What is TTD?
A way of capturing the structure of argument. An example of TypingNotCategorizing
It's a simple threaded discussion forum, but one where contributors are required to give each posting a "type" selected from the following set : ("counter-argument", "counter-example", "gainsay", "supporting argument", "supporting example", "agreement", "extrapolation", "exegesis", "summary", "reminds me that", "joke", "troll", "garbage") to describe the relation to the parent post.
Some points :
"garbage" would be the default, forcing the user to choose another type
all comments would be displayed, but would be sorted by type. (Ordered as above)
moderators would be able to change the type of a comment, but not it's contents nor remove it.
All those "would be"s sound pretty hypothetical, does it exist?
Originally I wasn't going to release the code because it's pretty lousy (it was my first attempt at doing CGI in Python, and I suspect very badly structured. Instead, if you're a coder looking to do something with TTD, you're better off starting with a good existing ThreadedDiscussion forum and trying to add typing to that.
But then I figured, why try to force your decision. The code exists, it does work (apart from the TrackBack thing, which probably wasn't a very good idea anyway.) So [here http://www.nooranch.com/critical/ttd.tar.gz here it is]
Yes, but what's the point?
The point is that you use it pretty much like any other Threaded Discussion, for the same kind of questions and answers, debates, flame-wars etc. With one extra action required : ticking a checkbox as you post. However, when a new user comes to a discussion that already has 40 or 50 replies they see the structure of the existing discussion much more clearly. They can see where the arguments are, the questions, the side-tracks into amusing annecdotes. And they can organize their browsing accordingly. Do they think the counter-arguments are most important? Or the evidence? Or the jokes? Or the exegeses?
If it's easier to find the responses of a particular type, you'll be less likely to repeat them. Say you want to request some clarification on a particular post and there are 15 replies. In a normal threaded discussion forum you can check out all 15, but you might be tempted to simply ask your question again, because 15 is a lot of threads to scan. In TTD, you'll see that, of those 15, in fact only two are further questions, the others are adding supporting or contradictory evidence. Two is a lot fewer to check than 15 so, you acccept the cost of looking at them first, to see if you're question is already addressed.
More notes and links
Who Marks Up?
Is the contributor always the best person to assign a type to the post? -- AdrianHoward
: not necessarily. I'm thinking of experimenting with either moderators being able to set the type or anyone(!) being able to change the type of a post. Possibly the right thing depends on circumstances. Although the initial contributor should put some typing in.
:Also note, as you can see from the discussion on the board, I think it's possible different communities and types of discussions need different sets of types. My first ideas were inspired by thinking about PhilosophicalMarkup, but it's very obvious that there's a need for a richer vocabulary of suggestions and speculations. This is very much work-in-progress. -- PhilJones
Threaded Discussion is Hierarchical
Is that the best structure for complex arguments? See ADocumentIsNotATree
Hmmm, sounds good (I hadn't been as convinced when I saw this page on MeatBall, even though it contained the same thing :)
Two comments :
- Wouldn't it be interesting if the types available for an answer depended on the type of the original post ? - or at least if the type of the original post had an impact on the default type and the way they are classified. For an open question, you expect answers, opinions ... (alternatively, when posting the user could also in some way decide what kind of answers to expect - this could be implied by the type but there could also be an "expected-response" type - one type for the relationship to the father, one for the expected relationship to the sons Troll is alsmost only a statement about the answers expected).
: phil : Interesting idea, an application where a post can specify the range of types of response.
- I wonder if splitting up the types (choosing from disagree-neutral-agree and argument-example-question ... something like that) would make things better or worse ? It might confuse the user, but it can be a way of having more than one type. This would require making a more complex (possibly some kind of hierarchy) representation of the types, but it could be hidden behind a name given for every combination of types. Sounds awfull / unusable / pointless like that, and heck, maybe it is. Combined with the previous idea and possibly a knowledge of user preferences, the system might offer the user an educated guess of where the post is likely to fit.
