Note: This was originally posted in the discussion area, but motions to move it to the top level were made.

My own desire to improve my rationality coupled with some posts criticizing LessWrong not too long ago led to an idea. For reference, the posts I mean are these:

The comment stream on that last link brought up whether or not LessWrong is fulfilling EY's vision of a "rationality dojo" of sorts (along with what I assume is the idea that LessWong is supposed to be fulfilling that vision).
Now, I'm not the one to evaluate whether LessWrong is or is not a true "rationality dojo" (not to mention reducing the problem as not to argue over semantics), but I do have an idea for how it could perhaps do so, and an idea I personally find quite exciting.
The idea is simple:  that of a LessWrong workbook.
What does that mean? For one, it means an awful lot of effort, primarily at distillation. I quite enjoy reading posts and appreciate the overall prose style used here. A lot of posts are easy reading, contain nice anecdotes, hypothetical real-world scenarios, and the like. Distillation would entail extraction of the "nuggets" that an aspiring rationalist should know. Teach me rationality in bullet points, flash card modules, and other mini-homework-sized methods (which I'll get to).
I see the above as reducing to some form of "Rationality Overview." What is rationality? What's the point? What is the end result? What are the typical tools used to be rational? Finally, perhaps, what does the life of a rationalist look like? Perhaps pages containing some helpful "mantras" or questions to ask one's self periodically, such as during some period of daily meditation? Inspirational/helpful prose like the  Twelve Virtues,  The Simple Truth, and the Litanies of  Gendlin  and  Tarski.
Secondly, I envision a literal "workbook" type of section. I was intrigued by a  comment  on my  first post, in which  JenniferRM  wrote [in discussing my  ongoing de-conversion]:
Unfortunately, I don't know of  any resources to help people traverse the path you're facing in a series of small safe steps.
Why not? Shouldn't there be some example of guidelines for evaluating beliefs and attempting to rationally adjust them according to new evidence, though processes, or trustworthy tools? Hence the birth of my workbook idea.
Most specifically, I envision the following:
  • Most interesting to me was the idea of some form of "rationality comb." An iterative evaluation process. Again, I hardly consider myself the one to design this, but perhaps something like:
    • Take 5 minutes and brainstorm about the beliefs you think affect your actions the most
    • Focus on the first belief, set(1):belief(1)
    • Can you recall how you came to hold this belief?
    • What are some common alternative views to your belief?
    • Do you think you could provide  testable  justification for your current belief over the above alternatives?
    • If not, can you  imagine  leaving your belief for one of the alternatives?
    • And so on...
    • Then repeat with set(1):belief(2). When set(1):belief(n) is finished... re-brainstorm for 5min to come up with set(2):belief(1)...belief(n).
  • A series of "homework" problems on Bayesian Probability, perhaps including EY's  tutorial  and other helpful material.
  • Brain teasers or similar items to focus on attentiveness to details, weighing evidence, knowing the limits of what you can know given certain information, etc. I think LW has already provided some good  examples  of neat things like this (even if they would require refinement).
  • Questions that intentionally try to deceive the reader with some form of  fallacy  or  bias
  • Tutorials on how to have rational discussions, rules of engagement,  reaching a mutual conclusion, etc.
This is just a stab at what the "meat" of the workbook could contain.
My primary interest is in the first bullet: a step-by-step guide to examining one's life. I think it needs to be iterative, since not all beliefs in need of rational attention will be apparent on the first pass, and each subsequent step needs to be able to be done in bite sized chunks. Steps should have concrete starting and ending points that seem achievable to promote continuing effort and a sense of accomplishment. I left out quite a few steps... suggestions could be made as to what kind of information should be sought in attempt to confirm or disconfirm one's belief/hypothesis (and being cautious if you find yourself looking for  expensive evidence in favor vs. cheap evidence against), how to know be wary of hypotheses that  explain too much, and so on.
The point is to have a handy reference guide to not only becoming a rationalist, but also to have a concrete guide to applying those skills to your own life. Steps summarized from trusted methods help avoid pitfalls, help us learn from what's already documented, and (I would suspect) help us avoid subconscious biases or shields we might be tempted to apply to cherished beliefs if left to our own devices/methods.
By this time, I'm hoping I have the point across. I've tried to link heavily to show that I'm not really proposing new material -- the content seems to be here. My idea consists of is  rearranging this content into a compact "dojo-like" format.
So, what are your thoughts? Is this done elsewhere already (if so, my apologies for wasting your time)? Does this sound helpful? Is it plausible/feasible? Would  one workbook be able to apply to varied personalities/learning methods? And what are  your thoughts (if you like the idea) on what it would contain?
New Comment
26 comments, sorted by Click to highlight new comments since:

One nice way to do this might be to wait until EY releases The Art of Rationality, and then the LW community could develop a workbook of exercises and sample problems to go along with it. In many cases, the sample problems could be borrowed from textbooks on, for example, probability theory.

Very good suggestion. I'd read him mention writing a book. Maybe if he sees this post he'll be reminded that young Padawans are waiting in eager anticipation. And, yes, if LW-ers could put together a nice workbook to go along with it, that would be fantastic.

