[00:14:29] 10Scoring-platform-team (Current), 10MediaWiki-extensions-ORES, 10MW-1.31-release-notes (WMF-deploy-2018-02-06 (1.31.0-wmf.20)), 10Patch-For-Review, 10User-Ladsgroup: Clean up ScoreLookup implementations - https://phabricator.wikimedia.org/T185534#4033952 (10Ladsgroup) Yup [00:18:46] \o/ [00:20:04] 10Scoring-platform-team (Current), 10articlequality-modeling, 10User-Ladsgroup, 10artificial-intelligence: Article quality campaign for Persian Wikipedia - https://phabricator.wikimedia.org/T174684#4033971 (10Ladsgroup) It looks great, I just finished a workset. I just need to mark the old campaign as inac... [16:35:34] ARGH [16:41:14] I don't have the paper I meant to write. I have background for a paper. [17:02:05] how many days do you have to turn background into an actual paper? [17:03:53] awight, pitch the paper. [17:03:58] That's a great way to make progress. [17:04:03] "I think that ..." [17:04:20] * halfak wants to read the pitch [17:04:38] The thing is... what I wanted to write is not about "political economy" [17:04:46] 10Scoring-platform-team (Current), 10MediaWiki-extensions-ORES, 10MW-1.31-release-notes (WMF-deploy-2018-02-06 (1.31.0-wmf.20)), 10Patch-For-Review, 10User-Ladsgroup: Clean up ScoreLookup implementations - https://phabricator.wikimedia.org/T185534#4035643 (10Halfak) 05Open>03Resolved [17:04:48] I ended up making a huge number of notes about the pol econ [17:04:55] but they don't fit. [17:05:02] together [17:05:52] does anything interesting jump out of them, 2-3 points maybe? [17:17:13] 10Scoring-platform-team (Current), 10articlequality-modeling, 10draftquality-modeling, 10editquality-modeling, 10artificial-intelligence: Re-train models with revscoring 2.2.0 - https://phabricator.wikimedia.org/T188775#4035669 (10Halfak) There was an issue in the makefile and then a reboot on Weds, so I... [17:17:23] +1 for picking a subset [17:17:56] Also, ask yourself what thought process lead you to write down your notes in the order you did. [17:18:20] Maybe there was a natural progression in the lit. Maybe you have a loose structure you don't like, but has some good qualities. [17:19:13] will share shortly [17:20:22] kk [17:30:36] https://github.com/adamwight/jade-pol-econ/blob/master/jade-pol-econ.pdf [17:30:40] halfak: ^ apergos if you're curious [17:30:51] Meanwhile, I'm gonna go gnaw my hands off for lunch. [17:32:23] back shortly meeting [18:20:20] back [18:27:31] so my q awight after the world's quickest skim is, what do you want the reader to come away with? "I looked at X, asking the question Y, came away with toughts A,B,C, and future is Q" something like that right up front to frame it [18:28:12] I feel like once you hve decided on a framing that works for the work you have or can easily complete, the rest will 'fall into place' so to speak [18:30:25] also your dumps patch is applied to master so you are nublocked for more vagrant role testing (thank you in advance) [18:39:45] apergos: Thanks for the suggestions! [18:40:22] mostly I'm just intereste in what the frame is, so I can digest your content :-D [18:40:51] You're right, that the frame is exactly what I don't have yet. [18:44:53] * halfak clicks on PDF [18:46:26] Starts too quick. Would be more natural to start with something like: Wikipedia is a big 'ol wiki, Open wikis have quality control problems, Wikipedia needs to do QC at scale, AI is really useful for making AI scale, ... [18:48:10] Also, I think that bias in AI needs to be introduced more deliberately. E.g. by taking an AI-bias paper and summarizing it. E.g. I know Solon likes to talk about different origins of bias -- the dataset, the algorithm, the system around the algorithm [18:49:15] What I need to tie this together is to explain how what we're doing challenges the status quo in commercial AI [18:49:24] Pacing comments are great too, thank you [18:49:35] * halfak attempts a re-outline [18:49:58] awight, might find something useful in discussing the power disparity [18:50:10] I like to talk about JADE as a means to empowering AI users [18:50:30] Which naturally disempowers AI developers [18:50:43] In some ways anyway [18:50:51] +1 I want to talk about how status quo auditing, if any, treats humans as data entry automata [18:51:09] how so? [18:51:59] It's completely alienated