 Welcome. This is our February metrics meeting. We have a few welcome students organizations. I'm going to read them out, and then we can all clap for them. We have Jorge and Maria, who are both conversions and full time joining us. And then we have a few new interns who get to be here and learn with us. We've got Andrew Metab and Aaron in Legals. So moving on, we've got a whole bunch of anniversaries and some anniversary stuff. And then I particularly wanted to call out some pretty major milestones that we've hit in January. So we did a release of the Android app by the Mobile Apps team, which is amazing. Coms has published an annual report, and it's the first time that we have a web version as well. It is a beautiful phone. And our language engineering team did a content translation beta. So all really great things. I'm going to invite Damon to talk about visual editor. OK, so real quick announcement. Visual editor triage meetings. We talked about this at the all hands. This is coming. This is really, really important. This is your chance to participate in the engineering process of shipping visual editor. One big piece of this is that to the mailing list will be sent a list of queries that you can use to view all the bugs that are going into visual editor. And also, there is a set of instructions that enable you to nominate bugs or issues or tasks in fabricator to be included in the triage conversation. This is really important because if you have an issue you want to be discussed, that is how you actually get it on the list is through nominating these bugs. Those instructions will be sent out to the Wikimedia L list and also the tech list as well. That was it. Thank you. We have a soft do it as my co-presenter for this particular metrics. And we're going to go in a global south update. Thank you. I'm playing Eric today. Eric is playing me. So he'll be in charge of the heck. Exactly. So Cico is playing Anasuya today just to keep things confusing. Cico, please start us off while I try to get the presentation going. On working? Yes? Great. So today we're going to give a global south update. We've drawn some folks from across the organization to share with you what we've learned so far about the global south and where we're going next. Let's see if we can get some slides. Well, you can speak from them if we can't get it to work. The internet is tired today. Can we do it in non-present mode? All right. Do I? Can I have a clicker? Assuming we get it going. While you sort this out, can you sort of? Yeah. Yeah. Let's go to slide three and I'll start talking while we sort out the rest. Does that sound good? OK. So what is the global south, first of all? We define the global south as these countries highlighted on this chart. The global south is really most of the world. Interestingly, it's also most of the world's internet users online today. What it's not, oh look, we do have slides. Thank you. Is my clicker work? OK. Sorry about that. What the global south is not is most of our contributors. There we go. It still represents a very small percentage of our contributors and there's still missing a fair amount of content as well. What else do we know about the global south? It's not a monolith. Every country, every community has its own challenges and its own opportunities. There's a variety of solutions that are needed that look different depending where you are in the world. So why do we care? We're lucky enough to have Jimmy Wales here in the office today, the founder of Wikipedia. And we thought we'd ask Jimmy. Jimmy, why should we care? Well, I mean, I travel a lot all over the world and I meet with Wikipedians, but also Wikipedia readers. And one of the things, the story I want to tell is actually an interesting one. Several years ago, I was in, I think, Chennai in India. I'm not sure exactly where I was, and had a meetup of Wikipedians. And we had about 20 people around the table and we passed around a piece of paper for everyone to write down their mother tongue, their home language. And as the paper went around the table at being India, by the end, there were eight different languages represented around the table. And so we were discussing what are the challenges. And most of these contributors were working in their mother tongue language, but they also work in English and so forth. And at the end of the dinner, one of the young men who was there came up to me and he said, I don't really understand why you care about all these local languages because everyone in India who has a computer can read English. So what's the point? Well, he was wrong then, but he is really, really wrong now. When we look at the growth of the internet across the developing world, we see stunning increases. I was just recently looking at Nigeria as one example where the internet usage has gone in the past four years has more than doubled. And it's nearly 50% of the population now using the internet and Skywarding. So that boom that we all experienced in the late 90s and early 2000s when everybody got online is now happening in the developing world. And that means that it's no longer in any way safe to assume that people can read English or French or whatever colonial language might be relevant there. Instead, there are hundreds of millions of people, billions of people coming online in the next three to five years who don't speak English at all. And those are the people that we should care about. These are also people whose need for information, need for knowledge is actually in many ways different from ours. If you look at people across Africa, if your only language that you have is, for example, Swahili, the problem that you face related to information online is very different from the problem that English speakers face. For English speakers, one of the fantastic benefits of Wikipedia is not to make information available to us, but to actually make a summary available. So you type in Queen Victoria, and instead of looking at millions of web pages on all kinds of different things, you go to the Wikipedia entry, and there it is. There's the basic information that you wanted to know in the first place. If you type Queen Victoria in Swahili, you might not get very much at all. Nothing that you're able to read. And so the purpose of Wikipedia in that context is, in many cases, not just to help navigate this enormous body of information, but actually to make that information available to you for the first time. This, of