[18:26:58] halfak, question when you're back [18:33:26] Ironholds: I think he just left [18:33:47] YuviPanda, that's okay! I had an idea for a piece of research, or rather, a twist on a piece of research I did a while back, using ORES [18:34:00] ah, nice [18:34:21] I was going to ask how it does with evaluating historic data. If it's good at it, we could run the model over things the AbuseFilter has caught and validate how well it works and if it penalises different classes of user [18:34:57] nice [18:35:07] computationally, it's easy to use revscoring against historical data [18:37:12] yeah, it's the accuracy I'm interested by [18:37:52] a related piece of work would be testing the accuracy against historic data in order to identify whether the nature of vandalism has changed over time ;) [18:41:47] * YuviPanda nods [18:42:05] Ironholds: are you going to be in SF for the things and stuff? [18:42:31] I am! I show up just for the all-hands and then stay for a week and a bit after [18:42:42] nice [18:42:58] you? [18:43:17] Ironholds: I theoretically live here still [18:45:00] hahah [18:45:02] excellent [18:45:04] we should hang out! [18:45:29] Ironholds: indeed! [21:14:08] o/ [21:14:12] Ironholds [21:14:29] hey halfak :) [21:15:00] Saw idea. Definitely could be interesting. [21:15:18] I've been using our wikidata model to find vandalism that hasn't yet been reverted. [21:15:26] So the idea definitely works in theory. [21:15:42] Historically, I expect that vandalism has been changing, but probably not enough for a substantial effect. [21:15:56] We'd need to manually label ~200 edits to know for sure though [21:16:11] sounds doable! [21:16:31] And more fun than my current problem of "this course has a test question they never taught us how to answer" [21:19:20] Boo. [21:20:05] actually I may just be misreading the question. It's identifying parameters estimated at 0 in a quadratic function based on various lines [21:20:21] isn't that just, are the slope or intercept 0? [21:20:47] So, if it intercepts (0,0), W0 is 0. If it's a straight line, W1-N are 0. If neither, neither? [21:21:38] Not sure what you are saying, but an intercept is usually at a zero value for one dimension. [21:21:45] e.g. the value of y when x == 0. [21:22:10] *nod*. This is a confusing question. [21:22:45] Actually this entire series is so far really off-putting. They want us to use a non-commercial-use ML package and decided an appropriate case study of why ML is useful was "predict how much more money you'll make now you've done this course!" [21:24:29] Heh. That kinda sounds like a fun question that would make prediction models very real for some people. [21:24:46] It's kinda cool seeing an algorithm that could affect your livelihood. [21:25:02] Someone is definitely modeling your earning potential. [21:25:35] totally [21:26:00] but it also feels really really problematic to be in a pedagogical role and reinforcing "keep taking the course you're paying me for" with basically bribery [21:26:10] like, the other examples are fun and real too; house price modelling, for example. [21:26:30] That's a good point. [21:27:23] Still, I don't see it as really *that* problematic. It's perfectly legitimate to take classes for the purposes of being able to maximize your earning potential. [21:27:55] Also, it would be legitimate if you found a way to model earning accurately that did not suggest that taking this course would significantly affect your earning. [21:28:07] That answer should be acceptable for full credit, of course. [21:28:49] haha [21:28:56] well, it's not a test question, it was part of a lecture [21:29:05] and yeah, it's totally legitimate for us to have that rationale [21:29:16] I disagree that it's a legitimate pedagogical approach to teaching it, though. [21:29:45] When it comes to mathematics, what makes one example pedagogically more sound than another? [21:30:19] IMO, it's the closeness to loved experience. I want my students to develop an intuition for how their studies describe real things that they are constantly interacting with. [21:30:25] *lived! [21:30:36] But still, that's why I agree that housing prices are better. [21:31:09] word [21:31:55] So, the prof is making a point about capitalism. The prof could use housing prices modeling give people an