The new wiki is off to a flying start - it's less than 15 hours old, and already it has over 20 articles and five authors. It's probably about time we worked out what it was for.

I created it because as things stand, I can't point my friends to Less Wrong and say "come and learn about rationality and take part in these fascinating and potentially important discussions!" The discussions we have here assume years of reading Overcoming Bias and close attention to what's been said there and here; it must be practically impenetrable to newcomers. So for me the primary goal is simply to provide a glossary, to give newcomers a fighting chance of understanding what on Earth we are talking about and why. I think it can do more than that, but before I come to that, let me say a little about what I think it's not for.

The way I would currently like to see it, the wiki is not there to duplicate what is already done elsewhere. So it's not a place for discussion - that's what this site is for, and the features to support discussion here are far stronger than they are there, what with voting, threading and so forth. By the same token, it's not a place to advance your ideas - it's better to do that here, where people can comment on them and where it's clearly tagged as the work of one author rather than some sort of collective conclusion.

I'd like to avoid duplication in other areas, too. Anything that can go in Wikipedia instead of our wiki should do: we will get better results if we and they are editing the same biography of Eliezer Yudkowsky, rather than creating a fork. To that end, I've created a {{wikilink}} template that can go at the top of an article, linking to the article with the same name in Wikipedia. Have a look at our current article on Newcomb's paradox - there is far more detail in the linked Wikipedia article, but there are some things we carry because they (rightly) won't: the sometimes non-standard vocabulary we tend to use around it (eg "Omega") and links to related articles in Overcoming Bias/Less Wrong on the subject, which they might not choose to keep (since Wikipedia is not a link farm).

Similarly, we don't want to provide our own index of heuristics and biases, since there's one on Wikipedia and another on the Psychology wiki, and most of what they lack on the subject we can fix there rather than trying to address by duplication.

It's often easier to say what a thing is not for than what it is for. What have I missed out here that we should be using the wiki for; am I right to discourage what I set out above; what else do we need to say about how best to use it? Because we could be discussing anything in a given week, but a wiki evolves more slowly, I'd like to hope that if in a year's time I meet someone who seems open to the ideas we discuss here and wants to learn more, it's the wiki I'd point them at rather than this website; it might eventually be the best starting point on how to become less wrong.


25 comments, sorted by Click to highlight new comments since: Today at 7:11 AM
New Comment

I'd like the wiki to do three things:

  1. Provide a good landing page for newcomers. Let newcomers see what we’re up to, why a person might care about it, and which past posts can help them with which of their own questions.
  2. Provide a good review, or summary of what has been discussed or figured out where, for current community members who’d like a better birds-eye picture of our project.
  3. Provide a list of "open problems" and "articles someone should write", together with a picture of how progress on those problems would contribute to LW's project, and an index of progress to date. So that would-be authors can see useful avenues to contribute, and so that some portion of LW posts can visibly contribute to a cumulative project, instead of being like a succession of randomly entertaining newspaper articles that no one much cares about afterwards. (I’m not hoping all LW posts will do this; but it would be nice if some portion would, and if LW readers build some cummulative competencies over time.)

I agree that discussion, and temporary or controversial content or scratchwork, should be at LW instead of at the wiki. I guess I also agree that completely standard, non-LW-flavored content should be put in Wikipedia, but I’m unclear on where the boundaries are: I’ve liked LW posts that mostly do summarize standard stuff, except the standard stuff they summarize is linked in to our specific goals and prior knowledge and open questions.

I put up a bad rough draft of a questions-first index, and began editing a page for one of the questions. I’m imagining each question on the initial list eventually leading to a list of progress to date (links to OB/LW posts, with sentence-long summaries) and open questions. It gives a better idea of what I’m imagining for the wiki. If anyone has time to look at my pages there and say whether that’s what the wiki should be or shouldn’t be, the feedback would be helpful.

Points 1 and 2 are very much what I have in mind. Sorry to say though, I don't think 3 is a good idea.

  • We can better make LW a cumulative project by working on what we have done, rather than what we haven't done. Similarly, we can avoid duplication with proper coverage of what we do have.
  • The best articles will come about because people feel the need to write them; I think high quality articles will rarely result from people reading a list of desired articles and deciding to work on them. Where a specific open question deserves more attention, there's nothing wrong with an article about it, as per Yvain's very popular article Why support the underdog?.

So I think that your current list would best live in your own personal namespace, as "User:AnnaSalmon/Open problems, and summarized progress to date‎" - thoughts?

I'm not really sure I follow you. Yes, we can avoid duplication with proper coverage of what we do have, but how is that a reason not to mention the things that still need to be explained? You say that we "can better make LW a cumulative project by working on what we have done, rather than what we haven't done", but why should those things be mutually exclusive? People work on the things which happen to interest them - a list may inspire a person with otherwise no inspiration to work on something on the list. The net result is more articles written, not less.

