Acquiring some skills is mostly about deliberate, explicit information transfer.  For example, one might explicitly learn the capital of Missouri, or the number of miles one can drive before needing an oil change, or how to use the quadratic formula to solve quadratic equations.

For other skills, practitioners' skill rests largely on semi-conscious, non-explicit patterns of perception and action.  I have in mind here such skills as:

  • Managing your emotions and energy levels;
  • Building strong relationships;
  • Making robust plans;
  • Finding angles of attack on a mathematical problem;
  • Writing persuasively;
  • Thinking through charged subjects without bias;

and so on.  Experts in these skills will often be unable to accurately and explicitly describe how to do what they do, but they will be skilled nonetheless.

I'd like to share some thoughts on how to learn such "soft skills".


Usefulness of non-true stimuli

If you read a chemistry textbook, it makes sense to ask after each sentence: “Is this true?”.  If the answer is “no”, “no”, “no”, for a sufficient number of sentences, you should probably abandon that book and look for a better one.  Chemistry textbooks are supposed to be made out of statements you can trust -- statements you can add to your file of “trusted explicit claims”, in such a fashion as to make you better at chemistry.  When a book fails at this property, its main value is lost.

Not so, IMO, for soft skills.

You can test ideas in your "inner simulator"

Your “inner simulator” is CFAR’s version of the distinction between profession and anticipation.  Basically, your “inner simulator” is the part of you that can play movies forward to determine what to anticipate: “Do I have time to turn left before that car reaches me?”; “What will she do, if I approach and say ‘hi’?” (that is: what does my inner movie-player show as the next scene, when I play it a movie in which I walk up to her and say ‘hi’?).

Your inner simulator is probably more accurate than your explicit models in domains where you have a rich experience base, such as social phenomena, and day-to-day physical phenomena.  It is probably worse in areas where you have good book-learning behind you (e.g., you may have an accurate conscious model of the bystander effect, but still mismodel this when you anticipate without conscious adjustment).  Your inner simulator is also the place where learning *must* land, if it is to affect your automatic system 1 responses (such as the perceptual patterns, and the trigger-action habits, that play into many soft skills).

IMO, most “soft skills” books are not trying to add explicit statements to your store of “trusted explicit/verbal statements”.  Instead, they are trying to evoke experiments to try out in your inner simulator -- bits that you can then keep, or not, according to whether they feel promising when you imagine trying them out.  Later, you can try the promising bits out in the actual world.

Example: Roommate

To see how this can work, imagine you’re in a tricky social situation.  Perhaps your roommate, Fred, is easily offended, and also keeps leaving the kitchen in a state of total mess.  You’ve brainstormed a number of options for talking to him, but they all seem likely to end badly, and so you find yourself in the self-help section of the bookstore, looking for, well, help. 

As you browse, you notice a lot of advice that you’ve heard before -- advice like “imagine what things feel like from Fred’s perspective” and “explain what’s in it for Fred; find a way to appeal to his pride and self-interest”.  You could’ve generated a lot of this advice yourself.  Nevertheless, much of it is advice you hadn’t actually tried, in Fred’s case.  You find yourself moved to actually try it as you read -- the stories in the books pull you to actually want to see things from Fred’s perspective, and you begin spontaneously picturing how he might be feeling.  You also find different sentence-stems in your head for how you might start the conversation -- sentence-stems seeded, in part, from the stories you read in the books -- and some of them seem promising.

Some of the books also contain statements that, as far as you can tell, are outright nonsense.  One suggests that the only reason you or Fred have any problems is that you weren’t praised enough as children.  You try on that perspective as well, but it feels yucky and nothing new clicks into place, and so you move along to the next part.

In this case, the books are acting, not as a source of trusted information, but as spur to your own process of anticipating, perceiving, desiring, and planning -- and, in this way, they are useful.

Example: Reading a good “woo” book

As I read “Bonds that Make Us Free”, I read many explicit statements I disagreed with (such as statements about a Christian God).  I took these in as poetry: I tried to imagine the world the author saw himself in, and to see myself in the same world, so as to have more access to the way he was parsing human phenomena.  