: phil : I did think about organizing the types into a hierarchy, but the two worries are real : how do people navigate to and find the type easily, and what's the right classification. Remember, the guiding spirit of TTD is that it's meant to be a quick-win. A simple extra action (1 decision / 1 check-box) to make a normal threaded discussion forum more useful.
Hmm, I'm adding extra types everywhere here, I'm not sure it would help. Other possibilities :
- answering to more than one topic at the same time (usefull for a sumary for example, or adressing a point that's been made severall times so it isn't answered in the same way several times). The relationship to each other topic may not be the same (though that complicates matters quite a bit, maybe it should be skipped)
: phil : that raises the most obvious question (to me) about TTD, why not just go all the way to a wiki with typed links ;-)
: Well, you probably still need some kind of user identification (Hmm, though one can wonder if the point is to represent the opinion the people have of what's written, or merely the logical relationship between the posts). I think the wiki system would need quite big an overhaul to work effectively for debate (see MeatBall:WikiForDebate and MeatBall:DoubleWiki). You probably need user identity (much more convenient than trying to find some kind of consensus), representing multiple points in the same page (whereas in a wiki you would tipically have a page per idea). Though a wiki with typed links would be interesting :)
- Adding types without posting content - rating, or changing the type of the subject (labelling it as garbage, a troll, etc.). Allowing third party labelling can remove the use for some categories (though people may still want to post stuff under garbage / troll).
: *phil : Well remember in TTD it's the links that are typed, not the posts, so I'm not sure what this would mean exactly. But it's another interesting feature for any discussion forum. *
: But each link is attached to a post, right ? So there may be some confusion as to wether the type applies to the post or to the link. A typed link without a post behind it would simply ne a label.
- Similar to the first idea, allowing users to select in which order to view the postings (which types are first ...)
: phil : Ordering / filtering by reply-type is one of the main purposes of TTD. In the example I'm running at the moment it's fixed, but certainly it should be part of the reader's UI
Wow, I'm designing some big DancingBearware there :) maybe it's more ambitious / complicated than type threaded discussion, and maybe it doesn't add that much. I see it as a system where organizers are as much valued in the community as posters. A lot of the work can be into finding similar arguments and linking them together, classifying the best arguments, allowing to relabel your post as "hey that bloke over there just said the same thing only better, go see him instead" with the corresponding impact on the thread representation (your post may become a "reformulates, but not as well" answer to the better post) ... of course this may bloat and require a dreadfull rating system / trust metric, but it may be possible to take some of these ideas and make something out of them that could work for a small community -- EmileKroeger
: phil : not quite as grand as your suggestion, but this reminds me of why TTD was meant to support TrackBack. Basically a post in a TTD discussion could be a link to any external resource, entered via trackback. So I could post something to my weblog, trackback it to TTD, and have a place-holder for it in a slot in the TTD hierarchy. Still debugging that though :-( (See also ThreadServer)
Well, the main reason for TTD, and the difference with something like ClaiMaker, is that this is a quick hack. The idea of TTD is to be minor increment to ordinary threaded discussion forums, the kind you see stuck to sites all over the internet.
So, in my conception, who'd use it is anyone who currently uses threaded discussion forums ... SlashDot users, Kuro5hin users, extremevbforum users. And the kind of knowledge that is being marked-up is just the usual chatter on those fora. (Note, TTD isn't meant to be suggesting any particular ontology for marking up the arguments. I'd expect each community to establish their own. )
But actually I may change this stance; as SituatedSoftware points out, a non-scaling, prototype code might suit small groups. The other place I'd love to see it tried is on the intranet of a university department. I keep meaning to organize this with my friend who teaches in the local philosophy department.
TTD is the future ... or is it?
I said TTD was one of the coming technologies of the next couple of years over on TrendPredictions/October2004.
