This ebook is kind of dopey, but it's one of the few resources I've seen where someone who's reasonably good at learning stuff tries to dissect and communicate the mental mechanisms they use for learning:

http://www.scotthyoung.com/blog/Programs/HolisticLearningEBook.pdf

Here's a quick summary.

  • You can learn things faster and better by improving the strategies you use for learning stuff.
  • "Holistic" learning is opposed to "rote" learning.  Holistic learners make lots of connections between different things they learn, and between things they learn and things that are personally relevant to them.  An example might be this diagram of various concepts in electrostatics, which I no longer know how to interpret.  Another example might be me remembering about that diagram when reading the book.
  • Holistic learners understand concepts in many different ways in order to really "get" them.  They focus on building mental models instead of memorizing facts or procedures.
  • If you understand a body of knowledge well enough, and forget a specific thing, you should be able to reconstruct your understanding of it based on related things you understand.
  • The book refers to a "model" as something specific you understand particularly well that you can explain other things in terms of.  For example, your "model" of a subspace (in linear algebra) might be a plane cutting through 3d space.  Not all subspaces are planes, but thinking of a plane could be a way to quickly preload a bunch of relevant concepts in to your head.
  • To learn holistically:
    • "Visceralize" concepts by summarizing them with a specific image, sound, feeling, and/or texture.  Example: when learning programming, think of an array as a bunch of colored cubes suspended along a cord.
    • Use metaphors to understand things better.  Whenenever you learn something new, try to figure out what it reminds you of.  If it's something from a totally unrelated domain, that's great.
    • Explore your understanding network, ideally by solving problems, in order to fix glitches in your understanding and refresh it.
  • Holistic learning works great for some subjects, like science and math, but it's not as good for others, like history and law.  It also helps less with concrete skills, like playing golf.

The author sells various information & coaching products in this vein, but as far as I can tell the ebook I linked to is the only free one: http://www.scotthyoung.com/lmslvidcourse/2.html.  (If anyone pays for any of these, they should summarize them (to understand them better) and post the summaries to LW ;].) I'm definitely interested in hearing about other resources people know of on the mechanics of learning.


Someone once told me that if you're a grad student studying under a Nobel laureate, you're much more likely to later win the Nobel yourself.  (I just searched the internet for evidence regarding this claim and couldn't find any, so I'm now less confident in it.)  This claim suggests that doing good research is learnable.

The person who told me this thought these research skills couldn't be described with words, and could only be transmitted through actual research partnerships.  I think it's more likely that they can be described with words, but no Nobel laureate has bothered to sit down and write a book called "How I Do Research".  (Please leave a comment if you know of a book like this!)

Even if your fluid intelligence is static and difficult to improve, that doesn't prevent you from improving the mental algorithms and habits you use to accomplish tasks.

New Comment
16 comments, sorted by Click to highlight new comments since:
[-]gwern350

Someone once told me that if you're a grad student studying under a Nobel laureate, you're much more likely to later win the Nobel yourself. (I just searched the internet for evidence regarding this claim and couldn't find any, so I'm now less confident in it.)

Challenge accepted. Begun 8:40PM, finished 8:47PM. Damn, I'm good. Citation:

In a study conducted with 92 American winners of the Nobel Prize, Zuckerman (1977) discovered that 48 of them had worked as graduate students or assistants with professors who were themselves Nobel Prize award-winners. As pointed out by Zuckerman (1977), the fact that 11 Nobel prizewinners have had the great physicist Rutherford as a mentor is an example of just how significant a good mentor can be during one’s studies and training. It then appears that most eminent scientists did have people to stimulate them during their childhood and mentor(s) during their studies. But, what exactly is the nature of these people’s contribution.

  • Zuckerman, H. (1977). Scientific Elite: Nobel Laureates in the United States. New York: Free Press.

GS lists >900 citations of this book, so there may well be additional or followup studies covering the 40 years since. Or, also relevant is "Zuckerman, H. (1983). The scientific elite: Nobel laureates’ mutual influences. In R. S. Albert (Ed.), Genius and eminence (pp. 241-252). New York: Pergamon Press", and "Zuckerman H. "Sociology of Nobel Prizes", Scientific American 217 (5): 25& 1967."

How did I find it? A few wasted searches like 'factor predicting Nobel prize' or 'Nobel prize graduate student' in Google Scholar, until I search for 'nobel laureate "graduate student"'; the second hit was a citation, which is a little unusual for Google Scholar and meant it was important, and it had the critical word mutual in it - simultaneous partners in Nobel work is somewhat rare, but temporally separated teams don't work for prizes, and I suspected that it was exactly what I was looking for. Googling the title, I soon found a PDF like http://www.pages.usherbrooke.ca/rviau/articles/principales_communication/eminent_scientists_demotivation_in_school.pdf which confirmed it (and is interesting in its own right as a contribution to the Conscientious vs IQ question).

no Nobel laureate has bothered to sit down and write a book called "How I Do Research". (Please leave a comment if you know of a book like this!)