The book also contained many stories, all of which “rang true” in my inner simulator (they matched the world as I anticipated it), but which formed new patterns when placed next to each other.  I felt my intuitions update as I read -- I felt the stories take patterns I had previously half-seen, and pull them into full conscious awareness.

I left the book with a changed perception of how rationalization patterns affect close relationships, and with an increased ability to separate from my rationalizations and see the people close to me.  I left also with some new freedom from social shame.  These changes did not come about via trusting or deferring to the author; they came via trying on his perspective, and finding that pieces of the pattern he was pointing to “clicked”.

Try it!

One way to try this is to read a book.  But, since books’ usefulness is not just about their accuracy, it can *also* be surprisingly effective to just write your own book -- or, at least, to write 5 minutes of it.  You can think of this as a way of getting system 1 unstuck from its default pattern.[1]

Set a 5-minute timer (I):

Pick a soft skill, X, that you’d like to get better at.  Then, set a 5-minute timer (yes, an actual one — thinking for “about 5 minutes” doesn’t work nearly as well).  Spend those 5 minutes explaining to yourself, in writing, how to do X.  (For example, if you picked “networking at conferences”, you might spend the 5 minutes brainstorming on what the key tricky bits are, and on strategies for navigating them.  E.g., “I’m not sure how to start conversations.  So, at the conference, I can watch and see what sentence-stems other people use.  I can also just start with ‘Hi, my name is Anna’, or with ‘Oh, are you Dr. so-and-so?  I loved your paper on such-and-such’.  That starter sounds promising, actually; I should try to skim abstracts and Google at least a few papers before the conference….”)

(You might consider giving this a try right now.)

Set a 5-minute timer (II):

This exercise was created by Zak Vance, and is one of my favorites.  It’s seriously worth 5 minutes of your life to try this one out.  Pick, again, a soft skill, X, that you’d like to get better at (e.g., “networking at conferences”).  Now pick a different skill, Y, that you’re already highly skilled at (e.g., “programming” — Y can be a soft skill or any other sort).  Now, set a 5-minute timer, and spend 5 minutes explaining (in writing, or aloud to a friend) how skill X is really actually just the same as skill Y, in the sense that anyone who is fluent in skill Y already knows all they need to know to be good at X — they just need to apply their Y-skill to X.  (Your goal, as you do this, is to create a very short guide that enables anyone who already knows Y to hit the ground running with X.) You can see my example (taken from the seed “networking at conferences is really just the same skill as programming") in footnote [2].  

It’s important, of course, not to believe everything you generate in such an exercise — after all, it was secretly written by a *beginner* in Skill X.  But, again, you can use it to brainstorm ideas to try out, and to thereby get your system 1 intuitive search pattern out of any local optima you may be caught in.

Some good “soft skills” books to try reading:

Bonds that Make Us Free, by Terry Warner
An Open Heart, by the Dalai Lama
Feeling Good Together, by David Burns
Eat That Frog, by Brian Tracy
How to Win Friends and Influence People, by Dale Carnegie
Self-therapy, by Jay Earley 
Focusing, by Eugene Gendlin (the audiobook is much better than the paper, IMO)
The Core LW Sequences, by Eliezer.

You might also just try going to the bookstore and locating a new “soft skills” book, by scanning through a few and seeing if any “speak to you”.  Books of the relevant sort can be found in the business, self-help, continental philosophy, and spirituality sections, as well as in sections focused on particular soft skills such as writing or problem-solving.

As you read:

As you read a soft skills book (or the results of your own 5-minute timers), you may wish to ask about each paragraph:

  • Is it true?  Can I use it to update my explicit model of Skill X?  (Yes, explicit models are still useful!)
  • Even if it’s false -- is there something near it that is true?
  • If I free associate from here, do any of my past experiences click into a new focus?  
  • What feels fruitful/interesting about this idea?  About ideas I can free associate to from here?  Does this improve my implicit space of hypotheses?
  • As I read this, can I visualize myself carrying out soft skill X in a usefully different way?  Do I find my intuitions changing?  
  • How can I climb as thoroughly as possible into the frame, feeling, groove, or worldview that generated this book?  And once I do that, does anything new click into focus? 

Do consider adding your favorite books, or book-reading strategies, in the comments!


