[15:33:36] ottomata, hmn? [15:34:06] ottomata1, you rang? :) [15:34:30] halfak, I've discovered, finally, the only thing in the world longer than the thesis you wrote [15:34:35] the thesis you're reviewing [15:34:47] if I vanish someone please just grab my feet and drag me out from the pile of paper and empty pens [15:35:12] Ironholds, someone reads & reviews theses? [15:35:21] ;) [15:35:32] me [15:35:40] should I just be like "oh it's fine" [15:35:44] or "I AM NOT CITED ENOUGH BOO" [15:49:50] the latter's a common gut reaction... why u no cite me papers? [15:51:40] In non-blind review, I have no concerns about saying "Cite my work" [15:51:57] hopefully I'll one day be halfak and it becomes an author-problem instead "my gosh, do I have to cite them _again_?!" [15:52:20] :P [15:52:21] In blind review, I'm conflicted because the person receiving the review can't take your advice with the appropriate level of salt. [15:52:54] halfak: agree on that, I'm having that problem right now, actually, and it's an interesting one [15:52:54] halfak, see also the AFT problem? :p [15:54:02] Nettrom, yeah, we ran into it about 3 years back; either Cliff Lampe or one of his students were reviewing a paper of ours [15:54:02] I'd like to hope student [15:54:02] but it was like "HERE ARE 20 LAMPE PAPERS USE ALL OF THEM OR ACCEPTANCE GETS IT" [15:54:02] :P I think that was good advice though. [15:54:03] that citation suggestion was relevant [15:54:15] yeah, but it looked hella-weird [15:55:57] Ironholds, I agree. It's always a little funny to get a recommendation to cite a paper during a review process. [15:56:15] But given the papers that fail to build on previous work, I think it's good that we maintain the practice. [15:56:27] Or even engage in it more aggressively. [15:56:54] yeah [15:56:56] hey J-Mo :) [15:56:59] Generally, I'll suggest that papers I am reviewing cite other people's work. I like to leave it to other reviewers if they thing that my work should be cited. [15:57:12] hi Ironholds! [15:57:13] That's worked OK so far :) [16:25:18] Nettrom: around? I was looking at your "The Success and Failure of Quality Improvement Projects in Peer Production Communities" and it's super great. [16:29:22] ^ \o/ [16:30:30] ragesoss: thanks! [17:23:51] darnit, I wanted J-Mo [17:54:24] If anyone's interested, here's a diff that takes FOREVER to process: https://en.wikipedia.org/w/index.php?title=December_2004&diff=prev&oldid=8659925 [17:54:34] And by FOREVER, I mean longer than 10 seconds. [17:56:18] ahhaha [17:56:43] ottomata, you pinged me yesterday or earlier! What's up? [17:56:47] mountaingoats? [17:56:52] morning DarTar [17:56:53] leila: back at my desk [17:56:56] hey Ironholds [17:56:59] o/ DarTar [17:57:03] yo [17:57:05] oh Ironholds, I thin I emailed you, about killing your job [17:57:10] oh, gotcha [17:57:12] yeah [17:57:14] YOU"RE FIRED [17:57:20] "killing your job" [17:57:22] am I? Oh. Okay. [17:57:26] haha [17:57:27] here is my volunteer agreement [17:57:33] accepted! [17:57:41] ok i am HUuuUNgry, lunchtime! [18:03:51] hi DarTar. [18:03:55] and everyone else. :-) [18:04:02] hey [18:04:54] o/ Leila [18:05:18] DarTar, let's figure out what to put as ETL for the link recommendation. some background: we're aiming for a paper submission on 5/8, which means the work should be at good shape at that point, however, we have not found the silver bullet, yet, we're experimenting with the algorithm as we speak, and we're finding ways to test how good it works. [18:06:02] it's a complicated stage, in the sense that there are a lot of complex data extractions that need to be done, and a lot of intense research. [18:09:09] what I'm concerned about now is that if we say 5/8 as ETL, and by then the algorithm doesn't get to the point that is ready for production, we have promised something we can't deliver. and I'm leaning towards what halfak always says. it's hard to commit to deadlines for research. [18:09:49] ETL? [18:09:51] ETL == Extract Transform & Load? [18:09:54] ETA [18:10:04] we can definitely commit to have the research "done" by mid-May, but "done" is very subjective. [18:10:08] sorry, Ironholds, yeah ETA [18:10:23] would more brains help? I'm sure we know additional algorithmic people interested in this [18:10:25] +1 for "done" being hard [18:10:34] WHat about "something useful" by 5/8? [18:10:36] we can call something "done" by being happy with the current state of the algorithm if we want to. but I'm not happy with that. [18:10:49] something useful == "a result we can make use of" [18:10:57] or a dataset we can use for other things [18:10:58] leila. sorry – I’m back, fighting with wi-fi [18:11:02] if you take it literally, that works halfak. [18:11:06] np, DarTar_clone [18:11:33] now imagining a tentacular