intuition about income segregation in their own neighborhood too. [21:34:50] agreed, but I don't get the impression it's an intentional point about capitalism [21:35:10] (also this proprietary library uses serialised data files so you have to use their tool, and doesn't work on Python3. You'd hate it ;p) [21:41:13] Bastards! [21:41:23] Why would you ever use non-free programming technologies!? [21:41:29] The free ones are so awesome. [21:41:52] If you're going to use FOSS *anything*, it should be the programming libraries and tools. [21:42:31] https://twitter.com/halfak/status/683399634526404608 [21:42:42] 99% efficiency gain in Wikidata. [21:42:53] Two orders of magnitude! [21:43:50] halfak, and not only that, not only that! [21:43:58] it only supports CSV as an export format! Not TSV! [21:44:07] niiiice! [21:44:22] * halfak lights on fire and becomes the deliverer [21:44:31] TABS ARE FOR TABLES [21:44:36] :D [21:44:39] TAB-les! [21:45:37] Every time I get done with a major refactoring/API design project, I'm always sad that I can't show someone how awesome it is. [21:46:00] There's no way to just hold my hands up and say "look" and have someone *see* what I saw though the last month of really hard work. [21:46:12] well, now you can! [21:46:22] okay, I have successfully outputted it as a CSV at least [21:46:43] I bet that profs files have \r's in them [21:53:33] * Ironholds smh [21:53:42] I need to get better at machine learning so I can teach a better course on it [22:11:40] That's the spirit! [22:12:16] halfak, and unrelated to this entire chain of thought, is there such a thing as the publication of, I guess, position papers in HCI? [22:12:33] Think long, well-researched arguments about meta-HCI. Specifically "stop publishing badly thought-out gorram design patterns". [22:12:34] Yeah [22:12:51] oh good, then maybe I'll start this as a paper instead of an extra-long blog post. [22:12:56] These tend to go to the journal more than the straight up conference. [22:13:01] yeah, makes sense. [22:13:09] But they still do make it to the conference. [22:13:25] I'd advise that you do some basic analysis of a random sample of CHI papers or something to make your case for the problem. [22:13:25] I have a long, long rant about design patterns that don't include anti-patterns and I'm pretty sure I can reference it enough to at least /look/ like a journal article. [22:13:30] I will! [22:13:43] in fact, I will be contacting the authors of those papers and asking them to fill out a survey! [22:13:46] +1 then. CHI loves talking about CHI [22:14:08] specifically looking at the publication of design patterns that lack anti-patterns and the rationale the authors had for not including anti-patterns. [22:14:47] the short summary of my argument is "it is the year of our lord 2016 and if your design pattern for peer-production or pedagogical communities lacks ideas of what to do with toxic users or behaviour it is practically useless and why publish something for the sake of publishing it" [22:15:17] it should go in the same pile as designs for buildings that don't factor in material flammability. [22:15:55] Sounds like an interesting discussion [22:17:53] should I be reading that with British or Midwestern glasses on? ;p [22:18:42] lol no [22:19:06] "interesting discussion" means I can imagine some disagreement and it will be interesting to see how that discussion plays out. [22:19:11] gotcha [22:19:15] and not "that's nice" :D [22:19:29] That doesn't sound too terrible [22:19:30] :P [22:20:23] "doesn't sound too terrible" == basically, it's the best thing I have ever heard and I don't know how to express that, so i'm going to imply that it's at least a little terrible. [22:20:39] \o/ midwest [22:20:50] We've been confusing the NSA and advertisers for years. [22:23:49] hahaha [22:24:00] luckily y'all are so lovely that the NSA doesn't really have to be worried [22:24:29] "I have serious issues with this country's principles. I guess I'll have to frown and quietly say 'that's not how I would've done it' to myself and go home" [22:30:19] "I'll bring it up when I'm shooting the shit with the guys next weekend." [22:32:54] hahah [22:34:37] BTW, "shooting the shit" means talking about inconsequential things -- not necessarily anything to do with guns or fecal matter.