Bay 12 Games Forum

Please login or register.

Login with username, password and session length
Advanced search  

Poll

Does The test server work for you? Are you willing to sped time helping me test it?

It works for me.
- 6 (31.6%)
It does not work for me.
- 1 (5.3%)
I'd be willing to help test it.
- 6 (31.6%)
I'm not interested in testing it.
- 1 (5.3%)
I might be willing to help test it.
- 5 (26.3%)

Total Members Voted: 14


Pages: 1 ... 3 4 [5] 6 7 ... 42

Author Topic: Agora, A better forum (Open Source Project): Now with Github and test site.  (Read 76212 times)

Eagleon

  • Bay Watcher
    • View Profile
    • Soundcloud
Re: A more effective internet forum
« Reply #60 on: October 07, 2013, 07:48:14 pm »

On closer examination of centrality algos and what information they require to function I realized that what I'm suggesting would probably only be useful to me. That means I'm the only person that should be putting it in there, haha. This is another reason I'm reluctant to contribute - I tend to go off on half-baked tangents that may or may not pay off. Sorry guys.
Logged
Agora: open-source, next-gen online discussions with formal outcomes!
Music, Ballpoint
Support 100% Emigration, Everyone Walking Around Confused Forever 2044

Anvilfolk

  • Bay Watcher
  • Love! <3
    • View Profile
    • Portuguese blacksmithing forum!
Re: A more effective internet forum
« Reply #61 on: October 07, 2013, 10:06:33 pm »

On closer examination of centrality algos and what information they require to function I realized that what I'm suggesting would probably only be useful to me. That means I'm the only person that should be putting it in there, haha. This is another reason I'm reluctant to contribute - I tend to go off on half-baked tangents that may or may not pay off. Sorry guys.

Are you kidding?!

You almost just described how research is done! Heck, It happens all the time that people are solving a problem that's already been solved only because it's termed differently. I didn't know about this notion of graph centrality, and but a quick look at the Wikipedia page tells me it'd be a great way to display an argument graph. One of many, perhaps, but absolutely relevant! These could simply be visualisation options: one based on time, another on importance/centrality, perhaps another based on how credible arguments end up being, etc etc.

I will say that speaking over the forum can get totally confusing, and I've been having a hard time figuring out exactly what you mean. It seems I've been similarly confusing, judging by Antsan's post, so we should definitely try to be clearer about what we mean.

------

Regarding the dangers of verbosity in the system we've been talking about: there's always a set of assumptions whenever you start any project. Ours has definitely been that there's an unfulfilled need for more serious, structured debate somewhere on the Internet. People are getting tired of meaningless interactions. I don't care what you ate for breakfast, or what it looks like. I would, however, like to engage in a lively debate about beliefs or whatever technical issues I happen to know. If there's an easy system that allows me to make my points, and at the end tells me whether what I'm saying makes social sense (within the context in which I have spoken/written), then I'd feel that's pretty great. Yes, even if I end up being totally wrong.

Hell, we could be deciding the system we should be building with the system we should be building! Which I guess is part of why it's important that it exists!

------

Big explanation alert

Regarding structure, what I mean is this. As a first step, imagine the Bay12 forums, EXACTLY as they are now. Now, add in another BBCode tag, like:

Code: [Select]
[argument attacks="argument1, argument5, argument115"]I think the existence of a debating system at Bay12 would be really cool[/argument]
So you could explicitly define, in your text, what you think is an important point for the discussion, or an argument. You also tell it what other arguments people have previously proposed it attacks. Every point made, like every post, gets its own ID, that people can reference - just like using quoting for posts!

Every time someone posts a new argument, the "outcome" of a debate is calculated. All points are highlighted, or marked, with a colour, anywhere in the spectrum from green to red. Green means fully accepted, i.e., acceptability 1, and red means totally refuted, i.e. acceptability 0. There's the entire spectrum of colour in between, or acceptability in [0, 1], for each argument to take. So at the end of this thread, we've have a fairly obvious clue of what Bay12 prefers in general, and we'd get a pretty clear picture of the kind of system that would work best for Bay12.

You keep the forum-like structure, with subforums and subsubforums and threads and everything. For the algorithm and system that reasons about arguments and tells you whether they end up being acceptable or not, that does not matter. That's why I mean it's a form of "meta-data", if you're thinking in terms of the argumentation system only.



Now get rid of the very strict forum situation. You could still have the forum and subforums, indicating the types of subjects debated. But each thread would be a graph instead of a list of posts. For instance, instead of making a post this long, I'd maybe cut it up, and make each paragraph of so an argument. Since each argument is supposed to be made to counter arguments other people have made, I'd probably declare a bunch of attacks as well. Instead of each argument within a thread being shown one after the other, like here, you'd be able to visualise a thread with a number of different "views", like some of the stuff Eagleon has been suggesting. If you want to see newer arguments, perhaps try to find a way to display the graph where the higher the argument is, the earlier it was posted. Perhaps you're interested in getting the gist of a discussion: then you'd use centrality, like Eagleon said. Or perhaps you're full of yourself and want to see your impact on the debate, and you'd get a view of your arguments in the middle, and the arguments they affect the most around, and so forth. There's many options!