[1] System 1 stuckness of this sort seems to be almost ubiquitous. For example, I type and drive quite a bit, but my typing and driving skill are pretty similar to what they were several years ago; my conversational skills improve more, but they seem to have some of this same "trying the same things again and again" flavor. In this context, even random noise seems helpful to jumpstart learning.

("Learned blankness" seems related here as well; one somehow gets stuck in an ontology; the goal of soft skills books is often to help a person jumpstart out of it.)

[2] The example I generated from this seed, in 90 seconds:

Networking at conferences is really just the same skill as programming.  The three programmers’ virtues of Laziness, Impatience, and Hubris will take you all the way there, as will basic principles such as analytic thinking and code reuse. 

Re: Laziness: Many folks who set out to network at conferences work hard, and push themselves to talk to lots of people.  But in fact, some conversations are predictably much higher value than others.  Be Lazy: plan a bit during your plane-ride to the conference (or during a boring talk!), so that you know *who* you’re hoping to have *what kind* of conversations with, and can position yourself accordingly.   

Also, when you approach a new person, see it as an opportunity to *practice* and *debug* your greeting, as a re-usable code module, instead of treating it as a one-off task to separately sweat about and make throw-away code for.  And, afterward, review it briefly in your mind — and see if you can create a way to refine it. 

You might also look for “test cases”, much as you would when debugging — you might approach people who you already know by reputation, or who a friend of yours has secretly already approached, and see if your interaction pattern is similar to what you’ve heard.  You can use “test cases” of this sort to find out how your results compare to others, and to gain valuable info for debugging your own routines. 

Note that, to show a real example, the text above is the first thing I generated when I gave myself 90 seconds to type from that seed -- it is a typical case, not a selected case -- and certainly not a vetted guide to networking. Nevertheless, you can perhaps see how this sort of rationalization(!) process might be helpful to seed learning.


New Comment
29 comments, sorted by Click to highlight new comments since: Today at 11:58 PM

I'd nominate C.S. Lewis's The Screwtape Letters as a soft skills book, that gives your System 1 a lot of vivid, specific ways human thinking goes wrong to chew on and find ways out of.

One of my favorite examples is this passage which made a big impression on my System 1 about things that do bore me but I nonetheless get sucked into. (Context note: Screwtape Letters is written as a series of letters from a senior devil to a junior tempter about how to lead human lives astray)

You will find that anything or nothing is sufficient to attract his wandering attention. You no longer need a good book, which he really likes, to keep him from his prayers or his work or his sleep; a column of advertisements in yesterday's paper will do. You can make him waste his time not only in conversation he enjoys with people whom he likes, but also in conversations with those he cares nothing about, on subjects that bore him. You can make him do nothing at all for long periods. You can keep him up late at night, not roistering, but staring at a dead fire in a cold room. All the healthy and outgoing activities which we want him to avoid can be inhibited and nothing given in return, so that at last he may say...'I now see that I spent most my life in doing neither what I ought nor what I liked.'

For me, thinking of this passage (and the horror of coming to say "I now see that I have spent most of my life in doing neither what I ought nor what I liked") helps me feel very motivated to snap out of boring, sterile activities. It makes the consequences clearer, and, it gives the action of closing the site/app/etc a sense of thrilling defiance.

Oh, I like this post! Thank you for writing it!

> Do consider adding your favorite books, or book-reading strategies, in the comments!

Here are some book-reading strategies I've found useful:

  • When I feel angry, offended, frustrated, or some other emotion that seems to want me to stop reading or to not let the words get all the way in, I ask myself what I want from the book/author that I'm not getting. For example, while reading Nonviolent Communication, I felt something like angry whenever there was a sentence I didn't totally agree with (which was most of the sentences). Turns out I wanted a trusted adviser, someone who would guide me in my studies so I didn't have to create all of the structure for my education myself. While that desire was active and pointed at whatever I was reading, obviously false statements were quite painful. When I recognized that this was going on and that I would not find such an adviser in Marshall Rosenberg, I was able to receive whatever he had to say in a much more opportunistic, "let's see what I *can* do with this information" sort of way.
  • In addition to trying on the perspective of the author, I also like to try on the perspective of the ideal reader: I pretend that even though I semi-randomly selected this book off of a shelf, it was secretly written to me in particular. The author knows all about me and and is actually extremely wise, and these are precisely the things I need to hear to solve my most pressing problems. While occupying that perspective, I ask myself what it is I need to hear from this book, and what's true about me such that I need to hear it.
  • Whenever something sounds trite or vapid and yet is emphasized by the author, or by large swaths of society in general, I try re-reading it with the assumption that *I'm* the one who's been trite and vapid up to this point, and now I have an opportunity to learn what is life-changingly important in this phrase/paragraph/book.
  • I try to be brash and selfish when I read. I try to have as many valuable insights as possible without any concern for whether the author meant anything along those lines. Sometimes this feels rude, and I find myself being pulled into a dull, studious, subservient sort of mental space. It's those times when I remind myself to be brash and selfish. (I also use this reminder when I encounter a vapid thing and need find what's life-changingly important about it.)
  • I highlight things and write in my books. Rather than highlighting things that seem "important" (or that could conceivably appear on a test), I mainly highlight things I have emotional responses to. Sometimes I use several colors of highligher to track several kinds of emotions (especially confusion, desire/curiosity, and satisfaction/insight). What I write in the margins is an attempt to give voice to the emotion. I think that most of what this does is keep the book in constant dialogue with System 1.

Now I don't have to write the LW article on this that I've been meaning to write for years. Thanks! ;-) (I might still write it, but now I don't have to.)

It's funny that when I read Anna's article I reach to tag PJEby (sadly nonexistent LW functionality). I guess my inner simulator is working.

I'm not sure that "read a book" is the best meta-strategy for acquiring soft skills. Sign up for a course might be a much better medium. Especially for people who do have money.

I'm not sure that "read a book" is the best meta-strategy for acquiring soft skills. Sign up for a course might be a much better medium.

Depends on the person. I've noticed that some people acquire knowledge much better when they are taught, and some -- when they learn by themselves.

Neither of those seem to me optimized by "read a book".

"Learn by yourself" is often, but not always, optimized by "read a book".

When it comes to learning social skills, interactions with other human beings is often central. You don't have that with a book. It's useful to have a trusted environment that's supportive.


Books are simply the best way to learn.

Be sure to read a good book though, because otherwise you can esaily lose motivation.

Books are simply the best way to learn.

This is a vast overgeneralization. Note for example that many books are poorly written, and some subjects don't have good books. Moreover, different people prefer absorbing information in different fashions. I'm pretty heavily on the books-yay end of things, but I'm not under an illusion that books are magically good in and of themselves.


Note for example that many books are poorly written

True. That means you just have to take the right book.

and some subjects don't have good books.

Not really the fault of books now.

Moreover, different people prefer absorbing information in different fashions.

Acceptable thing to say but an rather vague one as "preference" can easily be like that stupid "i has add lol cant read" (an rather extreme one, but humans are not automatically strategic and all that I guess), but admitting that good books has and always been a standard for a wealth of great information is not an understatement.

Most of this is less about books but rather about the information (utility). The only thing that matters is that they get the best bang for their buck. And for that.. books. Correlation, casuation.. eh.. let's keep it simple.

Soft skills are hard. I'm extremely good at learning soft skills. I try as hard as I can to teach one whenever someone desires one, but they're damn near impossible to teach. I've thus far only been successful at teaching someone how to do something related, in a way that their mind can grasp and run with, so that a few years later they will have developed the soft skill "on their own" (At which point, they will be doing exactly as I said exactly how I told them to, and they will tell me about this cool new thing they figured out how to do, and question why I didn't ever tell them to do it that way).

So rather than teach how to learn soft skills, I'm going to describe a few of the most useful soft skills I have.