I suspect that if the workbook came included with the book, a lot fewer people would buy it though. Possibly it could contain the address at which you could access the workbook online, and maybe print it out, although even such a trivial inconvenience could still be a barrier to a lot of people.

I think that it may be worth moving this article to the main page.

I agree re. an included workbook and to clarify, when I said "go along with it" I meant "related/parallel content," not "physically attached."

I think a downloadable PDF (easily accessible, easily revised/updated) for little or no cost would be optimal. I'd be willing to pay for it to cover the time and effort invested by other LWers. I'd also volunteer my time for free to the project as an organizer/assembler to assist those willing to supply/decide the actual content. I have a decent amount of experience with writing technical manuals and a fair amount of experience using LaTeX, which would lend itself well to the writing of an organized, footnoted workbook type of publication.

I think that it may be worth moving this article to the main page.

If perhaps two more agree with this, I'll do it... otherwise forcefully suggest it and I'll do it based on your recommendation alone. Again, I really have no idea what constitutes a top-level-worthy post!

Please move this to the main page. I'd love to see it done; I don't know if there's anything I could contribute in the way of time and knowledge.

I also second (third?) the suggestion for trying to make it a companion to The Art of Rationality. Desrtopa, why do you think fewer people would buy it if they came together? Is it because books with workbooks attached look intimidating?

Is it because books with workbooks attached look intimidating?

Yes, that and it conveys the impression that reading it is work,

I liked the idea of a separate, electronic format for several reasons:

  • Those who don't buy the book can still have the assistance of a practical guide (though hopefully most would buy the book...)
  • I see the two as serving different purposes. While I don't know what Eliezer's format/content will be, perhaps the book could be viewed as sort of reference "textbook" for education, with the workbook serving as a guide to actually apply such information to your own life (just learned about bias x? Great, here's how to identify it in your life)
  • Lastly, I like an electronic format simply for the accessibility of it (have internet? Super -- download it and continue where you left off) as well as the ease of updating. One great usefulness of updates could be to make various versions with different problem sets -- once you run out, the next edition will replenish your supply.

Thirded; move it to main.


I think that it may be worth moving this article to the main page.

If perhaps two more agree with this, I'll do it...

Move this to the main page.

Possibly it could contain the address at which you could access the workbook online, and maybe print it out, although even such a trivial inconvenience could still be a barrier to a lot of people.

Bothering to access the workbook online could be a practice test of instrumental rationality in and of itself.

And if you explicitly frame it as such, the entire book has primed them to pass this particular test.


I know there had been talk at the SIAI house about debiasing workshops (built around flashcards, as I was told). Does anyone know the status of that project?

There was a thing where we gave 90% confidence intervals for questions off of little cards; I think that was done in the time going up to Burning Man

I have been asking around about this. No one seems to have heard of it. Are you perhaps thinking of Divia's anki cardsets?


Someone mentioned the idea at Burning Man. Possibly Jasen? I think it was mentioned at Future Camp. I only know that it was suggested that SIAI was preparing (two-week?) workshops for this Summer, and was anticipating many more attendees for that than Visiting Fellows.

It might have been Anna's habit cards.


Will_Newsome has already done some distillation of the sequences.

Fantastic -- thanks for providing that link!


So did this ever get off the ground? Seems like a fantastic project, and if not, I'd like to push to make it happen.

I imagine something like Lovasz' Combinatorial Problems and Exercises, maybe also in a format amenable to spaced repetition. How many of us are qualified to contribute though? I certainly am not.

We could also write separate exercise books for different topics, and then each of us could specialize and sort of distributed pair-teach-learn in Bittorrent style. For example, Patrick has been helping me a lot (and I think he enjoys teaching me). I'd obviously like to return the favor, perhaps in another subject.

I'd been tossing around the idea of a popular "How to Improve Your Life" sort of site. The user answers questions regarding what they are disappointed in, some 'dance around the bush' type questions to work around people's self-bias and questions to judge education level on various necessary topics.

The system would then weight the answers and calculate what improvements would have the highest cost/benefit. My assumption was that most people would suffer from too little time, irrational beliefs or health issues. A site is then suggested to support improvement.

There are tons of free sites for too little time (GTD), improving health (mostly calorie counters) and knowledge (spaced repetition sites, Khan Academy, etc). There is nothing I know of to step people through rationality. When I say 'step through', think an expert system similar to software 'wizards'.

I've been leaning, and this post helped shove me over, that helping address the lack of a rationality site is the more immediate good. (A rational/informed person wouldn't need a site to point out what areas need improvement, and then google it)


I plan on starting to outline a rationality specific version of the above and code it for a LAMP stack. Anyone else interested, or have a better suggestion? If no better suggestions, I'll start a sourceforge project.

Very interesting idea -- it would be great if some would help you with the questions and I agree that you may need some tricky ones to pull out the biases and actually target the most immediate areas for improvement. I like the idea!


I think this is a great idea, and I'm willing to offer time toward such a project.

I had a related, overly-ambitious LessWrong project idea (it contained a variant of this one). But I'll post that in a new thread, to keep this one focused.

I look forward to reading about your project.