques seem very useful in analysis and requirements gathering. I've found them less obviously useful (though certainly not harmful in any way) during coding, debugging, and testing. In these stages there's simply too much to be confused by to address it all, and whatever I'm confused by that's relevant to the task at hand rapidly calls itself to my attention. For instance, when a bug shows up in a production system, the very first and natural question to ask  is "How the hell did the system do that?!" On the other hand, the planning kata may be very helpful with the early stages of system design, though I haven't yet had an opportunity to try that out.

Was it Worth $3900?

Overall, I found the workshop to be a worthwhile experience, if an expensive one; and I recommend it to you if you have the opportunity and resources to attend. There are a lot of practical techniques to be learned, and you only need one or two of them to pay off to cover the cost and time. Even if the primary value is simply introducing you to books and techniques you explore further after the workshop such as Getting Things Done or Thinking Fast and Slow, that may be enough. Most knowledge workers are operating far below the level of which we're capable, and expanding our effectiveness can pay for itself.

Before attending, it is worth asking yourself whether there's an opportunity to learn this material at lower cost. For instance, did I really need to spend $3900 and 4 days to learn about Pomodoro? Apparently so, since I'd heard about Pomodoro for years and paid no attention to it until January. On the other hand, a $20 book I read on the subway was fully sufficient for me to learn and implement Getting Things Done. You'll have to judge this one for yourself.