You say that "high quality articles will rarely result from people reading a list of desired articles and deciding to work on them", but provide little reason for why this would be so. Indeed, beginners may feel intimidated by having to come up with their own subjects and would probably prefer having a ready list of example topics they could work on. And even if what you're saying is true, that high quality articles will rarely result from people reading the list isn't the same as saying they never result from reading such a list - even a list that's of limited use is better than having no list at all.

OK, let me come at it from another angle. How are we going to agree as an editing community on what should be in a list of what hasn't been written?

We could use NPOV if necessary. E.g., we could note that "many have expressed an interest in techniques for reducing akrasia, and in gaining a better understanding of akrasia", perhaps with links to the relevant LW comments. Similarly, we could note that "there has been much discussion of anthropics/Newcomb/whatever, and most commenters seem to agree on (a), (b), and (c), but there remains quite a bit of controversy on how to coherently conceptualize (d)"...

I think the basic technique for agreeing about what should be on a wiki, is just to summarize the spread of disagreement when significant disagreement occurs on a particular point.

Somehow, it seems unlikely that "beginners" are going to contribute major new articles on LW. In order to get the right to post, they'll have to join the community and fit in well enough to get upvoted on their comments on other people's posts. Once they've done that, they'll be well on the way to understanding the jargon (even if there are holes in their understanding.) At that point, they'll know what areas are of interest, and if they have a contribution to make in an area we don't often talk about, finding it on a list of things we think are missing is less likely than noticing a lacuna that was invisible to everyone else.

I believe there would be value in a list of Contentious Subjects, but Articles to be Written will be unlikely to cover the new material we're not aware we're waiting for.

Um, well, hmm. Thanks for the input. I agree that good articles will sometimes come from people autonomously feeling the need to write them. And I'm not trying to have the community open problems list be the sole place from which good contributions spring. But as an intermittent posts author, I would find it helpful to have a better idea of what LW is trying to do, what loose corners people are puzzled by on which I might have thoughts, etc. And I was hoping some (though not all) authors or potential authors would feel the same.

A list in my personal namespace, representing only my own views on what LW has done and aims to do, wouldn't serve this purpose at all.

What do other people think about whether point 3 is a good idea? I'll insert a makeshift poll to gather votes, but comments with reasons would be good too.

Also, ciphergoth, in what sense will you be using your wikia administrator status? Are you advising against purpose 3 as one wiki contributor among many, helping to work out what our aims should be, or are you deciding it as our administrator?

Poll: Should aim 3 above (providing an [incomplete] list of open problems, and wanted articles, together with an indication of how those open problems would contribute to LW's overall project) be one of the aims of the LW wiki?

Vote this comment up if your impression is "Yes".

Hum. I voted this comment up to 2, but now I see that it's down to 1 again. People who disagree shouldn't vote this down but rather vote up the "no" comment, right?

Yes. The same happened to me, and I just assumed that the "poll" user had removed their own upvote after I voted to make voting zero-based, but it looks like we shouldn't trust the results of this poll. I don't think a poll is a good way to settle things like this anyway; it's better to hear what people think.

Vote this comment up if your impression is "No".

I realize that I was hardly the first or only person to suggest a wiki, but what I had in mind was something like the resources available at for people wanting to know about the scientific evidence for biological evolution and the nature of Creationism.

You don't find biographies of Richard Dawkins or even Charles Darwin there, and you certainly don't find the histories of minor footsoldiers in the Culture Wars. What you DO find is factual information, lists of claims, and arguments for/against those claims, complete with links to evidence and substantiation if needed.

That's the sort of thing I was hoping people would create, when they set out to make a rationality wiki.

I'm considering starting a style guide. I'll try to reflect current usage as best I can. Does anyone have any suggestions? Eliezer, I'm looking to you specifically as you have written the most.

Some points of style off the top of my head:

  • Less Wrong and Overcoming Bias are each two words.
  • Per ciphergoth's suggestion, we are sticking to the Wikipedia convention that article title should be all in lowercase, except for the initial letter and proper names. In the article body, should we do the same or use EY's more liberal cap style?
  • I'm half seriously considering to only cite OB/LW authors if they are not EY
  • Article references should be based on the References template. When creating an article, add {{subst:References}} to the bottom, save the article, and then fill in as needed.

Why subst the references template?

If someone were willing to make a template that we could pass references as parameters to, that would be great. I might be up to it later, but as it is, I think we'll do fine inserting the code and then adapting it to the articles' needs.

