Charts become a compelling possibility

January 31, 2005 1:11 PM

As MeFi grows, charts become a compelling possibility. I think it would be interesting to graphically chart (is that redundant?) MeFi posts on the basis of number of responses, keywords, and dates. It might be instructive/interesting to see what subjects touch nerves, and whether an idea that hits a sensitive spot in 2000 has the same effect in 2005 (and so on).
posted by humannature to feature requests at 1:11 PM (17 comments total)

What about silence? Silence is ever-so damning, or something, but how do you quantify the amount of it?
posted by raysmj at 1:22 PM on January 31, 2005

So do it already, hn.
posted by mischief at 1:26 PM on January 31, 2005

What about silence? Silence is ever-so damning, or something,

Confusing is, I think, the word your looking for, ray. When a statement is greeted with silence, (especially here in the virtual realm with all the usual cues eliminated), is difficult to figure out the meaning of that silence. Whether it's telling you were offensive, obtuse, or just boring.
posted by jonmc at 1:28 PM on January 31, 2005

Personally I wanted to chart what time of day was best for Ask questions, using the metric of quantity of responses. Too lazy to actually do it, mind you. Someone bust out the RRDTool and get to work!
posted by sohcahtoa at 1:35 PM on January 31, 2005

Silence can also mean you agree with or appreciate something but don't have anything to add.
posted by walrus at 1:39 PM on January 31, 2005

I should add a few thoughts. I do see that charts exist for use overall, and I see that there are ways to find 'most commented' posts. What I'm about is that a lot of data is being collected here, and the people who post here, while not a truly representative slice of the public at large, are generally bright and articulate (probably the average MeFi user has more education than the average citizen). This suggests that something compelling could be done with the aggregate data that would have an added value. That is of course essentially what Google does, with a larger and probably more representative slice of the 'net.

Charts are the most immediately compelling mode of using the data. Posts would probably have to be ground up through some sort of analysis engine, which would both introduce and remove certain biases in favor of arriving at some sort of determination as to the value/meaning of each post/comment. There must be a whole lot of bright folks out there working on such engines. In fact, there's no reason not to use competing methods of analysis to produce charts and other means of data display.

I think trolls, porcupines, lurkers, and assorted chatroom creatures would actually be more significant in these engines than their actual posts often are. An advanced analysis engine could, perhaps, examine how posts subsequent to a troll changed in tone. (This would require 'ratings' of 'trolls' as judged by other users.)

Anyway, I'm sure I'm not the first to think of this, but I can't find a MetaTalk post that brings up this subject exactly. I suspect this sort of thing will happen here eventually. Just wondering what the current thinking on this was.
posted by humannature at 1:39 PM on January 31, 2005

I hope you are all both damned and confused by my silence.

That is, before I broke it.
posted by cosmonik at 1:42 PM on January 31, 2005

You mean Charts like this?
posted by Steve_at_Linnwood at 1:47 PM on January 31, 2005

Re Steve_at_Linnwood: No, not exactly charts like that. Those charts exist for any site that wants them. I mean something that tells us more than growth in usage for the site, or most commented posts. Something that speaks to mindset, mood, persuasiveness of certain ideas under certain circumstances.

Charting silence is an interesting idea, too. What does it mean when a post does not get any comments?

Or charting anger. Which posts get 'angry' comments, which ones get 'humor'? As defined by an AI program of course. Or several AI programs workiing in competition.

As far as 'do it already' is concerned, well, it's not "my" site. I don't have the right or means to impose a feature here. But I'm not the sort to demand that someone (who is offering a 'free' service after all) create a feature just to serve my whims. Since this is a chat area to talk about potential features, that's what I'm doing. If I was a MeFi mover and shaker (something I've no right to claim) and if the MeFi community indicated that, God yes we want that feature (something I have no right to assume), then certainly I'd assume the burden. (By contacting a bunch of people better trained to handle it than myself, of course. But they are out there, and so is grant money. I write about such research, I just don't do it myself.)

For now, I'm just floating a balloon. I just think there's a lot of possibly-compelling (possibly lucrative - and MeFi costs money like everything else) data here waiting for a new use. Google, by simple comparison, is surely finding all kinds of new uses for the data it collects, and I suspect it's safe to say we ain't seen nothing yet.
posted by humannature at 2:00 PM on January 31, 2005

While we're on topic, is it even kosher to hit this site with a robot to collect data on, say, the most frequently used words (S@L, there's so much more interesting data to collect than shown on waxy.org)? If we were to contemplate such potential abuse, is it proper to clear the idea with #1 beforehand, perform the automated crawl at night, intentionally limit the bandwidth of the crawl, all of the above, etc.? Would an automatic IP ban be the reward of an abusive crawl?

[on preview]
possibly lucrative
By the way, the type of data mining I have in mind would in no way be lucrative or attempt to profile usernames. Perhaps just fun examples such as a MeFi version of this recently mentioned flickr color picker.
posted by fatllama at 2:04 PM on January 31, 2005

redundant.
posted by The God Complex at 2:44 PM on January 31, 2005

That flickr project is an example of 'intelligent' (i.e., somehow digested, designed, processed) and innovative assembly of collected data, in order to create something greater than any single piece of data could be, something compelling in a new way. (Isn't that the nature of the 'net overall, after all?) Re lucrative: What constitutes 'proper' use of data is not something I want to impose my own opinion upon. Possibly most users would be wary of use of individual pieces of data, such as user profiles, and certainly there are some 'acceptable' and some 'unacceptible' uses of data. But I do believe MeFi has the right/duty to find means to use its data in certain ways support itself and grow as it deems fit. And of course to offer something new and interesting, which again seems to be the purpose of the site.
posted by humannature at 2:49 PM on January 31, 2005

Oh. Silence measurement again. How do you do it? Maybe by gauging hits on a post vs. comments, offering a 'lurker quotient'. Silence, too, can be measured. I think.
posted by humannature at 2:52 PM on January 31, 2005

"Maybe by gauging hits on a post vs. comments [...] Silence, too, can be measured. I think."

Wouldn't this involve specialised links though (i.e. collecting the stats on mefi itself)? It would seem difficult to measure hits on links otherwise, as if you're not going to comment, you're probably less likely to look at the comments.
posted by walrus at 3:22 PM on January 31, 2005

I would love to create a program that detected 'anger' in a thread, and then just randomly would pick sentences from comments, italisize them and make a few computer-generated snarky comments. For example....

Oh. Silence measurement again.

Why must you always make the thread about you. Always with the Silence measurement again. comments. You nitwit.
posted by Arch Stanton at 3:38 PM on January 31, 2005

Shouldn't be too hard. Take the Eliza code, change the responses a bit ...
posted by walrus at 3:54 PM on January 31, 2005

What are MeFi's busiest times? I can't tell from Waxy, as the times given aren't timezone-stamped, and there's no time entries in the top 20 lists.
posted by bonaldi at 7:31 PM on January 31, 2005