[14:29:19] o/ kevinbazira [14:29:46] I was just looking and it doesn't seem like your IRC cloak has been created. [14:30:17] Sure? 0_0 [14:30:19] Could you disconnect and reconnect from freenode to see if is applied? [14:31:04] I want to invite you to a few channels where a cloak is a sort of soft requirement required. :) [14:39:59] Hmm. Still didn't work. [14:40:03] It's probably OK. [14:40:17] I just sent an invite. [14:40:21] kevinbazira, ^ [14:43:40] Thanks for the invite! [14:44:07] halfak: did you want kevin to get access to analytics-privatedata-users too? [14:44:16] that's the main hadoop access one [14:44:24] or not necessary for now? [14:44:38] ottomata, hmm. If we're moving scoring data there we should probably add him to that one too. [14:44:45] ya prob good to just get it [14:44:46] will add it [14:44:58] And if I forgot that for kevinbazira, I probably forgot that for accraze too. [14:45:25] 10Scoring-platform-team, 10LDAP-Access-Requests, 10Operations, 10SRE-Access-Requests: Grant LDAP groups and deployment shell access to Kevin Bazira - https://phabricator.wikimedia.org/T234209 (10Ottomata) [14:45:31] 10Scoring-platform-team, 10LDAP-Access-Requests, 10Operations, 10SRE-Access-Requests: Grant LDAP groups and deployment shell access to Kevin Bazira - https://phabricator.wikimedia.org/T234209 (10Ottomata) [14:56:21] o/ groceryheist [14:56:29] o/ halfak [14:56:30] So re. the null - non-adoption hypothesis. [14:56:34] yeah [14:56:40] i think both steps are hard [14:56:44] What kind of interesting insights do you think we might glean? [15:00:48] so I can see at three plausible explanations: 1. the gap between what designers have in mind and what's actually done in implementation and use 2. The ability of communities to find their own solutions to their problems in open collaboration means that designers don't have a clear view of their needs (this is complicated by the fact that at least some members of the communities obviously wanted [15:00:54] this tool) 3. A tool might be adopted for reasons that won't translate into the outcomes we expected to measure. [15:01:44] 2. is the "ecology of tools" idea [15:02:29] I don't think this is super novel, but more reinforces some classic CSCW lessons but in a different context and technology. [15:06:26] 1. is like what we found yesterday about watchlist preferences. If a team building an AI doesn't have a clear picture of when it's predictions will be made visible that limits their ability to predict whether it will be used. [15:07:22] BTW: It's also on my list to run the analysis for the huggle-only cutoffs. [15:22:55] huggle-only cutoffs? [15:23:43] I think there's something to be said about the impact of algorithms. It's mediated by UI implementation and adoption. Future studies of the effects of AI need to explicitly account for UI and adoption patterns before the effect of the AI can be explored. [15:23:53] groceryheist, ^ [15:24:06] Sorry for the delay. I got pulled into some other work. [15:31:17] sorry I don't mean "huggle-only." I ores-only, no rcfilters. [15:31:39] " I think there's something to be said about the impact of algorithms. It's mediated by [15:31:42] UI implementation and adoption. Future studies of the effects of AI need to explicitly [15:31:45] account for UI and adoption patterns before the effect of the AI can be explored. [15:31:48] " [15:31:50] ^ yeah that's good [15:33:51] groceryheist, that argument leads to the critique of "algorithmic bias" explorations not accounting for the whole ecology. But this is more basic. The algorithms *entrance* into the ecology is mediated (by UI and adoption) and that mediation is important. [15:34:12] *the critique --> your critique [15:34:14] :) [15:49:33] o/ accraze [15:49:36] Good morning :) [15:49:52] mornin halfak [15:50:06] We have a staff meeting today so we should probably do async [15:50:12] ^ standup [15:50:18] yeah just saw that on the calendar [15:50:33] sounds good [15:52:01] Good morning accraze :) [15:52:47] good evening kevinbazira! [15:54:15] halfak: yeah let's keep thinking in that direction for now. [16:04:40] 10Scoring-platform-team, 10Analytics, 10Analytics-Kanban, 10EventBus, 10Patch-For-Review: Change event.mediawiki_revision_score schema to use map types - https://phabricator.wikimedia.org/T225211 (10Nuria) a:03Ottomata [16:06:03] -- Async Standup -- [16:06:04] Here's my update [16:06:13] Y: All Jade stuff-- Cleaned up warning handling for all api modules, updated comments, finally fixed all codesniffer errors, fixed a merge conflict and rebased against master...BUT now the Jenkins build is failing (quibblevendor/nodejs+grunt/decreased coverage??) [16:06:29] T: Fixing Jenkins build, I'm getting some errors related to stuff I didn't touch, so need to figure out what's going on. Hoping to get this all sorted out today [16:23:44] Y: 99.9% done with this tag of session orientation. I just want to clean up a couple of things before I ping accraze for another round of review. Otherwise, I did some work sending a welcome email for Kevin, got some data for Nate, and went to a bunch of meetings. [16:24:44] T: I'm reaching out to UI standardization and the Front-End Engineering group to get their recommendations lined up for Jade. If I get that far, I'll get back to turning Jade wireframes into mocks. [18:17:03] halfak: the models finished scoring the edit samples so I'm unblocked to run the threshhold analysis [18:58:33] \o/ awesome [20:29:26] ottomata, what's the URL for the score event stream? [20:31:59] Also what do you think about collaborating on a blog post about it? [20:38:33] I just worked out the URL after trying a bunch of things: https://stream.wikimedia.org/v2/stream/revision-score [20:39:24] 10Scoring-platform-team, 10editquality-modeling, 10artificial-intelligence: Why is jawiki's goodfaith model so bad? - https://phabricator.wikimedia.org/T230953 (10Keegan) As discussed with halfak on IRC, we're not aware of much inter-editor communication or coordination around counter-vandalism work at jawik... [20:44:21] I really want to get the word out about this. I think it's pretty cool. [20:44:36] xinbenlv, you might be interested in this too. [20:44:45] It's a stream of ORES scored edits. [20:45:04] This'll make it so that you don't need to do an ORES score lookup for every edit. [20:58:19] awesome [20:58:30] good to know [21:29:54] Arg. I couldn't get the last tests for the session PR done in time after all. I ran into a weird issue (which is why we have tests!) [21:30:39] I'll need to finish this up in the AM. [21:30:40] o/