Arguably, Feynman's various books and collections constitute such a thing. Alternately, Turing Award winner Richard Hamming wrote The Art of Doing Science and Engineering: Learning to Learn, which is just that. (I was disappointed and thought his talk was much better.)

[-]tgb50

Is this the talk that you're referring to?

Yes, although I find that version unpleasantly narrow.

Nice. This seems like a topic worthy of further exploration.

[-]satt20

For instance, I was already aware of Nobelists begetting Nobelists (though I didn't know it was so common!), but I've no clue how much of the correlation's actually a chose-the-right-field effect rather than a chose-the-right-mentor effect. It could be that most of the gains from having a Nobel-winning mentor come from riding the same bandwagon as them, rather than the mentor being especially stimulating. Zuckerman might have explored this issue but I don't know her work well enough to say.

[-]wattsd100

Hamming's "You and Your Research" and Herbert Simon's "The Scientist as Problem Solver" are good "How I do research" papers. Hamming's paper was described in the other comments. Simon won both a Turing award and a Nobel prize.

Simon's paper is here: http://repository.cmu.edu/cgi/viewcontent.cgi?article=1425&context=psychology Hamming's: http://www.cs.virginia.edu/~robins/YouAndYourResearch.html

Thanks for the links, but I didn't care for Simon's paper at all. I recall Hamming's inspiring me for a few hours at least - perhaps just making up the time spent reading it :)

Simon's writing style seems a little strange to me for what its worth...

There are few others who have worked with with him and described their impressions of how he worked. Those might be more readable, but Hamming's lecture/paper is hard to beat in my opinion.

http://web.cs.dal.ca/~eem/gradResources/HerbertSimon.pdf http://www.isle.org/~langley/papers/has.essay.pdf

I attempted to summarize the three papers and incorporate a few other things a while ago, inspired in part by a post by Cal Newport of StudyHacks on the methods of Feynman and a few others. Incidentally, Cal has colloborated in the past with the author of the Holistic Learning ebook in the OP.

Cal's post: http://calnewport.com/blog/2012/06/18/impact-algorithms-strategies-remarkable-people-use-to-accomplish-remarkable-things/ My summary of Simon's Methods: https://sites.google.com/site/wattsd/simplesimon

The summary is still rough and incomplete, so the sources might be more interesting/useful.

Yes, I got a very strange vibe reading that Simon paper, as funny as parts of it (like the concluding advice on how to make good use of your friends) were, and as seminal a figure as he has been in AI and related fields.

After thinking about it, I think the issue is that Simon is coming from the Good Old Fashioned AI point of view of messing around with random Lisp code without any kind of principled background such as statistical models, and this leads to a kind of subtle semantic drift on all sorts of points and vocabulary - a kind of Uncanny Valley effect. Just similar enough to disturb one.

That's some nice doing-research porn. I liked Cal's summary. Too bad he PC comment wars over a joke about Feynman "chasing skirt".

Also recommended: Heilmeier's Catechism.

That's some nice doing-research porn. I liked Cal's summary. Too bad he PC comment wars over a joke about Feynman "chasing skirt".

Also recommended: http://en.wikipedia.org/wiki/George_H._Heilmeier#Heilmeier.27s_Catechism.

[This comment is no longer endorsed by its author]Reply

This claim suggests that doing good research is learnable.

Alternatively, people who are better at identifying promising researchers, mentors, and topics are also better at winning Nobel prizes.

Thanks for the summary!

Someone once told me that if you're a grad student studying under a Nobel laureate, you're much more likely to later win the Nobel yourself.

I would guess that the first order selection bias is from being accepted by the Nobel Laureate - I assume a Nobel Laureate has a good eye and can be picky.

my personal summary:

Building constructs: within an area link all ideas as much as possible (biology) Building models: Simplify concepts - abstract from them to create something that you can use (evolution by natural selection) Highways: Linking constructs (biology to economics trough evolutionary economics)

Acquire. Test - Have I seen/listened to the idea before? Understand. Test - Do I get (at a surface level) what this idea means? Explore. Test - Do I understand where this idea comes from, what it is related to and what outside ideas can be connected with it? Debug. Test - Have I removed inappropriate links between this idea and others?Have I removed false conclusions based on connections that donít actually exist? Apply. Test - Have I used this idea in my practical life?

A) Acquiring Ideas 1) Speed Reading 2) Flow-Based Notetaking B) Linking Ideas : FOR DIFFICULT OR CRITICAL INFORMATION 1) Metaphor - used to relate unfamiliar to familiar ideas. (process: ask for it. take the first that comes. test it) 2) Visceralization - used to translate information to preferred format (process: identify concept, pick mental image, does it move trough time or is statis?, add more sensations and emotional impacts to your image) 3) Diagraming C) Handling the Arbitrary 1) Linking 2) Pegging 3) Information Compression -> Notes compression (write down major ideas and all related ideas. rewrite) D) Extending Ideas 1) Practical Usage - Look for ways to applu ideas 2) Model Debugging - typos are not bugs, use very different questions to do a shotgun debug, spread practice times out 3) Project-Based Learning - 1-3 month project that uses knowledge you want and don't have