GrahamLally wrote a good comment in response. But I've been a bit dictatorial and moved it back to the TTD itself :
There are a couple of meta-comments I'll make here though. Clearly, it's not easy to have a TTD discussion without TTD beoming the topic of conversation itself That shouldn't be surprising, AllMediaTalkAboutThemselves (it's one of the ways NetworksCreateValue, by talking about themselves they actually move to a more abstract representation of the discussion and the problem)
But at the moment, discussion about TTD tends to swamp any other discussion you have there and even prevents it starting.
So I have two hats :
- one as cooly objective observer of genres of OnlineWriting. I think .... "hmmm. Maybe there is a flaw in the whole thing." I can't see there being a problem once TTD is established. Eventually users will get comfortable with the typing of the links, and they'll understand how to break up their response into smaller chunks and to select the right type. In fact, learning to do this is part of the value of TTD as it helps you have more structured discussions.
: But this isn't really very dynamical or "BottomUp" thinking, is it? Because we have to consider the ecology of the MediaSystem as a whole. In evolution, a new trait such as a strategy or product must be InitiallyViable in a world where it is new and rare, not merely be fit when it's already well established and in the majority. So yeah, this extra difficulty of a) thinking how to classify your first few posts, and b) worrying about how to decompose your response into classifiable chunks might very well kill it. In which case all wild claims about it being TheNextBigThing are to be discounted.
- wearing the other hat I think "Fuck you! TTD is my baby and I love it! It's cute and it's right! And one day, you will love it too, you bastards!!! Just you wait and see, bwooaahaahaahaahaah!! ... please (sniff) give it another chance, please .... :-("
More importantly, all suggestions as to how to make it more initially viable and obvious are very welcome.
This is why I'm trying to not come across as to pooh-poohy (is that a word?) about TTD - because I like the idea :) And without the TTD prototype, I suspect my brain wouldn't be mulling over how it could be improved whilst keeping the fundamental premises in sight.
If I get the time, I might try and throw some perl scripts together to do something, but in the meantime, here's my thinking.
Why force people to split their missives up into small, atomic blocks, when they already have them to an extent, in a manner that's easier to read, write and think. Basically, parse incoming text and treat each paragraph as a separate chunk of text which can be related to any other chunk of text.
Is a single relation type appropriate, or could it adopt a more layered/conglomerated approach, a la Slashdot say, so that a poster can assign a few different response types to their answer in one go (e.g. "[humourous, counterpoint]"). Similarly, other people who don't agree with the categorisation offered, or want to add some where there is none, can "rate" the relation as they see it. This still lets conversations be sortable and/or machine-readable.
Allow people to classify their paragraphs as they type in a semi-wiki-like style, e.g. precede them with "[explanation]". This means you can classify your response while not having to break it up into lots of actions and break your flow.
Allow any paragraph to be related to any other paragraph in the thread/system (maybe via some unique URI). By default, replies to a message are naturally attached to the preceding paragraph and have some generic "response" relation type, but it may serve as a response to another point elsewhere too.
This partially comes out of my love of e-mail, I admit. I think you could easily combine TTD with e-mail lists, allowing people to classify their paragraphs as per #3, and then getting people to "moderate" relations in a web-based GUI later. (You could easily post via the GUI too.)
Maybe I'll give it a go, maybe I won't. I figure I might as well post the ideas here to see what people think anyway... -- GrahamLally
Where has the thread headed?
TTD is the prefix, what is the suffix?
Seems like this structuring of conversations could also have a reviewing/rating/ranking component
Some good information(value="6")
Not very helpful(value="2")
Way off topic(value="1")
Could be another navigational aid
Rup3rt, a couple of points.
I didn't look at rating systems because everyone else is already trying that :-) TTD was an attempt at a different approach. Though of course, you could combine the two ideas in an actual system.
Remember TTD gives types to the relationships between posts, not the posts themselves.