One of the important bits of all this, is that underneath all of this stuff that's got to do with "how do I find a debate I want to engage in" (I usually only visit Other Games and Creative Projects, and avoid certain topics), or "how can I see/visually organise a debate", is a graph. And that graph does not need to be independent from thread to thread, which means one of our arguments could be attacked by something someone said in another thread. Heck, it could be attacked by a point made by someone in a NY Times online article, if they adopted the system!

I don't think we should visualise the entire Bay12 forum as a plain graph. It'd be IMPOSSIBLE to figure anything out. But forums, subforums and threads can become meta-data helping you get to where you want to be, and they can be visual cues for the graph visualisation. Like a nice box around all the arguments proposed in this particular thread. You might get some incoming and outgoing arrows, which is fine. They tell you there's another topic talking about similar issues! You might want to follow that arrow into another box and participate there too!

What you don't want to do is make any arrows disappear because then you can't really make sense of why the debate outcome turned out that way. The only reason you'd want to make arrows/attacks disappear is if they were super duper downvoted and they don't really have an impact, or if they are removed by a moderator.


Sorry for the long forum post, but I hope this gives you a better, more concrete idea of the kind of system we've been thinking about.

Eagleon

  • Bay Watcher
    • View Profile
    • Soundcloud
Re: A more effective internet forum
« Reply #62 on: October 07, 2013, 11:17:07 pm »

Regarding the dangers of verbosity in the system we've been talking about: there's always a set of assumptions whenever you start any project. Ours has definitely been that there's an unfulfilled need for more serious, structured debate somewhere on the Internet. People are getting tired of meaningless interactions. I don't care what you ate for breakfast, or what it looks like. I would, however, like to engage in a lively debate about beliefs or whatever technical issues I happen to know. If there's an easy system that allows me to make my points, and at the end tells me whether what I'm saying makes social sense (within the context in which I have spoken/written), then I'd feel that's pretty great. Yes, even if I end up being totally wrong.

Hell, we could be deciding the system we should be building with the system we should be building! Which I guess is part of why it's important that it exists!
I tend to agree, especially since it's much simpler than pushing for what I want out of it. Still, I wish I had a formal background in programming this stuff so that I could tangent a little better, haha. If nothing else, I still think I'm right about letting users fragment a composite argument themselves, rather than requiring them to make precise ones to begin with and punishing them if they don't. I'll take a crack at the basic model (no fragmentation, simple directed graph, a separate visualization from BoboJack's) again tonight/tomorrow, I've brushed up on an unrelated project and my programming juices are flowing nicely.
What you don't want to do is make any arrows disappear because then you can't really make sense of why the debate outcome turned out that way. The only reason you'd want to make arrows/attacks disappear is if they were super duper downvoted and they don't really have an impact, or if they are removed by a moderator.
Question - if you allow for removal of arguments, what do you do about arguments that have been orphaned into their own little isolated graph? If you make attacks disappear this way through user action, it could happen quite frequently, no matter how high you set the threshold to destroy them. Would those nodes just disappear as well? Any time I hear or talk about disappearing nodes/edges through user moderation I kind of cringe for this reason - if there's a faulty argument/attack I think it should be left alone with the arguments against it left intact, or at least simply hidden, so that people don't keep making it and making more orphan networks.
Logged
Agora: open-source, next-gen online discussions with formal outcomes!
Music, Ballpoint
Support 100% Emigration, Everyone Walking Around Confused Forever 2044

Antsan

  • Bay Watcher
    • View Profile
Re: A more effective internet forum
« Reply #63 on: October 08, 2013, 03:44:48 am »

I don't think we should visualise the entire Bay12 forum as a plain graph. It'd be IMPOSSIBLE to figure anything out. But forums, subforums and threads can become meta-data helping you get to where you want to be, and they can be visual cues for the graph visualisation. Like a nice box around all the arguments proposed in this particular thread. You might get some incoming and outgoing arrows, which is fine. They tell you there's another topic talking about similar issues! You might want to follow that arrow into another box and participate there too!
That makes sense!

Quote
What you don't want to do is make any arrows disappear because then you can't really make sense of why the debate outcome turned out that way. The only reason you'd want to make arrows/attacks disappear is if they were super duper downvoted and they don't really have an impact, or if they are removed by a moderator.
I think that views that do not display all edges in a given induced graph could be very useful - maybe you only wanna see how arguments relate to a certain post or you can't figure out what's going on and you need to get rid of some edges, so you want to find a planar subgraph to display or cut down the number of edges per node to some maximum.
If the displayed graph gets recalculated while browsing the missing edges aren't that important - the user will find those while exploring. Missing edges can be indicated in some way.


