[14:37:28] Hello. [14:43:25] Good morning guillom [14:43:33] * halfak jumps on bike to head to the University [14:43:36] back soon [15:47:40] o/ [15:51:32] hey halfak :) [15:51:39] Hey Ironholds [15:52:12] BTW, still interested in modeling circadian rhythms with pageviews? [15:52:28] http://arxiv.org/abs/1212.5943 - Modeling page-view dynamics on Wikipedia [15:53:43] halfak, ooh, cool! I should reach out to those authors. Thank you for the pointer! :) [15:53:50] my weekend is currently being spent in committees [15:53:56] and I need to poke Nate [15:54:17] Ironholds, was thinking of inviting them to a RG meeting. [15:54:30] If agree, I'll try to get them a slot next week. [15:54:34] *you agree [15:54:52] halfak, love it. We can always kill two birds with one stone? I can reach out about future research and whether they'd like to present, wi'you CCd in (unless you have pre-existing relationships) [15:56:07] I do. I know David. We've been talking about this work on and off for a while. I actually forwarded something about it to our mailing list a few weeks ago. [15:56:20] So maaaybe I'll CC you into our conversation :D [15:57:46] halfak, works for me! :D [15:57:50] I would love to talk to them [15:57:58] you're just trying to get me to attend again. I see your bribery ;p [15:58:10] :D [15:58:21] * halfak wants more Ironholds in his diet [16:00:12] halfak, too much Ironholds may cause earache, tiredness and a desire to go biking. Tolerance to Ironholds builds up over time. Long-term use of Ironholds has been linked to an increase in the body's amount of tolerance of bad puns. Do not eat Ironholds before driving heavy goods vehicles. [16:00:38] Is iron holds a benzodiazepine? [16:00:53] I think it's a stimulant [16:01:13] I mean, given that my diet is z-drugs and nicotine at the moment, it's probably some of both by volume [16:01:23] ha [16:02:09] morning bearloga :) [16:02:57] good morning Ironholds halfak guillom J-Mo! :D [16:03:04] hello hello [16:03:15] morning bearloga! [16:03:19] o/ bearloga [16:04:08] I concur on the tolerance of bad puns. [16:04:32] Although I had an existing condition :p [16:05:37] guillom, was it...punmonary? [16:05:57] :D [16:06:03] I played an AMAZING game last night. It's called Keep Talking and Nobody Explodes. The person at the computer has a bomb they need to defuse in 5 minutes while everyone else has a bomb defusing manual. You know that scene in action movies where the person needs to know which wire to cut? It's that scene but made into a game and with a bunch of different puzzles. [16:06:57] bearloga, okay, but is it better than FTL? [16:08:05] Ironholds: it's different. It's a game about communication and solving puzzles together asymmetrically. [16:08:14] Ironholds: perhaps you're thinking of Spaceteam? [16:08:26] no, I just consider FTL the canonical game of this century ;p [16:08:44] ah, so you ARE mistaken. [16:08:49] okay, noted. [16:09:03] * Ironholds gasps [16:09:08] * Ironholds throws a single glove at bearloga [16:09:16] <3 FTL [16:09:38] bearloga, your man's gotta answer for his words, Burr. [16:09:39] I play it when I'm traveling because my laptop can handle it. [16:16:24] good morning lzia! [16:19:46] hello bearloga. :-) [16:19:54] hi everyone. [16:22:13] Ironholds: I'm pretty sure I've seen papers say "statistical software R" or maybe I'm just loopy this morning [16:22:23] bearloga, yeah, but it's a programming language ;p [16:22:30] it is also software! [16:22:45] but that's like calling Linux 'some code some people wrote'. True, but very broad to the point of practical inaccuracy [16:22:52] also I don't want it to sound like R is SPSS [16:25:12] Ironholds "statistical software and programming language R"? [16:26:44] Statistical software in the form of a programming language and standard library R [16:27:44] FYI, the meeting room is currently occupied, so it might take us a few minutes to get the hangout set up. [16:28:04] halfak: good