Wikify AskMe April 22, 2005 1:02 PM   Subscribe

MegaPony plea to Wikify AskMe! (Way too much inside.)
posted by loquacious to Feature Requests at 1:02 PM (19 comments total)

I wrote the bulk of this a while ago intended for this thread, or one like it, as there have been many. Almost daily there are requests for AskMe ponies. I don't say this in critique of the threads or any of the posters or participants, but as an agreement that AskMe needs something more. I am not a web designer, database engineer, or programmer. This might be terribly long, or it might have fit inside an existing thread, but the more I thought about it the more I wanted to bring this to the forefront and see what people thought about it in its own space.

I feel that the main weakness of Ask.Metafilter is that it is a linear stream. There's no tree. It's flatter than a pancake on a bunny.

The document displayed at the main link for AskMe is not conducive to broadcasting an indiced and categorized one (or some) to many transmission like a reference document should except within a set time frame and with a limited/vanilla archive and search interface.

It puts an undue burden on the answer-seeker to perform due diligence in searching for prior threads to be useful, in asking questions or searching for answers.

The same questions and answers are going to keep on being repeated no matter what we do as long as AskMe is in a date-sequential and set-expiration-date mode. We're asking users to conform to a ill-suited system and learn arcane rules in pursuit of exchanging information that should flow more effortlessly and be better organized, or risk (often mild to occasionally severe) public rebuke.

Could we scrape what exists already in AskMe and prepare that for insertion into a wiki-like user-editable database, and just kinda stop throwing ponies of usually the wrong color, shape or size at it?

There's no logical reason (outside of cycles and bandwidth) that we shouldn't stray at least a little from the clean, minimal aesthetics of the blue in the green, especially if we want it to actually be useful and accessible to many over the long run.

It seems that AskMe could be hybridized into a sort of dynamic wiki and stream, with a variety of metrics and ways to delve deeper into the aggregate of information displayed on the front page rather easily.

Even if we had to start rough and loose, or use a third party, not entirely handrolled solution, and refine it into something more wiki-like as we went along.

I feel that a little hoariness and hairiness is needed in a networked user-created knowledge repository. Cruft. Nooks and crannies for things to crystallize in. You need budding branches - along with the trunk and the roots. Seeds. At e2 it's called freegel, and it is the surface or membrane between the completed, existing database, the uncompleted portion of the database where individual strings of text exist as nodeshells, titles or placeholders (Sort of like Stubs in Wikipedia) to the potential one article or a collection of them, and on out into totally open, undefined conceptual space.

Threads (or nodes) shouldn't close, and should be easier to find. Titles of threads should be mutable. Threads should bifurcate or fuse as needed as questions are refined or expanded. There should be version tracking, and a mode of off-thread communication or discussion available in-page that doesn't rely on third party applications or protocols.

For Ask.Metafilter, we should be willing to put the asking of questions and giving answers on hold for a month. Or even three. While we changed over and experimented.

Take a snapshot of the database, mine it and stuff it in, say, 4 different engines, user interfaces, and/or database combinations. While you're setting up these configs with the snapshot, let AskMe run or whatever in the meantime until they're ready.

Make a party or celebration or something of the temporary new question-and-answer freeze as the new rollout officially hits the main page. Try out one new config per week with a set of basic user privileges for logged in users, like where/how to index stuff, voting/flagging and more - all geared towards applying our collective knowledge of how AskMe should work and feel. At the end of the month pick out the best config.

Let us swarm it and give us the basic tools to try sorting it out and nesting and branching it. Folks that are really into that whole sorting and linking thing tend to be really good at it, and it seems pretty likely to me we could get a good, valid, MeFi-like seed built quickly to crystallize off of and bring in the laggers.

Picture something with the first hand experience, direct practicality and grass-roots purpose of AskMetafilter. The editability, indexing and categorization of Wikipedia. MeFi's acknowledgement of comment authorship,'s article crosslinking, and a front page like Wikipedia and e2 where a list of new questions or answers would tick off and old and new stuff gets churned up together.

But hopefully refined like MeFi.

If we make AskMe a true repository rather than a comment stream, and enable some sort of user-level tools, as users comb through the nodes and threads not only will things be put in their proper place but the proper place will have been made for them. Things will coalesce. A working, highly complex but well organized system will arise if given the chance.

The main beauty of AskMe - aside from the extravagant wealth of answers - is that the question and/or topic is pretty much completely open. I think we can keep that and improve on it, while organizing and freely (if not gleefully) eliminating the linear clutter, repeat questions and chatfilter questions without subtracting from AskMe's uniqueness in the user-created knowledge repository realm.

