[16:44:46] Research showcase starting in ~15 minutes! [16:44:56] I'll be here to collect questions (or on Youtube) [17:01:54] and we're live! [17:51:59] Question: The main page method is quite ingenious. I wonder how much the conclusions depend on the sample of articles linked from the main page on that day being representative of topics as a whole? (The main page links are usually being deliberately curated by the editor community) [17:52:21] thanks HaeB -- agreed and will pass up [17:55:25] (not a question, but to add for those following along: This paper studied possible bias in this main page selection in another context, for enwiki https://meta.wikimedia.org/wiki/Research:Newsletter/2015/May#Notable_women_%22slightly_overrepresented%22_(not_underrepresented)_on_Wikipedia,_but_the_Smurfette_principle_still_holds ) [17:56:38] yeah the methods here are super smart [17:57:20] Any thoughts on attempting to replicate this for other country blocks e.g., Turkish? [18:00:28] Any attempt to validate the backdoor way of teasing apart direct and latent information demand by comparing to clickstream data for other contexts? [18:00:56] got it groceryheist : i'll see if we have time and try to get them in [18:02:38] Nate TeBlunthuis btw [18:04:05] I was enjoying your IRC name (Nate TeGrotenhuis) :) [18:09:44] (Not a question) The internal traffic by country data has been used to examine censorship before, e.g. https://cyber.harvard.edu/publications/2017/04/WikipediaCensorship https://meta.wikimedia.org/wiki/Research_talk:HTTPS_Transition_and_Article_Censorship [18:11:30] ooh thanks - that is a colossal study [18:16:35] isaacj wow I never updated my irssi config [18:18:47] groceryheist: ahh -- well glad I could point it out then! what have you been working on recently? [18:27:26] The "Effects of Algorithmic Flagging" paper I did with halfak is coming out in CSCW soon [18:28:14] also working on projects about ecological relationships (competition and mutualism) between subreddits. [18:28:29] isaacj: trying to finish PhD [18:30:46] issacj how about you? The knowledge gaps whitepaper was cool! [18:33:05] seems like you're doing some interesting nlp stuff too [19:56:13] oh that's super exciting about the CSCW work! i will look for that :) and Aaron mentioned in his Research Showcase the ecology piece so I've been looking forward to that! I'm a sucker for a good framework [19:56:38] groceryheist: good luck with the PhD! a lot of good work to pull from [19:57:47] yeah, a lot of work around the knowledge gaps white paper. i've learned a lot about how much research and surveys are out there :) [19:58:29] doing a lot of thinking and prototyping of how to make some of our tools language-agnostic. so yeah, taking topic models that had been originally developed for English Wikipedia and making them work for all languages of Wikipedia (by moving from using the text to use article links) [19:59:32] and been thinking a lot about article "importance" and what features you would need to build recommender systems that are more flexible to the different ways in which communities might want to "prioritize" content on Wikipedia for improvement [20:10:40] thanks for the encouragement. Both the language-agnostic topic models and importance models sounds like important stuff