  • Maximize information input in a way that maximizes timeless usability. It works like this: prioritize observations in opposite order. The observation with the highest priority is the observation that is most distant from the desired observation while still being distantly related. The observation with the lowest priority is the observation that was first intended. For example: when performing a scientific experiment, the lowest-priority observation is the result of the experiment. Literally everything you could possibly observe related to the experiment is of higher priority than the direct result of the experiment. The reason for this is that distant observations have a tendency to be more useful for the future than direct observations.
  • Minimize friction to maximize usability at detail. When welding, move the welding rod and heating device at the exact rate that keeps the rod molten, and the welding pieces not-molten. When using a common carbon fiber dremel cutting blade on metal, move the blade through the metal at the exact rate that just barely melts the metal in the direction of the cut. When hammering a nail, make the nail move at a consistent rate that is fast enough to slide past the wood, but slow enough to not split the wood (this is like air resistance at sub-sonic vs super-sonic speeds). When talking to people (assuming you want to maximize usability of the other person's mind), make statements and ask questions at a rate that doesn't overload the person's mind. When drawing a picture with a pencil, move the pencil across the medium at a rate that glides across the medium without catching on the medium.
  • Inference jump to maximize efficiency of skill gain. When developing a theory from a conceptual base, assume a related postulate to be true so that you can obtain more data from experiments. The postulate is most likely not true, but that's not important. Experimenting under the assumption that a postulate is true is very useful in determining why said postulate is not true. Believe nothing. Treat all beliefs and assumed truths as inference jumps, because doing so maximizes their use. In order to make inference jumps with highest usability, all previously beliefs should be considered irrelevant. Failing that, consider the possibility that what you know to be wrong, or believe to be wrong, is in fact true, and use that as an inference jump.

I personally think the conceptual base of Less Wrong is contrary to efficient soft skill development. However, I think the conceptual base of Less Wrong is a potentially good platform to use to begin soft skill development. The most efficient way I can think to do so is to learn to use the conceptual base of Less Wrong, then make the inference jump that everything you've learned from Less Wrong is uselessly inefficient. This is the pattern I see everywhere people strictly adhere to a conceptual base. It's almost always a good platform to expand further from, provided the expansion stops using the platform as soon as possible.

Could you give examples of how those three points play out in social interactions?

Some of the books also contain statements that, as far as you can tell, are outright nonsense. One suggests that the only reason you or Fred have any problems is that you weren’t praised enough as children. You try on that perspective as well, but it feels yucky and nothing new clicks into place, and so you move along to the next part.

I don't think the problem with that statement is that it's outright nonsense. The problem is rather that it isn't actionable and doesn't tell you something you can do about the situation.

Well not in current situation, but if you believe it then you that would teach you to avoid dealing with people who were not praised enough as children.

Well not in current situation, but if you believe it then you that would teach you to avoid dealing with people who were not praised enough as children.

I normally don't have good information about the childhood of people I'm interacting with and the amount to which they were praised.

Very nice post. A few recommendations of books that I found useful to read in this way:

  • Never Eat Alone (Keith Ferrazzi)
  • Pitch Anything (Oren Klaff)
  • The Charisma Myth (Olivia Fox Cabane)
  • The Flinch (Julien Smith)
  • The Trusted Advisor (David H Maister)

ETA formatting

I will second The Charisma Myth and The Flinch. I have mixed feelings about Never Eat Alone, but if you live in a large city/on a college campus, Ferrazzi's advice is likely worth reading.


First of all,I'm french so please forgive my languages mistakes Hi, thanks a lot for this article, I'd never heard about the inner simulator's notion before, now I've a word to put on it :), in a second part, do you think these skills are still semi conscious once they're learned?don't they pass on a conscious level, in my opinion once a skills or more generally a knowledge is learned, it pass from a semi conscious level to another which is more conscious if not totally. take the example of the maths problem, once the solving method is known (for a particular problem) you have to use this method in a conscious way. However it's not unconscious either, even if you spend a half life to study, this knowledge will not come out automatically, you have to found it in your memories which is a conscious mechanism. thanks for reading and sorry if it's a bit confused, in my native language it's easier :)



How to Win Friends and Influence People, by Dale Carnegie

Will try.

But I will admit I'm not sure positively people will react if they see it on my bookshelf..

Solution: get an eBook edition or get a used hardcover and take off the dust jacket.

Also, it's not going to make you meaner than the general population. It just teaches you how to do consciously what some people can do unconsciously.

If it's morally good for me, as an autistic person, to improve my social/manipulation skills such that they're closer to the average NT, then why would it be immoral for you to improve your social skills? Unless there's some morally optimal level of social skills that is quite conveniently the level of the average person, this seems strange.

Solution: get an eBook edition or get a used hardcover and take off the dust jacket.