try [16:28:52] Good grief it's freezing in here. [16:29:13] Ironholds: btw we need to try spaceteam. [16:29:19] bearloga, yessir! [17:02:14] halfak: you coming to this? https://www.google.com/calendar/render?tab=mc&pli=1#eventpage_6%7Ceid-YzZyNHZsdWtzZDYxMmczcG40N3ZwbDlpYmsgYW90dG9Ad2lraW1lZGlhLm9yZw-0-0- [20:20:02] Hey folks. Anyone know someone from Wikipedia Zero who would be interested in working with some researchers to study the effects of Zero? [20:20:14] lzia: https://phabricator.wikimedia.org/T116321?workflow=create [20:20:20] hi halfak [20:20:29] o/ YuviPanda [20:20:33] I saw your question about dockerfile. I think it's useful esp. with the jupyterhub stuff [20:20:47] OK. I'm just not getting it. If you do, I'll be happy to merge. [20:20:57] one problem is that it'll be using versions different from what we'll use in production, but I guess that's ok [20:20:59] Neal is pretty cool, so I'm happy he is contributing :) [20:21:10] YuviPanda, yeah, I think that'll always be true. [20:21:14] halfak: it just solves all versioning problems and what not. [20:21:23] halfak: yeah, just as long as we're aware of that. [20:21:26] Eventually, we should have state-of-the-art revscoring well ahead of production so that we can push stable code. [20:21:59] oh totally. just that it sucks when you are debugging stuff and it ends up being 'oh shit different versions' so knowing it already in top of head lets us rule that out if needed [20:22:16] YuviPanda, +1 [20:22:31] Already living that on a few of the bugs I've been tracking down. [20:22:35] Could you go say what you just said on the PR? [20:22:52] can you link it? gmail's being ownky [20:22:54] wonky [20:22:59] nvm I'll go to github directly [20:23:03] YuviPanda, thanks. [20:23:03] https://github.com/wiki-ai/revscoring/pull/201 [20:23:25] YuviPanda, I had already created another task, wait a sec [20:24:04] BTW YuviPanda, I have a PR for logging config and I've hit a dead end on celery queue sizing (backpressure), so I filed a bug. Main celery maintainer already responded. [20:24:09] YuviPanda: https://phabricator.wikimedia.org/T116101 [20:24:27] YuviPanda, can you merge the two tasks? or let me know and I'll do it? [20:25:36] lzia: ah sure let me do that [20:25:38] halfak: awesome [20:26:11] halfak: you basically 1. make change, 2. rebuild docker image and 3. test [20:26:25] that's fairly fast andn different people do it differently I guess [20:26:58] Do we maintain our own docker image or just the Dockerfile? [20:27:15] YuviPanda I'll be back in 30 min [20:27:16] see ya [20:29:06] halfak: I commented [20:29:09] halfak: just the Dockerfile [20:29:19] Thanks [20:51:39] * halfak bikes back home [20:51:43] back in ~ 45 min [22:06:18] Back looooog [22:06:29] lzia, ^ [22:06:45] * halfak calls out to the non-present Ellery [22:06:58] The backlog calls for you [22:07:13] * halfak refreshes the call [22:07:50] sorry halfak [22:18:10] o/ J-Mo [22:18:15] We should split up this card: https://phabricator.wikimedia.org/T113389 [22:18:30] I'd like to mark that one done and create a new one for our plans re. followup. [22:22:22] let's do it, halfak. mark that one complete and we'll start new ones. [22:23:53] https://phabricator.wikimedia.org/T116339 [22:23:56] J-Mo, ^ [22:27:40] excellent, halfak! thank you [22:28:27] lzia, lololol [22:28:30] :D [22:28:37] you are funny, halfak :D [22:39:12] J-Mo. Dumb idea that might lead to something better: What say we suped up the Teahouse with technological enhancements. [22:39:33] Not just gadgets, but some AI too. [22:39:35] love it. [22:39:47] in a meeting right now, tho. [22:39:52] Snuggle does an OK job of detecting good newcomers. I think we can do way better. [22:39:57] Oh yeah. I'll wait. :) [23:00:53] * halfak waits patiently for J-Mo [23:01:25] I'm in my next meeting… so many meetings… will you still be around in 30 min? Or is that too late? [23:03:24] We'll see. [23:03:29] The ideas will be around tomorrow