Last time, I wrote about 11 core rationalist skills. Now I would like some help from the LW community: which of these skills am I good at, which ones am I bad at? Just to recap, the skills are:

  • Actually want an accurate map
  • Keep your eyes on the prize
  • Entangle yourself with the evidence
  • Be Curious
  • Aumann-update
  • Know standard Biases
  • Know Probability theory 
  • Know your own mind
  • Be well calibrated
  • Use analytic philosophy
  • Resist Thoughtcrime

I'll post a description of each one of these skills as a comment, and if you think I am good at that skill, vote it up. If you think I am bad at it, vote it down. Don't be too shy - even if you are biased or uncertain - because over the course of many votes, these biases and errors will cancel out to some extent. (This is the "guess the number of beans in a jar by asking 50 people to guess and taking the average" method)

EDIT: We can also comment on each rationalist skill to say how well I am doing at that skill. Later today, I will do this myself. 

Thanks in advance! 

40 comments, sorted by Click to highlight new comments since: Today at 8:21 PM
New Comment

My first concern is that people may tend to excel in lists of items that they draw up. The areas where one is likely to be weak may include areas which one doesn't value enough to either list or develop. That said, I'll discuss the weaknesses I see in Roko on the listed skills. I won't discuss his strengths because I have some of the weaknesses Eliezer listed in "Why Our Kind Can't Cooperate" and its successor posts.

I would say that "be curious" is the place where Roko needs the most work, at least compared to our typical crowd. Because he has leadership potential, "know your own mind" may be a higher utility place for him to focus though, as its long tail risks are larger

I was going to say Black Swan risks, but that's laughably wrong, I couldn't know very easily where the largest Black Swan risks are, though curiosity seems likely to catch them best, but I know perfectly well what the known, established long tail risks of leaders not knowing their minds are.

Without cultivated "eye on prize" and curiosity I suspect that analytic philosophy is a weakness rather than a strength. Likewise, resisting thoughtcrime is more likely to be a weakness if not coupled to knowing one's mind.

by the way, can you point me to the literature on this:

I know perfectly well what the known, established long tail risks of leaders not knowing their minds are.


My first concern is that people may tend to excel in lists of items that they draw up.

to what extent do you think that I have done that here?

General statement: your current self-analysis is probably less accurate than the Outside View you get by considering past mistakes. Look back at your old blogging of ideas you no longer subscribe to, and look for the failures of rationality there. It's most likely that those are the areas you still need the most work on.

Thanks, I may well have to do forensics on my blog...

Roko, even though presumably know you better than most Less Wrong users since I've met you in real life, I'm finding it very difficult to estimate your rationalist skills.

When I take an online personality test, I try to think of examples from my life that demonstrate that my personality is one way or another. It's often difficult to come up with examples, and sometimes I will come up with examples that contradict one another.

Given that it's difficult for me to come up with examples demonstrating how I am, and that I live with myself every waking hour, imagine how hard it is for me to do the same thing for you. I gave you a downvote on probability theory for having trouble explaining causality to me, and an upvote on entangling yourself with the evidence because your website says your morality changed "almost overnight" after reading The Terrible, Horrible, No Good, Very Bad Truth about Morality, but what am I supposed to do for the others? Imagine something that would demonstrate presence or lack of a characteristic and then try and figure out if you would do it?

I guess that might work. An interesting variation on this exercise would be for you to supply this action that demonstrates your strength or weakness in advance. This would make things easier on the people being surveyed and might help you interpret your results better.

Feedback appreciated. It is an experiment - we'll see what comes of it in the end!

Roko Knows standard Biases: Have conscious knowledge of common human error patterns, including the heuristics and biases literature; practice using this knowledge in real-world situations to identify probable errors; practice making predictions and update from the track record of your own accurate and inaccurate judgments.

Roko resists Thoughtcrime. Keep truth and virtue utterly distinct in your mind. Give no quarter to claims of the sort "I must believe X, because otherwise I will be {racist / without morality / at risk of coming late to work/ kicked out of the group / similar to stupid people}". Decide that it is better to merely lie to others than to lie to others and to yourself. Realize that goals and world maps can be separated; one can pursue the goal of fighting against climate change without deliberately fooling oneself into having too high an estimate (given the evidence) of the probability that the anthropogenic climate change hypothesis is correct.

Roko Uses analytic philosophy: understand the habits of thought taught in analytic philosophy; the habit of following out lines of thought, of taking on one issue at a time, of searching for counter-examples, and of carefully keeping distinct concepts distinct (e.g. not confusing heat and temperature; free will and lack of determinism; systems for talking about Peano arithmetic and systems for talking about systems for talking about Peano arithmetic).

What is the difference between free will and lack of determinism? What's the definition of free will?

Well for one thing, there appear to be non-deterministic systems (possibly real ones, definitely conceived ones) that we would never say have free will. If x is a radioactive atom whether or not it decays in the next minute is undetermined. But the atom does not have free will. And actually, it is hard to make sense of what free will would be if it was just a lack of determinism since the extent to which an event is undetermined is also the extent to which it is random. One cannot control something that is random. But free will means having control over our actions. Ergo: if our actions are free they cannot be undetermined.

(And I know the particle decay picture is resolved differently with MW, but we don't have any other non-deterministic systems to talk about.)

See the SEP on compatibilism

The fact that you can ask the question "is free will compatible with physical determinism?" means that free will and lack of determinism are, a priori, distinct concepts.

Hmm. You can ask the question: "are most leprechauns taller than fairies?" too.

In itself, asking questions doesn't prove very much.

"are most leprechauns taller than fairies?"

Yes. p=0.95.

it does seem to show that "fairy" and "leprechaun" are distinct concepts, which is not entirely a contentless statement.

The point was that merely asking a question proves very little.

For another example, one can ask whether felines are smarter than cats (on average).

That's a coherent question - and its answer is "no".

it does seem to show that "fairy" and "leprechaun" are distinct concepts, which is not entirely a contentless statement.

But everyone knows that LEPRecon (Lower Elements Police reconnaissance division) is a job title and leprechauns are, in fact, fairies. This leads me to approximate the question and answer "most leprechauns are above average fairy height". This is to be expected for cops in general, above or below ground.

It seems that even if leprechauns are fairies the concepts are distinct.

I'll post a description of each one of these skills as a comment, and if you think I am good at that skill, vote it up. If you think I am bad at it, vote it down. Don't be too shy - even if you are biased or uncertain - because over the course of many votes, these biases and errors will cancel out to some extent. Thanks in advance!

The usual approach with this kind of thing is to include another comment as a karma sink for those who desire one. That can be expected to increase the accuracy of the feedback you infer.

I like the idea approach you are taking here. Proactive and rather bold. I've added a few votes here and there. I have to note, however, that I while I have a general positive impression of your posts I haven't built up a particularly clear model of how you think. This is, all else being equal a good thing. I mostly form models of people's thought processes when they have obvious biasses that I find damn irritating.

include another comment as a karma sink

Downvote some of the things you think I am bad at, or less good at. I appreciate the feedback!

Roko is Curious: Look for interesting details; resist cached thoughts; respond to unexpected observations and thoughts. Learn to acquire interesting angles, and to make connections to the task at hand.

[-][anonymous]12y 0

Why do you think that a large enough number of people here remember enough of the comments/articles you've posted well enough to be able to accurately give you this level of detailed feedback? I'm somewhat doubtful that the results will be useful to you at all, but we'll see.

Roko Entangles himself with the evidence. Realize that true opinions don't come from nowhere and can't just be painted in by choice or intuition or consensus. Realize that it is information-theoretically impossible to reliably get true beliefs unless you actually get reliably pushed around by the evidence. Distinguish between knowledge and feelings.

Roko Knows Probability theory: Have conscious knowledge of probability theory; practice applying probability theory in real-world instances and seeing e.g. how much to penalize conjunctions, how to regress to the mean, etc.

Roko actually wants an accurate map, because he has Something to protect.

You seem to want an accurate map. I've got no idea if this is motivated by having something to protect. Do you?

In my own analysis of my strengths and weaknesses, which I'll post once this exercise is over, I did place some significant probability on this.

Roko is well calibrated: Avoid over- and under-confidence. Know how much to trust your judgments in different circumstances. Keep track of many levels of confidence, precision, and surprisingness; dare to predict as much as you can, and update as you test the limits of your knowledge. Develop as precise a world-model as you can manage.

Roko Knows his own mind: Have a moment-to-moment awareness of your own emotions and of the motivations guiding your thoughts. (Are you searching for justifications? Shying away from certain considerations out of fear?) Be willing to acknowledge all of yourself, including the petty and unsavory parts. Knowledge of your own track record of accurate and inaccurate predictions, including in cases where fear, pride, etc. were strong.

Roko Aumann-updates: Update to the right extent from others' opinions. Borrow reasonable practices for grocery shopping, social interaction, etc from those who have already worked out what the best way to do these things is. Take relevant experts seriously. Use outside views to estimate the outcome of one's own projects and the merit of one's own clever ideas. Be willing to depart from consensus in cases where there is sufficient evidence that the consensus is mistaken or that the common practice doesn't serve its ostensible purposes. Have correct models of the causes of others’ beliefs and psychological states, so that you can tell the difference between cases where the vast majority of people believe falsehoods for some specific reason, and cases where the vast majority actually knows best.