Question - if you allow for removal of arguments, what do you do about arguments that have been orphaned into their own little isolated graph? If you make attacks disappear this way through user action, it could happen quite frequently, no matter how high you set the threshold to destroy them. Would those nodes just disappear as well? Any time I hear or talk about disappearing nodes/edges through user moderation I kind of cringe for this reason - if there's a faulty argument/attack I think it should be left alone with the arguments against it left intact, or at least simply hidden, so that people don't keep making it and making more orphan networks.
I forgot the name, but I think there is a way to detect nodes that are necessary in a component of a graph to keep it connected. If such a node is removed, calculate the new components and make them new topics in the given forum, maybe name them "[topic] - Split [n]", alert the administration and give someone the possibility of choosing a better name for the topics.

I'll start programming an engine for this today.
Logged
Taste my Paci-Fist

Angle

  • Bay Watcher
  • 39 Indigo Spear Questions the Poor
    • View Profile
    • Agora Forum Demo!
Re: A more effective internet forum
« Reply #64 on: October 08, 2013, 10:33:02 am »

So It looks like we have two or three different ptrogrammers- Do we want to coordinate our efforts through github or something?
Logged

Agora: open-source platform to facilitate complicated discussions between large numbers of people. Now with test site!

The Temple of the Elements: Quirky Dungeon Crawler

Eagleon

  • Bay Watcher
    • View Profile
    • Soundcloud
Re: A more effective internet forum
« Reply #65 on: October 08, 2013, 01:11:39 pm »

This is the nonsense I worked out on my whiteboard last night, and just pushed into an SVG mockup. I think that as you zoom in closer to individual arguments, particularly in tightly clustered and highly connected groups, you're actually going to want to focus on five or six at a time in a more readable way, to allow faster vote turnover. You could use fuzzy logic based on that small network, the network as a whole, or whatever set you like to provide the ratings shown

In reality, the change that I depicted ("Your mom misread god" disappearing) would probably result in a reshuffling of all seven nodes based on a few predetermined selection criteria such as high or low rated attacks - you could allow users to pin nodes to prevent this from happening, or prevent it in part, so that you could remain focused on All Hail Grod the Uneasy as an attack, keep Non-Falsifiable in play, and try to see how they relate to the central argument better.

To get this view (as far as I've worked out)
1. Pick an argument.
2. Pick something it's attacking, attach it to a single-axis scaling connection at the top of 1.
3. Collect what attacks 1, decide how to prioritize their display. Attach the two or three best to the other sides with another connection of the same type. Scale all nodes to the screen. Find connections between all nodes and draw them.
4. Collect what attacks 3. Prioritize their display the same or differently (could provide contrast, select for arguments that attack both 1 and 3, etc). Attach them to free sides. Scale, find connections, draw.
And repeat until you're satisfied with the number of nodes on screen. Then you can do the fancy work of figuring out connection paths that don't intersect your OGL objects, preventing overlap of argument blocks, working in floating force-atlas style network sections on the perimeter, allowing hiding and revealing of argument bodies and/or blocks etc. I'm willing to bet you could modify a standard visualization like BoboJack's to do this if the visualization allowed you to single out nodes you don't want animated, and if it doesn't, it could be rewritten to. You could get really fancy and send the nodes back off into the network like a bomb exploding, drawing new ones that have been selected in.

Lots of fiddly! I don't like UI work =P None of that would work for me the first, second, third, n^fanciness times, because there's always some boundry or glitch you aren't going to expect that ends up making it look like garbage all the way through, haha.
Logged
Agora: open-source, next-gen online discussions with formal outcomes!
Music, Ballpoint
Support 100% Emigration, Everyone Walking Around Confused Forever 2044

Antsan

  • Bay Watcher
    • View Profile
Re: A more effective internet forum
« Reply #66 on: October 08, 2013, 01:23:03 pm »

So It looks like we have two or three different ptrogrammers- Do we want to coordinate our efforts through github or something?
I don't know. BoboJack uses Haskell, I use Common Lisp and whoever else is in on this thing is probably using neither and I guess no one is really willing to switch languages.
If there is anything to coordinate via github besides collaborative programming...
Logged
Taste my Paci-Fist

Eagleon

  • Bay Watcher
    • View Profile
    • Soundcloud
Re: A more effective internet forum
« Reply #67 on: October 08, 2013, 01:44:20 pm »

Anvil, does your group have any larger arguments they've thoroughly deconstructed we could use as test data? I'm talking about something with maybe twenty or thirty nodes. Even if it's incomplete, it could speed things up considerably to have something besides a randomly built network to work with.
Logged
Agora: open-source, next-gen online discussions with formal outcomes!
Music, Ballpoint
Support 100% Emigration, Everyone Walking Around Confused Forever 2044

Anvilfolk

  • Bay Watcher
  • Love! <3
    • View Profile
    • Portuguese blacksmithing forum!
Re: A more effective internet forum
« Reply #68 on: October 08, 2013, 01:47:52 pm »

I'm at work and will catch up with the thread later, but in short: no, we don't. We've used randomly generated graphs for developing the algorithm, and we always try to use the smallest examples possible that exhibit a certain behaviour.