I enjoyed it a lot as an Audiobook.

I wish it came with an explanation what _exactly_ Impatience and Hybris virtues entail (given that both are generally described as non-virtues but I do seem to have the feeling that they can be good; same works for Laziness, but here I believe I have better understanding already).

I find that a fiction book that speaks to you is also a really great way of jump-starting experiments in your Inner Simulator. I usually find such books by browsing a book store and picking out (not necessarily objectively that good) stuff that speaks to me rather than working from a crowd-sourced list.


Here is an interesting flowcharts related to depression. Not sure how to interpret them, but I suppose flowcharting mental health triggers and actions could be very useful.

I'm glad you're talking about this because I spent a good deal of Highschool (age about 15-18) working on how I related with people and social situations and now I do it with aplomb, grace, and ease. Very good!

Your inner simulator is probably more accurate than your explicit models in domains where you have a rich experience base, such as social phenomena, and day-to-day physical phenomena.

Not true. At least not likely. If it were so, then most people would pick up soft skills in a rather automatic fashion without the need for too much conscious effort (actually, any conscious effort, since if the inner simulator is more accurate, your intuitions should lead you on the correct/correct-ish path. Ceterus Paribus, this should hold true for the majority of people (i would expect around more than 90%)

Also, I have a more general comment. A large part of the strategy advocatde here seems (to me) to be divided into rough categories: Keep trying stuff and keep reading stuff. You seem to indicate that learning hard facts should be discarded when you find them to be wrong, whereas learning soft skills involves some amount of reading "wrong ideas/facts/information". I get the feeling that you're keeping a double standard here. In the sense, that the timeline for learning a hard fact seems to be almost instantaneous while soft skills seem to be a little more long term (at least that's the sense I got from reading it, as if it was implied). To further illustrate this point, consider the following thought experiment:

  1. You read in a book that gravity doesn't work on cell phones. You drop your phone and find it falls. You've learnt that gravity does in fact work.

  2. You read that flattering people doesn't help. You avoid flattering people and then suddenly notice that your competitors who flattered their bosses/colleagues/friends have advanced their careers/social position.

See, what I'm getting at?

People's inner simulator is almost always more accurate than their explicit models. It's just less precise. The thing about your statement of [if it were more accurate, people would be using it more, and be successful at more things] requires a few initial assumptions to be true. The first is that people are able to use their inner simulator on purpose, which is usually not the case from my observation. The second is that people are able and willing to take the path indicated by their inner simulator when it contrasts with their explicit models, which is also usually not the case. Then there is an additional factor where the inner simulator can (and often does) output a result that is contrary to the facade a person is trying to keep up for social reasons, which provides an additional impetus to reject the inner simulator (even if it is indeed more accurate, and dropping the facade would be the best way to produce the desired long-term result).

In regards to learning/picking up soft skills. People do indeed pick up soft skills automatically. Soft skills are notoriously difficult to teach. Most people that learn a soft skill learn it on their own, through a long-term automatic process, after taking in information that sparks a complicated processing of that information. I've never met someone who learned a soft skill without it being at least partially automatic. There is additional complexity in this in that people are "born" with an "affinity" for certain types of soft skills, and have an extreme amount of trouble learning any soft skills outside of their affinity range.

The actual reason learning soft skills involves reading some amount of "wrong ideas/facts/information" is because the processing of information for soft skill use is different from the processing of information for explicit model use. Or more accurately, the mind has multiple ways of processing information, and is not at all limited to a neuron-only model. The method of processing information for soft skill use is more of a resonance/antenna model, which benefits from additional points of information no matter how wrong they are (as long as they are at least distantly related to something that has at least an iota of truth), up to a limit of what the resonance chamber can hold coherently.

Last, your two points in that thought experiment are explicit model only, and do not relate to soft skill learning or use. The difference would be that you have an explicit use in mind for each experiment. A soft skill necessarily has a soft use. The difficulty in translation of "flattering people doesn't help" is that such things are actually a general-scope statement, and they have to be for soft skill learning and use. So your thought experiment is akin to: "You read that white men can't jump. You ask a white man to jump, and he is successful." You're combining a general-scope statement with an explicit-scope experiment.

New to LessWrong?