Above all, someone said once in an AskMe-related MeTa thread that went something like "Keep your filthy paws off my comments." I think I agree with that. But I think also we need to push AskMe into something more indexed and indexable, and furthermore that we need a whole wild herd of ponies to get the job done.
posted by loquacious at 1:03 PM on April 22, 2005

posted by Stynxno at 1:36 PM on April 22, 2005

Ask.MetaFilter: Let us swarm it

loquacious, indeed.
posted by breezeway at 1:47 PM on April 22, 2005

Correct me if I'm wrong, but youir main point seems to be right in the first paragraph: "I feel that the main weakness of Ask.Metafilter is that it is a linear stream."

With which I respectfully disagree: the fact that the front page is linear means that I'm exposed to and end up reading questions on subjects that I never would have looked twice at had they been buried deep in, say, "Culture and Arts > Books > Graphic Novels". The hodgepodge of the front page is part of Ask MeFi's charm, and I read it for the same reason I read the SDMB: I might stumble upon something cool that I never new existed. I'd be much less likely to do this if it was more "organized." Sure, an occasional repeated question is likely to pop up from time to time... but so what? Someone links to the previous answer, we all have a few laughs, we move on.

Leave it be.

Also, is anyone else amused by the fact that the above screed was written by "loquacious"? On preview: breezeway is, apparently.
posted by Johnny Assay at 1:49 PM on April 22, 2005


All the problems you point out exist and are completely valid, but I highly doubt you're going to get Matt to overhaul the system as it stands. He speaks for himself, but based on past behavior . . . it'll probably be left the way it is until it is completely and irreparably broken.

I think some of the solutions you raise are interesting, but there might be other ways to represent the data while maintaining the front page of AskMe as is - for instance an outright tree structure. At the top level you have a simple page with links bearing the name of each AskMe category. You click one, it takes you to another simple page with a list of sub-categories and very general questions related to that category. You click on one of the sub-categories and it takes you to a page with more meta-sub-categories and general questions related to the parent sub-category - and so on and so forth. Multiple instances of the same question could be hyperlinked numbers which take you to each respective thread which is more or less a clone of one parent question.

The AskMe frontpage should continue as it is, but a link to a page with a topical tree-structure as above would greatly enhance the ability of end-users to avoid duplicate questions.

There isn't yet a sufficient crush of questions to render AskMe unusable, but I agree with your general sentiment that something major might need to be done before we get there. Matt's call, though.
posted by Ryvar at 1:55 PM on April 22, 2005

If there were even only maybe just a couple of redundant words we could build you a pony. And a stable. And a farm. And an estate with noded race track and indexed pavillion. Alas..
posted by peacay at 1:55 PM on April 22, 2005

I think tags are the data organization method that Matt has chosen for Ask.Me. While offering different views of the back end might be useful, eliminating the linear feed would be disasterous.

Further, the idea of 'wiki' implies constant editing and moving things about, an idea which I dislike intensely. If I give an answer, it's within the context of the question itself, and the answers that have already been given. I don't want my words edited, moved, realigned or anythinged else that takes them out of their context.

Organizing archives in a useful way = good. Changing actual content in any way = horrible.
posted by jacquilynne at 1:58 PM on April 22, 2005

Johnny: I agree with you, wholeheartedly. That's why places like wikipedia and everything2 have various "featured article" features.

On E2, there's even a "random nodes nodelet" that does nothing but serve up random stuff in a list. There's also a "random node" link that allows you to trawl randomly through the whole database.

I wouldn't even think of banishing that feature. But you can still have that randomness along with some organization, and have it be even better and more satisfying in that regard. (Check out the front page of everything2 and google at all the crazy features they've developed over the years. That's just the main feature set. There's even more tucked away here and there.)

Occasionally repeated questions are one thing, but dozens, if not hundreds of them, are a whole different story. Even still, the structure I'm suggesting allows for similar questions to be grouped together, and allow users to ask more refined questions if the existing one didn't answer their specific question.

And to maintain that freshness under this suggested restructuring, there could be a ticker-list on the main page that lists only recently asked questions, along with all the other ponies.

As for my username, it was given to me, not chosen. Yeah, it fits. And am I really going to take a ribbing for being too wordy even at a place like this? /green monkey.
posted by loquacious at 2:06 PM on April 22, 2005

Ryvar, jacquilynne:

Yeah, I agree that any even hypothetical changes should be pretty subtle. And I agree with your ideas.