However, it shouldn't be too hard to go on to www.debate.org, find one of the more refined debates, and use that as a source for choosing what is an argument and what isn't. I've seen a couple of debates with a level of formality that makes it easy to identify what is an argument and what isn't, and what attacks what.

Eagleon

  • Bay Watcher
    • View Profile
    • Soundcloud
Re: A more effective internet forum
« Reply #69 on: October 08, 2013, 03:26:10 pm »

I found a tool called Gephi while trying to figure out a reasonable format to share a common but human-readable graph to test on. Its graph streaming plugin and the existence of a plugin to import RDF data from a server seems particularly interesting - I have no idea if Virtuoso is any good, but this seems to indicate that we might be able to use Gephi as a backend for this project. I know that lugging around a large platform like that might be more trouble than it's worth, and lots of people hate Java, but there's plenty of interfaces to it for other languages that you guys could use for your own implementations. Worth considering.

If nothing else, Gephi can export to CSV, with a fair number of generic datatypes built-in (Data Laboratory is where you want to look if you'd like to specify more columns). Here's the files for this - Nodes (edit: fixed small defect in the state military section), Edges - if anyone wants to work that functionality into their program using a non-random data set.
« Last Edit: October 08, 2013, 04:40:47 pm by Eagleon »
Logged
Agora: open-source, next-gen online discussions with formal outcomes!
Music, Ballpoint
Support 100% Emigration, Everyone Walking Around Confused Forever 2044

Anvilfolk

  • Bay Watcher
  • Love! <3
    • View Profile
    • Portuguese blacksmithing forum!
Re: A more effective internet forum
« Reply #70 on: October 08, 2013, 03:29:38 pm »

Also, can't believe I forgot this, but before you go off and implement lots of things, it might be worth looking into:

http://wiki.lesswrong.com/wiki/Debate_tools

and http://debategraph.org/ in particular.

Unfortunately, none of these provide any outcome of a debate, let alone a formally justified one, which is what we'd like to have.

Urist McScoopbeard

  • Bay Watcher
  • Damnit Scoopz!
    • View Profile
Re: A more effective internet forum
« Reply #71 on: October 08, 2013, 05:06:26 pm »

I like the structure of debategraph a lot, I won't lie.

that's kinda what i've been envisioning in terms of organization.
Logged
This conversation is getting disturbing fast, disturbingly erotic.

tompliss

  • Bay Watcher
  • Goodbye ?
    • View Profile
Re: A more effective internet forum
« Reply #72 on: October 10, 2013, 05:06:51 am »

So It looks like we have two or three different programmers- Do we want to coordinate our efforts through github or something?
I don't think that's what is important now... at least for angle's project.

Here is how I see it working :
 - First, you (as in every concept designer involved here) need to study the existing tools, as they will give you some more ideas.
 - Then, you (Angle, as it is your project) should write some senarii, some user stories, without thinking about an interface. Those should be understood by everybody invloved, as they will be the core of the development.
 - Developers take the lead, there, building mock-ups and showing them for tests and approval.
 - After a few iterations on the mock-ups, something good should be ready to launch a kickstarter, or "simply" search for fundings.
Logged

Angle

  • Bay Watcher
  • 39 Indigo Spear Questions the Poor
    • View Profile
    • Agora Forum Demo!
Re: A more effective internet forum
« Reply #73 on: October 10, 2013, 09:58:11 pm »

User stories? Like "When users log on, they will see... When they post a new thread they will see...", or like "I used this, and it helped me.."?
Logged

Agora: open-source platform to facilitate complicated discussions between large numbers of people. Now with test site!

The Temple of the Elements: Quirky Dungeon Crawler

tompliss

  • Bay Watcher
  • Goodbye ?
    • View Profile
Re: A more effective internet forum
« Reply #74 on: October 11, 2013, 01:39:32 am »

First option :D
They're little scenarii that describe what is done with the solution, with few details on the "how".

They're good at dividing the solutions in a list of features, and seeing the links between them. And of course, they're very good at creating Task-Trees, that are very good at abstracting UI.
Logged
Pages: 1 ... 3 4 [5] 6 7 ... 42