router trying to kill DarTar_clone while he fights it off with a broadsword [18:11:48] ;-) [18:11:48] leila: wanna talk on a hangout? [18:11:51] * halfak eyes up J-Mo, J-Mo1 and DarTar_clone with concern about the coming clone revolution [18:11:58] Ironholds: wait until my clone^2 appears [18:12:01] sure halfak. will send you a call now. [18:12:09] order 66! ORDER 66. [18:12:15] leila, ? [18:12:56] For Science! [18:13:42] I for1 welcome our new J-MoBot overlords [18:13:55] Hey kim_bruning :) [18:14:05] Long time no see. [18:15:22] sorry halfak, I meant DarTar. :-\ [18:15:36] I'll update you halfak with the conclusion. :-) [18:15:38] Gotcha. :) [18:16:07] leila, looking forward to having better communication around "delivered" research products. [18:16:15] :) [18:16:31] totally halfak. it's complicated. [18:16:56] * kim_bruning waves to guillom [18:17:08] _o/ [18:18:59] \o_ [18:19:05] halfak: leila: Do you think it's because we're doing "softer" science here? Based on my experience, my instinct is that it's easier to assess success/failure with harder sciences, but I'm interested in hearing your thoughts. [18:19:26] (This can be a debate for later ;) [18:20:52] I don't think that dividing between "harder" and "softer" is productive [18:21:33] the criteria for success in any study following the scientific method, regardless of field, is the same: you have proven or disproven the null hypothesis to the levels of certainty that are expected. [18:21:46] that level of certainty might change, but the criteria does not. [18:21:54] guillom, I don't think so. [18:22:32] To add to what Ironholds says, the study isn't done when the experiment or analysis is complete. [18:23:01] Often we see things we don't expect when we look at the world in new ways. [18:23:02] Ironholds: good point. I think I'm simply always less comfortable with the uncertainty that comes from studying humans. [18:23:32] Meh. Is nueroscience a hard science? [18:23:38] yes :-P [18:24:10] Well, in neuroscience they study populations whose individual behavior is poorly understood using very similar methods to ours. [18:24:51] To me, it's all a hard science is you use controlled experimentation. [18:25:10] word [18:26:38] My point is basically: give me electrons and chemical molecules any day, and their behavior is usually more predictable than humans'. But I do acknowledge that this is based on my personal background, and I'm not saying it's a universal truth. That's what I was asking for opinions from people more familiar with this field. [18:26:53] yeah, I think that's fair [18:26:57] guillom, electrons are not predictable :P [18:26:58] but: how to phrase. [18:27:01] * Ironholds thinks [18:27:20] https://en.wikipedia.org/wiki/Uncertainty_principle [18:27:37] guillom, I guess the reason I react so strongly is because I have never felt that even dividing "hard science" and "soft science" is useful. And most of the time when I see it being used, it's being used by physicists to devalue chemists, or chemists to devalue biologists, or...all the way down. [18:27:52] ^ this [18:28:29] the thing that divides sciences is not hardness, or softness, or fields, it is methodologies, and many methodologies are shared between fields because they work in a variety of areas, and that sharing is undermined when we start segmenting the population by what you study [18:29:02] are you interested in increasing our knowledge of the universe, in the macro or in the micro? [18:29:12] Do you ask "why?" a lot? [18:29:18] ħ/2 is a level of uncertainty I can live with :p [18:29:20] then you are a scientist. Etymologically, culturally, and practically. [18:29:38] and everything else is just window dressing that evolves from humanity's weird mix of hierarchy and sentience [18:30:02] Ironholds: That's fair. [18:30:09] here comes the feminist theory of science [18:30:19] the correct theory of science ;p [18:30:23] well, a correct theory of science [18:30:30] Ironholds, :P [18:31:00] Ironholds: I used a bad shortcut to express my point. Thanks for calling me on it. [18:31:02] But seriously, there's some good stuff there. c.f. https://en.wikipedia.org/wiki/Donna_Haraway and https://en.wikipedia.org/wiki/Sandra_Harding [18:31:14] I do hope my point wasn't lost though :) [18:31:16] I need to read more Harding [18:31:30] "She has developed the research standard of “strong objectivity,” and contributed to the articulation of standpoint methodology." [18:31:32] guillom, it wasn't! And no problem: if I screw up (read: when I inevitably screw up, because I will, unceasingly) I am equally happy to be called on it [18:31:48] halfak, recs? [18:31:56] recs? [18:32:34] recommendations! [18:32:43] we clearly need y'all to build a /paper/ recommendation system next [18:33:16] "you liked