And Wiki is probably the wrong model to point out - I'm thinking something more like E2 which has the same concept of comment and article ownership as MeFi. Users don't get to change your actual stuff, but perhaps they should be able to tag it or vote for tags or catagories or something. On E2, anyone can add softlinks to any node, however, it the softlink table is full, it can take a great deal of repetitive work to get the softlink to actually appear in the table.

And yeah, E2's softlink system gets abused, but usually with hilarious results. It's more useful than not, regardless.
posted by loquacious at 2:10 PM on April 22, 2005

Occasionally repeated questions are one thing, but dozens, if not hundreds of them, are a whole different story.

I disagree with this statement. Go into the Ask.Me archives and find me any two questions that are really, truly the same. Not similar, not on the same topic, but with all the subtleties, shadings and specifics to the askers, actually the same. The duplication is probably minimal. And if you can find two questions that are the same, how close are the answers?

One of the things that makes Ask.Me great is the ability for individuals to get an answer to their actual question - advice, tangents, shades and all of that included. A searchable resource is useful for a broader audience, but it's a lot less useful to the individual.

I think your perspective makes Ask.Me more like an encyclopedia, and less like a reference librarian. There's lots of encyclopedias on the web, but not so many reference librarians.
posted by jacquilynne at 2:34 PM on April 22, 2005

loquacious: No, I know what you mean about E2, but I'm not really sure that it, or wikis, are useful models for AskMe - or if they are, they're too different from what we have now for people to readily accept.

The huge, gaping weakness AskMe has right now aside from the amount of traffic (which I'm honestly not sure how one would solve), is the inability to conduct a meaningful search for similar questions - and no, tags really don't cut it. They can help, but not the way that categorization and sub-categorization can.

Topically arranged tree structures are generally best for helping people find information when word/phrase searching and simple metadata are insufficient, hence my suggestion of a secondary viewing/searching method. The real problem here is - how would you establish such a tree? You could presumably allow people to select their location within a tree, but that would probably make a mess of things as new sub-topics are entered and people disagree about where in the heirarchy their *special* question belongs. Furthermore, filing all AskMe questions asked to date would be a monumental task of herculean proportions. I've got a lot of free time on my hands - but nobody has that much free time. Not even several somebodies with a more-or-less agreed upon filing system acting in good faith.

I don't really think there's an easy, or even complex, suitable solution I've heard of that addresses all these problems, which more than anything is why things will probably stay the same until the entire thing implodes on itself - or doesn't.
posted by Ryvar at 2:44 PM on April 22, 2005

Well put, jacquilynne. Mathowie has always said (imagine various MeTa links here) that he sees AskMe as building a repository of knowledge, but its primary charm and value is clearly about the present moment of asking and answering.

(now, if that moment could just be extended a bit. . .)
posted by nobody at 2:47 PM on April 22, 2005

I have always planned to move much of the archives (note that date sorting comes last, since it is the most useless of all) structure to the front page, by putting information in a sidebar that would let you surface the archives better, so that the site wasn't just a linear thing.

There are several metrics I could use to stuff things into the sidebar, I just need to decide on them. They could be any one of these:

- list of the top 10 most popular tags within a date range
- most answers in last 24 hours
- most favorited comments (notes on that in the earlier thread today)
- random list of category names

I definitely think the home page of ask metafilter has plenty of room to grow in this regard, to show you subjects or specific questions that you're interested in.
posted by mathowie (staff) at 3:15 PM on April 22, 2005

Thanks, Matt. I knew something like this was in the works, but I still wanted to throw my random ideas out there. If you don't ask, you won't recieve, right?

When you get around to deciding, perhaps you or someone should open up a MeTa thread that's just for AskMe pony wishes, and maybe a poll or something to decide what stuff people want.

It's all too easy and comfortable for any given person to simply say "It's not broken! Don't you dare fix it!", without imagining what could be, and how much better it could be.
posted by loquacious at 3:27 PM on April 22, 2005

Unfortunately, there are no real suggestions in this post, just a general sense that Wiki = Good and we should make everything more Wiki. Social Security? Wikify it! Public transportation? Dynamic user interface *and* stream! Nooks and hoary seeds!

Sorry loquacious - had some fun at your expense there :)

I'd suggest narrowing in on one thing at at time. Your big trip seems to be building a better homepage. Start with the data architecture that's available, the usage that most people are now accustomed to, the existing model, and try to design a better homepage. Throwing big concepts at a problem like ***how to best build a repository of all human knowledge*** is a fun party game when you're half drunk. But light a candle here.
posted by scarabic at 3:28 PM on April 22, 2005

For Ask.Metafilter, we should be willing to put the asking of questions and giving answers on hold for a month. Or even three. While we changed over and experimented.