too. [23:03:31] :) [23:04:14] cool. I'll ping you when I'm done. If you don't answer, I'll ping you tomorrow halfak :) [23:04:30] Good deal. Have a good meeting :) [23:11:17] TIL that you don't need to insert on a wiki page any more. Just having s on the page will automatically display them at the bottom. [23:11:32] \o/ [23:12:03] So basically, you only need for groups, and if you want them to be displayed elsewhere (like the __TOC__). [23:17:02] Indeed. Much better than big red "MISSING " [23:17:59] doing that still feels a bit dirty though ;) [23:29:13] alright halfak. how you feeling? wanna talk TEA!?!?! [23:33:48] So. [23:34:16] I'm imagining a cyborg baby of revscoring, snuggle, teahouse and gettingstarted. [23:34:36] oh lord [23:34:51] I'm listening. [23:34:53] We'd use the Teahouse as the primary social hub (as it already is) and outfit it with integration points. [23:35:32] So we'd work with hosts and/or hostbot to take advantage of the goodfaith/badfaith predictions and detection of negative feedback. [23:35:45] There's probably more we can do just routing newcomers. [23:36:24] I'm not sure how to integrate the gettingstarted tools yet. [23:36:45] But maybe the hosts can use those to direct people towards training or specific types of needs based on the questions asked. [23:37:31] by routing newcomers, you mean using Snuggle/revscoring API to better target GF newcomers, or something more? [23:37:58] More. I want to work out how hosts might take advantage of this and build that tooling. [23:38:13] Right now, HostBot works well with the scale of the teahouse. [23:38:22] But we might get substantial gains by pulling in more people [23:38:30] (say 50% of new editors rather than 10%) [23:38:43] In order to do that, we need to make hosts' work more efficient. [23:39:37] We can use snuggle/revscoring to make sure that we're pulling in the most promising newcomers who are most in need of support. [23:39:56] how do we make Hosts more efficient and preserve the personal touch that setsTeahouse Q&A apart from other forums? [23:40:04] Good question. [23:40:10] That might make this a dumb idea. [23:40:20] But I bet that some of the hosts might have some proposals. [23:40:29] I agree. [23:40:30] Or that by talking to them, we might think of something clever. [23:40:53] * halfak centers is inner user [23:40:54] Also agree. [23:40:57] *his [23:41:07] personally, I'd rather bring in more hosts. [23:41:21] well, not "rather" necessarily, but "also" [23:41:33] Yes. [23:41:47] can we make participating in the Teahouse appealing to more Wikipedians? [23:41:48] So. What makes being a host desirable/not-desirable-enough? [23:41:55] well [23:41:57] We can raise the status of the Teahouse. [23:42:01] another question we can ask the hosts ;) [23:42:02] but [23:42:11] We could get substantial gains just by publicizing that it WORKS [23:42:21] 1. acknowledgement of their work (your point) [23:42:31] Yea. That. [23:42:38] 2. streamlining processes so they do more of the stuff they like, easier [23:42:45] (better tools) [23:42:47] +1 [23:43:20] 3. reducing the frequency of negative events [23:43:50] That's something that we can optimize for. [23:43:57] And I think we can get very good at it. [23:44:08] so, by inviting more GF newcomers (and fewer bad faith ones), we reduce the amount of time they spend dealing with users who they don't want to deal with, but feel OBLIGED to give time to because, you know, it's the Teahouse [23:45:02] the biggest anti-pattern that I want to avoid (and I find myself fighting against this when hosts propose it): [23:45:10] is [23:45:29] generic responses, lots of links to help pages, etc. [23:45:52] I'd like to avoid increasing efficiency at the expense of "service quality" [23:46:20] the whole point of the Teahouse is that no one ever, ever, tells you to GRTFM [23:46:29] +1 [23:47:02] So, FAQs are good. [23:47:09] A revert warning is not an FAQ [23:47:09] NO!!!!!! [23:47:11] BAD! [23:47:15] Really? [23:47:19] well, [23:48:04] So, one thing we can do is have the "Question" textbox in the "post a question" form search past questions. [23:48:06] sometimes a necessary evil? But getting a specific answer from a patient person who's willing to explain something to you in the specific terms you understand, and answer your exact question rather than a somewhat-related one, that's valuable [23:48:23] Yeah. I'm with you. [23:48:42] yeah, that's been floated. Not my favorite idea, unless... [23:48:47] We need to do alchemy. [23:48:53] Turn lead into human empathy [23:49:06] So, I don't want to suggest an FAQ so much as the idea. [23:49:14] unless you are provided with a list of related questions that you are WELCOME to peruse, but you are still encouraged to ask the question yourself [23:49:19] I think the delivery of an FAQ is awful. [23:50:07] no one likes those awful customer service phone trees that try to shoe-horn you into one of 6 options, and that only grudgingly connect you to a real person when you press 0 three times [23:50:08] A canned answer that's been carefully edited, delivered in context, and (maybe) customized to the situation is a different kind of delivery. [23:50:29] What if the options in the call try were built by the people who answered the calls though? [23:50:29] Newbie Knowledge Graph (TM) [23:50:41] I used to work in one of those call centers. [23:50:49] I would have designed a better call tree. [23:50:51] that would improve performance, but still wouldn't be a person [23:51:32] Maybe queues or a rubric would be better. [23:51:40] I DO like this basic idea of leveraging previously asked questions, don't get me wrong. I just want to make sure we do it in a way that doesn't alienate [23:51:51] E.g. when writing a paper acceptance or rejection, I have a list of things to make sure to say. [23:52:07] That helps me write a response much faster. [23:52:23] Also, hey! Maybe we can help people figure out a particular situation. [23:52:25] hosts probably have their own heuristics. that would be something we could ask them, and get valuable responses. [23:52:39] oh!!!! [23:52:41] idea [23:52:44] Maybe the trouble isn't really in drafting a response, but trying to figure out more about the context of the question. [23:52:46] so, keep the Teahouse as it is [23:53:21] and use the database of previously asked questions to build a UI Q&A widget that anyone can use, any time. [23:53:52] wherever you are on the site, you type in your query and it probabilistically serves up a set of related Teahouse questions. [23:54:24] "why was my article rejected?" [23:54:55] ^^probably top Teahouse question of all time [23:55:06] Yeah. I like it. [23:55:19] Basically, we'd make it trivial to lurk in the Teahouse. [23:55:20] and you can rate the answers you get [23:55:27] so that the system learns over time [23:55:44] * J-Mo likes to pretend he understands ML [23:56:56] The ratings would also provide feedback/motivation loop for hosts. [23:57:01] Also, scary stuff. [23:57:03] in parallel, we refine our selection process for delivering Teahouse invites to a larger set of GF newbies, and also start delivering these invites as soon as someone meets a threshold (rather than a daily cron job) [23:57:24] oh god, we invented Karma [23:57:30] J-Mo, that could work better. I don't think we need to wait 24 hours. [23:57:36] 30 minutes would work for most people. [23:57:59] We might find that inviting people earlier is desirable too. [23:58:04] true [23:58:25] If we can catch them before their first edit session is done, we might see substantially higher response rates. [23:58:33] yes! [23:58:43] Even just so people know that when they come back, there's a Q&A forum for them. [23:58:50] Why are we doing this in MediaWiki again? [23:58:59] LOL [23:59:02] right? [23:59:33] * J-Mo looks around nervously… tries to convince himself that it's not a cabal if it's happening in an open chan