While it is helpful to talk about what LWiki is not, we will need to give some guidelines as to what is suitable. Translating the consensus that I hear between Anna, ciphergoth, and myself on the purposes of LWiki, here are practical tips on what articles should be made:

  • Any non-standard word, acronym, phrase, or concept used on OB or LW. Anything someone with some background in math, psychology, economics, or philosophy would not recognize. I am currently classifying these things as "Jargon".
  • A standard topic that is of particular importance to OB/LW. These articles should contain a short description (think dictionary, not encyclopedia entry), reference primarily on more widely-known resources such as Wikipedia or the Stanford Encyclopedia of Philosophy, and provide other references that are of interest to our community, but not suitable for adding to Wikipedia.
  • Study guides on a particular topic. Personally, I'd like to see the Mega-Cluster broken up into digestible chunks.
  • Maybe at some point, articles on applied controversies like Annoyance mentioned can be developed, but I think it's important to develop a strong rationality reference first.

I like Anna's suggestion of including open questions or possible points of investigation, but I think that would be better suited to a section in the article it is related to. Depending on how useful that turns out to be, we could set up an Open Question Portal.

Are there better ways to structure this or categories left out?

So, badger (or others), you'd rather I refrain from building a questions-first alternative index into the wiki, that has a link for each question, and that collects under that question both a concise summary of progress to date (dictionary-style, with collected links to relevant OB/LW articles and to some other background sources) and of gaps in our current analysis?

I was hoping the questions-list could function as an entry point for study guides for newcomers who want to answer particular questions, as well as for readers or prospective authors wanting to get a birds-eye view and see where the gaps are. (The links off the questions-list would include the detailed open problems within the entry on that particular question/topic, as you suggest.)

But whether a questions-based index into OB/LW is or isn't allowed on the wiki, it would be nice to know, so that I know whether to write it. My inclination is to just go ahead and then see if people like it, but I'm not sure if ciphergoth is trying to veto that in his capacity as admin, or if there's been enough opposition that I should heed that.

For what it's worth, I like your ideas.

It seems to me that much of the value of a wiki should be a distillation and structured introduction to the material.

Here are my thoughts on the open question topic:

I think the main open question page is useful. However, reading over the questions there, they seem more like fundamental questions in rationality rather than the very specific issues to investigate I thought they would be. On that basis, I think that page could be productively developed as a beginner's starting point.

I'd like to stay away from making pages corresponding directly to those questions though. "What specific biases or error patterns interfere with accurate belief-formation?" should just be called "Cognitive Biases", and "What practical techniques can improve the accuracy of individuals’ beliefs?" should be "Individual Bias Reduction". Then the question on the beginner's page can just link to the topic or study guide.

The study guide pages can be more conversational listing references like the bias page you made, and I would love to see more pages developed like that. We will need to make a decision about how study guides should be integrated into the whole structure. I considered doing them as subpages, but it probably would be best to do everything as "topic study guide".

If there are very specific open questions, then I think they should go in an open question section on the main topic article.

I'd like to stay away from making pages corresponding directly to those questions though. "What specific biases or error patterns interfere with accurate belief-formation?" should just be called "Cognitive Biases".

Why? Also, are you just suggesting different titles, or are you suggesting some larger change in the content I'd been imagining? I'd been picturing a starting-point page with general open questions, as noted, linking to combined {study guide / open questions} pages for each topic, that summarize progress to date (mainly by a list of relevant OB/LW articles and other resources, with a sentence of so description of how each fits into the topic) and remaining very specific issues to investigate.

I'm mostly concerned about conventions and broad structure, not content, at this point. I agree entirely with the content you have produced (at least as a rough draft, as you acknowledge). Content can be refined as we go, but I feel it's important to settle questions about overall structure early on.

For any particular topic, I'd like to see two paired articles:

  • "topic name" with a short description, references to Wikipedia, standard sources, and a straight list of related OB/LW articles, written in NPOV
  • "topic name study guide" with content like what you have done, written in a conversational tone

This might seem like a bikeshed argument, but short names make it easier to search for a topic or link to it from memory. I also like the explicit link between the reference and teaching articles if reasonable pairs exist.

Is there another short description that would work better than "study guide"? I can think of "outline", just "guide", "syllabus", "open questions", "questions".

Does anyone else want to chime in on this? Ciphergoth?

I'm still unconvinced of the merits of the whole plan, but several people seem interested and I'd be delighted to learn that I'm mistaken and good will come of it, so I'm just standing back and seeing what gets built.

I don't mind using short titles; your point about easy searching and linking makes sense I think.

I don't want the thing I was envisioning to be called a "study guide", though. Because I want the individual "study guide"-like pages to serve not only as a guide for newcomers but also as a summary of what we've produced so far as a community, and where the gaps in our analysis are. Your other suggestions don't really unite the purposes either. We could just do e.g. "Heuristics and biases, conversational style", but that doesn't indicate purpose the way "study guide" does and I don't much like the way it sounds. Maybe it would help if I understood what purpose or use-case you have in mind for the NPOV style articles, as a contrast class.

Yeah, I understand that. Is there a term you think would be fitting? "guide" is the best term I can think of. It also wouldn't have to be just two types of articles. We could have "topic", "topic study guide", and "topic open questions" or something similar. I think "open questions" still sounds a little awkward, but that's just quibbling if we can't think of anything better.