course, is going to have an enormous social impact on education, on politics, on everything. And we need to care about it. It's what our mission is, of imagine a world. This is the world we're building. So for me, this is one of the most important things. And I think everybody knows Wikipedia has had a huge impact on the world, but our real humanitarian impact starts here in the global south. That's where we really are going to make incredible impact on the future of the planet. Thank you. Thanks, Jimmy. And I think that point that you're making around the different contexts is really important. And it's something that we're thinking a lot about today, and will be spoken to a fair amount throughout this presentation is, again, it's not a monolith. Uh-oh. Can I? This is our vision, the sum of all human knowledge. And without really broad and diverse participation, we can achieve this. Much is still missing, and much remains to be challenged. Our mission is to empower and engage in participating in the production of knowledge, and the curation of knowledge is part of that empowerment. The global south is still, many communities are very underrepresented in that process. And even Asaf was just underscoring to me that even on English Wikipedia, a lot of information about the global south is still missing, and having global south contributors participate in that process is going to strengthen those articles, not just in English, and eventually across all of languages. So that's another key for us to keep in mind. And then finally, the strategic reason. The largest number of people that we have sort of the largest potential to impact is this, the 2.5 billion people who are already online, but who are not yet reading Wikipedia, who are not yet participating in our projects, who we haven't yet reached. Over half of these are in the global south, and so as we think about opportunities for the future, that's a really key area to keep in mind. This is sort of our target group for the next set of efforts. So what have we done around the global south to date? There have been a number of different activities and initiatives across the organization that have been aimed at supporting readers, supporting contributors, and supporting sort of those not yet reached in the global south. So Wikipedia Zero is working with partners in almost 50 countries. The education program, about half of their programs are in the global south. We did the catalyst programs in India and Brazil. Language engineering, of course, is doing lots in terms of fonts and display and so forth. And in grant making, we've been providing mentorship and support to individuals and groups in the global south to lead initiatives that help them get where they want to go with their projects. And then things like hackathons in India as well. What we see when we step back from all of these initiatives is that we've learned a lot separately. We haven't been particularly well integrated or strategically coordinated in achieving our goals together. And so we're taking this as an opportunity now to take a step back to figure out what it is that collectively we've learned from all of this work so far and then where we go next together in a more coordinated fashion. So to that end, that's sort of what the rest of this presentation is going to walk you through, particularly what do we know today about readers and what do we know today about contributors. And then Asaf is going to give a sense of some next steps coming out of that. So with that, I'll hand it back to Asaf. Thank you. Apologies for the technical difficulties. And we now move to what do we know about readers and their environment in the global south? The speaker is Smriti, who will be participating remotely. She's on the hangout. Smriti, are you unmuted? Yes. Hi, Asaf. Can you hear me? Yes, thank you. OK, excellent. Hi, everyone. So we just highlighted why global south matters. Now let's take a look at Wikipedia readers in the global south and their environment. Our current readers in the global south, they are an affluent minority. They are urban, highly educated, and use English, or one of the other major language Wikipedia's. But they're also less than 20% of the global south population. We are not reaching the majority. We are not reaching every single human being. So what is the majority? The prospective Wikipedia readers look like. And even though the answer varies from country to country, here are four important attributes. So number one, majority of the global south population access the internet primarily through a mobile device. In fact, for many, their mobile phone is the only connection they have to the online world. Number two, and this connects to the point Jimmy just made, which is as more and more people get digitally connected, they need relevant local language content. And this includes content not just in their language, but content that also applies to them. Number three, they lack Wikipedia awareness. Many of them have never heard of Wikipedia, and they might not even know what an encyclopedia is, which is a critical thing for us to address. And number four, they face multiple barriers in accessing the online world. Now, the barrier could be affordability. It could be low end devices. It could be slow networks and many others. But this is sort of the world they exist in. And given that what we know about the global south potential readers, how can we expand Wikipedia's reach to the majority of the population? So we're at the next slide, and at least in my view, the map is not rendering. But it's a map of the world with the global south highlighted. So in WMF and Wikimedia Foundations so far, to expand our reach, we have been focused on access. And we have primarily addressed the affordability barrier by implementing Wikipedia Zero. But as Wikipedia Zero grows, we have learned that just making the mobile data free is not enough. It doesn't mean people will automatically start using Wikipedia. As we saw in the global south reader profile, affordability is not the only barrier. Therefore, we need to look at three things together. Access, awareness, and local content. Access means we address affordability, but we also address other issues, such as our user experience on slow networks or on low end phones. We also need to understand global south country by country and find ways to build Wikipedia awareness along with locally relevant local language content. As was