labor: some % of users will see a simplified feedback form "did you like your machine prediction? Y/N" [18:52:13] The purpose is entirely to gather feedback for the AI [18:52:26] JADE is intentionally empowering an existing community [18:53:20] Ahh yes. I like it. [18:54:13] Hmm.. The money section seems strange and out of place. What is its purpose? [18:55:01] True that. The "resource lifecycles" section (Money, Knowledge, Labor) is my background research, trying to map out exactly what the material flows look like. [18:55:11] None of it will be in the thing I'm sending on Saturday [18:55:20] :-D [18:55:27] max length for Saturday is 1 page, anyway. [18:55:30] OK I'll skip that for now. [18:55:34] :D [18:55:39] Oh that'll be helpful. [18:55:53] Helps to know how to limit scope. [18:56:03] I feel like the Resource Lifecycles is something I still want to pursue later, in some way. [18:56:16] I also realized that there's some important work to do in analyzing paid editing. [18:56:21] Which we'll be able to do! [18:58:45] From the material angle, there's just making good estimates of how much paid editing is happening, how much goes unreverted, and estimating how much money is flowing into the paid editor pockets. [18:59:31] Then there's more subtle stuff like which articles are being paid edited. [19:00:10] Seems like things are getting off topic -- at least so far as JADE is concerned. [19:00:54] I don't know how to make a good political econ argument. But I could help fashion a power-related JADE argument. [19:03:36] That is solid. [19:06:22] OK cool. I'm going to work in an etherpad [19:06:51] https://etherpad.wikimedia.org/p/jade-pol-econ [19:17:34] OK I like that outline. [19:17:41] Want to give it a shot awight> [19:17:44] *? [19:18:01] Very helpful, thanks [19:18:41] :) [20:16:21] halfak: I don't see any literature about "fairness enforcement", maybe there's a synonym? [20:17:48] e.g. https://www.researchgate.net/profile/Toshihiro_Kamishima/publication/220766348_Fairness-aware_Learning_through_Regularization_Approach/links/590201894585156502a2e679/Fairness-aware-Learning-through-Regularization-Approach.pdf [20:17:55] Also http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.442.6732&rep=rep1&type=pdf [20:22:17] ty [20:28:48] wiki-ai/revscoring#1447 (w2vfix - 526ea83 : Sumit Asthana): The build failed. https://travis-ci.org/wiki-ai/revscoring/builds/351007844 [20:30:13] wiki-ai/revscoring#1448 (w2vfix - f03f410 : Sumit Asthana): The build passed. https://travis-ci.org/wiki-ai/revscoring/builds/351008357 [20:32:13] where's codezee [20:32:16] :P [20:33:26] halfak: that was the fastest ever pr/review cycle! :D [20:33:32] ha. [20:33:43] Got the IRC ping and am working on emails :) [20:38:15] i ruined it :/ [20:38:19] the moment [21:03:44] 10Scoring-platform-team (Current), 10articlequality-modeling, 10draftquality-modeling, 10editquality-modeling, 10artificial-intelligence: Re-train models with revscoring 2.2.0 - https://phabricator.wikimedia.org/T188775#4036469 (10Sumit) with wordvectors blockers now cleared, building drafttopic model on... [21:21:15] halfak: Feed like re-reviewing? [21:21:22] linky [21:23:22] ah same link [21:23:38] https://github.com/adamwight/jade-pol-econ/blob/master/jade-pol-econ.pdf [21:38:36] Oh that. Sure. I'm working on Annual planny presentationy things. [21:38:39] Will get to it after that [22:24:41] o/ [22:49:28] Arg [22:49:31] * halfak clicks on pdf [22:52:31] No strong opinions except for maybe to tone down the "us" and "them" [22:52:51] E.g. we can make "us" specific [22:53:16] us = users and those affected by an AI [22:53:33] Non employees/executors [22:55:34] awight, ^ [22:57:03] awight: is it intended that the boxes aren't closed at the right side? [22:57:57] maybe clarify what AI is at the introduction: Artificial intelligence (AI) has become… [22:59:25] add a [1] in the text when refering to Christian Sandvig work? [22:59:25] El búfer 1 está vacío. [23:00:00] I find odd the use of "For the moment," [23:00:52] should "agency" be "agenda"= [23:00:53] ? [23:01:57] is the use of contraction "we’ve" appropriate for a paper? [23:03:17] could you explain in some more depth how does the Aircraft Certification Service relate to this? [23:03:24] do they specifically talk about AIs?