I think it can be done simultaneously. AskMe can always hum along nicely, even while experiments are being done with the new UI.
There are three important things about what we have going here at AskMe:
  • a questions feed.
  • a dead-easy way to hold round-table discussion regarding the question and, as appropriate, solutions.
  • a sourdough starter that guarantees good growth: namely, the MeFi community.

    AskMe as any sort of "worth a daily look or two" system of sharing absolutely must provide those things.
    What we need to make this a better environment for the question-askers is:
  • a dead-easy way to find out if "this sort" of conversation has been held before.
  • a way to judge whether the community is going to be open to starting a similar sort of conversation again.
  • a dead-easy way to start the conversation.

    For question-askers, I think we're looking showing a visualisation of three dimensions of tagging. When the AskMe tag set is shown today, it depicts length (the tag name) and breadth (the tag size). Now we need tag strength-of-association as a dimension. This will let them "seek" an "area" to ask their question.

    I think a TSoA is probably derived by its "proximity" to other tags, as measured by how often it appears with other words and how "adjacent" they tend to be; and ditto with words shared by those other words (if {ABC} are always listed together (and never as pairs), the pairings (ie. {AB}{AC}{BC} ) probably have a stronger relationship than if only {AB}, {AC}, and {AB} pairs existed).

    The mega-challenge is in determining {A, A', A'} are all variations on a theme, and that {A'} and/or {A'} will add a little bit of "gravity" to the proximity calculation. But, y'know, a thesaurus might be just enough to do that trick...
    What we need to make this a better environment for the answer-givers is:
  • to be shown the kinds of questions we are most interested in.
  • a dead-easy way to contribute to the conversation.
  • a dead-easy way to usefully extend the conversation.

    IMO, the server's determination as to which questions to show us must:
  • randomly insertion undesired questions to test whether that disinterest is still in effect.
  • try to logically determine tendancies such that several factors are weighed in relation to one another, stuff that intuitively "makes sense" -- like, if you've answered in a question, you're probably going to be interested in it for a while longer, so it's good to show them (but not if it means bumping a "hot" question that you'd probably want to see; but if it's not so "hot," well, ... if you've asked that "sort" of question before then that might make it worth seeing... etc. Wishy-washy decision-making. Fuzzy logic.)

    The other half of the trick is to make it easy for us to either go back and re-contribute to the thread in a useful/unique manner; or contribute pure hard data in a useful/unique manner. This is the Wiki-ish aspect. We have some sort of conversation, and when we're done we can add notes to it.
    I'm not entirely sure that's really clear. On the front page, it'll mostly look like the current AskMe, in that you're shown a few pages of questions. The difference is that you'll be able to indicate how much you care about such questions; and this will influence which questions you're being shown and in what order. A feedback loop.

    Inside, you'll read/contribute to the answers/discussion centred on the leading question, as you do now. And/or, you'll read/contribute to a "scribbles" page that is fine-tuning what was previously said. Not a discussion, but a reference... but somehow, what happens on that page influences your front page, and I'm not entirely sure how. Some sort of notice about what's new, but perhaps only if we've shown very particular interest in a particular question.

    Oh, and somehow, these scribbles are also somehow easily linked to other questions that are being/have been answered. Leveraging the facts.

    Gahd. I hope I've made my idea/vision clear; it seems like a really innovative sort of thing.

  • posted by five fresh fish at 12:23 AM on April 23, 2005

    I am terribly disappointed that I am not being hailed as brilliant.

    On the other hand, I did write that late and night and I'm afraid to re-read it. It may be the very antithesis of brilliant...
    posted by five fresh fish at 2:48 PM on April 23, 2005

    fff sez : "The mega-challenge is in determining {A, A', A'} are all variations on a theme, and that {A'} and/or {A'} will add a little bit of "gravity" to the proximity calculation. But, y'know, a thesaurus might be just enough to do that trick..."

    Your long, rambling manifesto reminded me of the Porter Stemming Algorithm. While it won't address all of your points, it could help in resolving differences in tags ("travel vs. travelling" and such) and compacting the tag space a bit.
    posted by boo_radley at 1:44 PM on April 24, 2005

    You know, not everything needs to be a Wiki.
    posted by mrbill at 1:48 PM on April 24, 2005

    « Older Cold Fusion busted again   |   Here is my case study of MeFi for an ecommunities... Newer »

    You are not logged in, either login or create an account to post comments