Harding, you may enjoy: destroying the systemic kyriarchy" [18:33:26] Ironholds, that's something the GL lab was working on for a few years, but the lack of open data made it really hard. [18:33:33] Publishers hide their metadata [18:33:34] :( [18:33:36] :( [18:33:44] I mean, if you need someone to semi-illicitly scrape all of the things... [18:33:45] Recommendation: https://books.google.com/books?hl=en&lr=&id=MIQKZMyLJ0UC&oi=fnd&pg=PA298&dq=Feminism,+Science,+and+the+Anti-Enlightenment+Critiques&ots=BNcLe_zHHo&sig=asH2nD1hTZ5Jv62Oh7GewOJ374U#v=onepage&q=Feminism%2C%20Science%2C%20and%20the%20Anti-Enlightenment%20Critiques&f=false [18:33:54] thanks! [18:34:00] I may just get the entire book for ease [18:35:04] I borrow a lot of ideas about "enlarged perspectives" and "more complete objectivities" from Harding. [18:36:09] I guess the core of my point is that, if my research involves humans, I never feel that I really control the experiment, because there are so many variables, and so I never have a lot of trust in the data that results. But like I said, this may simply be because I'm a trained physicist, and that I lack the proper tools and methodologies for the study of humans. [18:36:33] a lot of the tools are actually the same [18:36:44] for example, DarTar insists on calling me a sociophysicist [18:36:47] mostly because it pisses me off [18:36:56] :D [18:36:58] It turns out that when you look at populations of people, they are about as predictable as collections of atoms are. [18:37:05] but also because I do a lot of work that involves applying "how the universe's building blocks work" methodologies to meatsacks [18:37:24] The fun thing is when the population has the potential to behave chaotically/complexly [18:37:27] but as halfak says, really both are fundamentally chaotic systems on any real level of scale [18:37:36] our P value is 0.95, same as anyone else's. [18:37:42] Then you run into issues in the same way as you do in physical chaotic systems [18:38:01] (he's about to start talking about maximum animal size and how mathematically beautiful the structures are. It's a rly good speech) [18:38:03] hmm. Interesting. [18:38:13] halfak, sometimes I think we spend too much time together. This is one of those times. I am actually okay with that ;) [18:38:14] https://en.wikipedia.org/wiki/Three-body_problem [18:38:20] ^ chaotic physical systems [18:38:28] halfak, so we decided that the deliverable is the research report and the data-set, the implementations and further improvements can be done after that if needed. [18:38:47] Ironholds, I never see you! :P We need to sit down and have a beer. I think it has been more than a year! [18:38:56] close to a year! June! [18:39:03] leila, that sounds good to me. [18:39:09] I think it makes a lot of sense too. [18:39:15] I'll see if *muttermutter* can fly me to MSP. There are a lot of people interested in *muttermutter* out there. [18:39:17] yeah, halfak. agreed. [18:39:33] what is muttermutter Ironholds? [18:40:07] http://www.carnegiehall.org/Calendar/2014/11/18/0800/PM/Anne-Sophie-Mutter-Mutter-Virtuosi/ [18:40:11] :D [18:40:20] (not actually related) [18:40:31] leila, tell ya friday, probably! (like: if I actually am allowed to tell people) [18:40:32] How do you make a trollface ascii emoticon? [18:40:42] it's an org I've been asked to do some work with and they want me to do a few talks [18:40:48] :D [18:40:57] I should try and make the talks coincidentlaly overlap with "places I wanna go anyway" [18:40:59] (I was gonna ask how it's related) [18:41:45] like what kind of work Ironholds [18:41:55] Software engineering! Not research, alas. [18:42:20] ow okay. that's cool! [19:05:19] wait, we're talking about Haraway and Harding in here now?!? [19:05:23] :-D [19:07:58] halfak, on twitter you will find a python3 fanatic who sits on the boards of...a lot of orgs. [19:08:00] you two should chat :) [19:08:19] ragesoss, totally :) [19:08:30] Ironholds, cool! Who is it? [19:08:44] paultag! He's in your mentions [19:19:58] I thought you two would get on ;p [19:29:59] Ironholds, do you use exclusively python3? [19:30:14] Yup he totally does! :D [19:30:14] I do not use Python at all [19:30:20] Awww too late [19:30:27] I consider Python too slow ;) [19:30:28] kim_bruning: halfak does tho [19:30:30] oh okay :-/ [19:30:35] only r? [19:30:38] C__ [19:30:39] *C++ [19:30:42] (I do too wherever I can) [19:30:53] C++? [19:30:58] wait what? WHY? [19:31:11] you don't strike me as a C++ kind of dude [19:31:19] I like C++ [19:31:21] have you grown a beard? [19:31:23] :-P [19:31:32] it's whip-fast, mostly rational and not as fanatically OOP as Java [19:31:53] those things are all true, but dev time is longer than for anything starting with p [19:32:01] (except maybe php ... boy has that changed :-p) [19:32:11] anything starting with p or r even ;-) [19:32:19] but ... interesting to know :-) [19:32:33] ever looked at anything functional too? [19:33:09] yeah, I've been investigating Haskell of all things, since it does what I consider my ideal language would do [19:33:17] (lazily evaluated, functional, compiled, strictly typed) [19:33:38] and dev time for C++ is indeed higher, but I can write a URL decoder that'll process 1m URLs in 900ms [19:33:44] that's absolutely worth it [19:34:00] I use the code too often for it to not work out in my favour ;) [19:34:17] that makes sense [19:35:21] (plus: most of my code is for reuse, so it stacks up) [19:41:53] Ironholds: the cluster is mostly ok, i just haven't sent a 100% go email yet because I haven't fully finished healing everything [19:41:59] but, you should be able to submit jobs without hurting things [19:42:04] neat! [20:23:24] hey, I after much tribulation and compiling missing packages, I got Wiki-Class to run and evaluate an article! [20:23:34] halfak: ^ [21:54:36] hey halfak, did you see the announcement about the new built-in edit tagging feature? https://lists.wikimedia.org/pipermail/wikitech-ambassadors/2015-April/001162.html [21:55:10] It's the same feature as the one used by AbuseFilter or VE/HHVM etc., but the new ones can be added/removed by editors [21:55:12] I didn't see the announcement, but I've been following the phab tasks. [21:55:17] ok :) [21:55:45] oh right, didn't see you in the long cc: list [21:55:51] I think it's great and I look forward to finding ways to take advantage of it. [21:56:00] e.g. posting revscores as tags -- maybe. [21:56:09] yeah, just wanted to make sure you were aware of it [21:56:20] Depends on what tags can afford. I'm assuming key-value pairs are a lot to ask for. [21:56:40] Can always use restbase too :) [21:57:58] I haven't really looked into RESTbase too much yet. [21:58:18] Is it basically one gigantic and fast data store? [21:58:29] that sounds like a nice tag line ;) [21:58:30] Basically, yes. [21:58:36] :) [21:58:38] o/ gwicke [21:58:39] heh :) [21:58:51] Someone has restbase as a ping word [21:59:00] My thoughts as well! [21:59:00] ummm ;) [21:59:10] I have "science" a ping word :) [21:59:11] nobody expects the Spanish inquisition! [21:59:52] I was wondering about the revscore status [21:59:52] There was a phab task about " Computer science " that trolled me for a couple weeks. [22:00:01] héhé [22:00:25] halfak: https://phabricator.wikimedia.org/T95551 [22:00:38] getting ready to deploy that [22:00:40] gwicke, just running off to a meeting. TL;DR: the prototype webservice is up. We're working on the manual "revision coder" that will allow a crowd to help us label revisions to train models. [22:01:11] gwicke, http://ores.wmflabs.org/scores/enwiki?models=reverted&revids=345678|5678947 [22:01:13] added you in the CC list [22:01:31] Woot! I need a cache :) [22:01:35] o/ [22:01:38] * halfak --> meeting [22:01:58] halfak: shiny! [22:03:38] We're an odd bunch, those of us whose reaction to seeing JSON is "this is cool!" :) [22:18:22] I like JSON :( [22:43:17] halfak: are you implementing a revscore webservice for Wiki-Class too? [22:51:25] ragesoss, yes :) [22:51:32] I figured it would be useful for wikied :) [22:51:37] And lots of WikiProjects :) [22:51:49] I could have that up this weekend if someone really wanted it. [22:52:03] I really want it! [22:52:48] halfak: btw, the new model is too big for Github, is there somewhere I should put it so you can get access to it? [22:53:02] (in case you'd want to use the new one in your running code) [22:53:30] halfak: have you done any thinking on how to linearize article quality? Like, how to represent fractional improvements between revisions that don't change the most likely score, but push the values up a bit for higher scores and down a bit for lower scores? [22:55:00] to be able to graph the predicted quality over time for an individual article, across its whole history. Or a simple-to-understand before/after for a diff. [23:00:08] ragesoss: I've thought at lot about that, since it's something I've wanted to use the classifier for. There's the concept of "stochastic dominance" that could be used, to allow for a finer-grained scale, but I'm unsure if the classifier is accurate enough to do it [23:02:43] ragesoss, yes. I have good measures of that :) [23:02:57] https://meta.wikimedia.org/wiki/Research:Measuring_value-added [23:03:50] I can apply these measure on a small scale today. Soon I'll be able to apply them across the whole wiki. [23:03:51] :) [23:04:00] So, class of 100, no problem [23:04:06] * halfak runs to next meeting [23:04:20] * Ironholds does the same \o/ [23:05:51] halfak: you've been doing all the things!