I really don't like the "moodle" type system where you try to give numbered ratings "names" like "way off topic". What if something is "way off topic" and yet the most interesting thing you ever read that you really want to draw everyone's attention to?
"D'oh!" "agreement" "Final Word"
Yup, conversations are most often rated by their durability. I saw from ThreadedConversations that most threads only contain two messages. I am still keen to spot the fulcrums of threads or conversations - these should show up in the patterns of TTDs.
Have you done any more on this, it sounds very promising, a way to focus on the conversation rather than scoring points, raising the collective value? Have you ever come across SpeechActs? - kk (requesting information :)
I still think the idea is very promising. But no-one else seems particularly enthusiastic. I think there are several possible reasons. The obvious one everyone mentions is that it's too hard to identify what type to give the link. Typing the link clearly needs some extra thinking by the user, and that turns out to be a show-stopper. Less the "choose a type" itself part I suspect, and more the implied constraints it puts on the post : you should really split a complex response into several different replies; and you must choose one of several possible types. I think these feel very constraining. The cognitive overload is too much. (Maybe the types themselves aren't that good too.)
Of course, the whole point is to force a little bit of extra work out of the user, in exchange for a global payoff. But no-one is buying. That brings us to the second issue. TTD was intended to solve the kind of situation you get on something like SlashDot where there are 800 responses to a post, some of them excellent, but it's hard to see the good and relevant stuff amongst all the junk and posturing. I think even the system I have now would start to show its value in such circumstances. But with my hardly visited toy, the number of responses is too small for the typing to add any value. Potential users see the extra pain but not benefit.
The obvious answer to that is for me to take a long argument or discussion and mark it up myself, manually. That way I'd be able to demonstrate the value of information organized this way. That will take some time though, which I don't have at the moment.
There are two other relevant points to make. One is sort of positive; the other is sort of negative.
On the bright side, there's all this excitement about tagging / FolkSonomies. That has something of what I was getting at : a little bit of locally based, easy to add, relative classification which nevertheless, when aggrogated, adds a lot of value. (What I rather badly labled TypingNotClassifying) So, I can claim that the success of tagging vindicates some of my intuition. And it may inspire people to look for more of this kind of thing, and to reconsider moving in a TTD direction. Less positively, it may be that the real key to tagging's success is the lack of perceived constraint as the user can add any tag he or she likes. TTD is, in contrast, more constraining. And perhaps permanently on the wrong side of some cost-benefit calculation.
The negative point is this. The ability to type links was one of the earliest features I added to SdiDesk. It's not that clunky. Yet I never use it myself on my personal copy. In the context of my personal information management, it hasn't been so useful. Of course, TTD is meant to make arguments navigable by outsiders. I don't need it because I know my stuff pretty well. In fact the most frequent link-typing I do is here on ThoughtStorms with Warp****Links. So it might not be important, or it might be an indication I'm wrong about the value of typed-links.
Like I say, it seems to me that the secret of "selling" TTD is going to be manually preparing some examples of large-scale arguments. I'll try to have a go at that sometime this year.
All other suggestions welcome. Cheers.
ps : no, what are SpeechActs? Do you mean in the J.L. Austin sense?
See also :
TruthMapping looks very similar
Meatball:TypedThreadedDiscussion (same original post, more comments and links to related ideas)
SimonBuckinghamShum and others at the OpenUniversity have been developing ClaiMaker, a project that bears some resemblance (and is more mature). See [[http://portal.acm.org/citation.cfm?id=775187&dl=ACM&coll=GUIDE&CFID=11111111&CFTOKEN=2222222 http://claimaker.open.ac.uk/, http://portal.acm.org/citation.cfm?id=775187&dl=ACM&coll=GUIDE&CFID=11111111&CFTOKEN=2222222
Scholarly publishing and argument in hyperspace]
Wiki Court : http://www.aaronsw.com/weblog/001175
Threaded discussion in blogs : http://steve-yegge.blogspot.com/2006/03/blog-or-get-off-pot.html