mentioned earlier, global south is not a monolith. So the combination of these three factors, of course, varies by region. So for example, we have seen different needs and realities for Africa and Asia. In Africa, access and awareness are huge barriers. Whereas in Asia, it's more about awareness and local language content. So for the purposes of this presentation, let me share some learnings specifically from Asia. Next slide, please. Great. So these pie charts, they show Wikipedia page views by language in five countries in South Asia. The blue color is English, which means that Wikipedia page views are primarily in English. But these are not primarily English speaking countries. If you look at India, for example, 80% of the people do not speak any English. They speak one of the several Indic languages, many of which have their own Wikipedia's. But people aren't reading Wikipedia in their own language yet. So what's going on behind these page view numbers? And to elaborate, let's double click and look specifically at India. All right. Again, the image is not rendering, but that's OK. So in India, among the regional language speakers, there's low awareness of Wikipedia and also low awareness that a local language version exists. There is no organic channel where a regional language speaker would see the local language Wikipedia. Also, due to the association of many white-collar jobs with English, there is this sort of false assumption that people have to learn English first and then they can access knowledge. They don't know about the local language knowledge resources. And not just that. Even among the affluent minority, Wikipedia has great friend recognition, but many people are not aware that it's an editable encyclopedia. Many think of it as a useful rank or a resource to read. Now, in terms of mobile data use, so we talk about huge growth numbers in mobile India and how out of a billion plus people, 800 million people are connected via mobile phones, et cetera. But that does not mean that unlimited mobile data is the norm. It's actually the opposite. People buy small data packs that they use for limited use cases. For example, small data pack to use WhatsApp or Facebook for a few hours. So that limited use obviously means that there is less chance and less ability to discover and use Wikipedia. Then fonts is another issue. Older phones do not support Endic fonts. What people normally do is they learn enough English just to essentially use their phone. And they might text each other in their local language by using Roman script to get around the Endic font issue. But they cannot read local language Wikipedia because their phones do not support the fonts. Now, given this oral scenario, would a combination of the three things we talked about access, awareness, and local language work? And the answer is yes. We have seen one great success story with one operator in Bangladesh. So can you go back to the slide? Thank you. So you guys remember these blue pie charts? If you see specifically in Bangladesh, it has less blue than other countries. There's a good chunk of the pie represented by orange that shows page views in the local language Bengali. So what happened in Bangladesh? And now if we go forward to the Bangladesh slide, this pie chart shows page views of one of our Wikipedia zero partners in Bangladesh called Grameenphone. 95% of their page views are in Bengali Wikipedia. And a lot of that contributes to significant usage of Bengali in the page views from Bangladesh. So what did they do differently? They essentially combined access, which was Wikipedia zero, with awareness and local language content. Grameenphone, they promoted, so on the right hand side, you can see the screenshot of Wikipedia zero. And on top is the orange banner saying free Wikipedia zero provided by Grameenphone. And it's in Bengali. So what they did was, they promoted Bengali Wikipedia on Opera Mini, which is the preferred browser for many people in the global south and in Bangladesh. So when these people came online, they saw the Bengali Wikipedia. Also, even though many phones in Bangladesh don't support the Bangla font, Opera Mini does. Thus that removed a big impediment. It allowed people to both read and write in Bengali. Therefore just kind of placing Bengali Wikipedia and the flow of how users were experiencing mobile data, the usage of Bengali Wikipedia went up. Now, this is very exciting for us because it shows that people respond when we raise awareness of Wikipedia in local languages. Next slide, please. So, sum up. Essentially, each country is different, but we can look at the three factors that can help expand Wikipedia in the global south. Access is the only part of the equation. We do need to work with our partners and local communities to raise awareness about Wikipedia as a knowledge resource in their own language, where they can contribute as well. To do this, going forward, we would coordinate among various Wikimedia Foundation teams. Wikipedia Zero, Mobile Product and Engineering, Communications, Grant Making, and possibly more. We would need to come together to understand the context of Wikipedia usage and how we can unlock growth in the global south. So, that's it. Thank you, Asaf, back to you. Thank you, Smriti. Moving right on, we will share some stories and things we've learned about contributors and our communities in the global south. So, continuing the theme of the complex language landscape in much of the global south, I will add that different people operate in different linguistic circumstances. Many people in the global south use more than one language every day. They have more than one language for different uses in different contexts. Maybe they are educated in English, but at home they speak Swahili or Urdu or Tagalog. Even if they're native speakers of Tagalog, for example, for any definition of native speaker you would care to choose, they may not necessarily be able to contribute the kind of nonfiction expository information that we expect in an encyclopedic article in that language. That is a very key factor, remember, when we think about why aren't people contributing more in Tagalog? And 98 or something like that, percent of the edits from the Philippines, for example, are to the English Wikipedia, not the Tagalog. So, what is the language of education matters? What is the language of the major newspapers? The secondary sources matters a whole lot to the viability of writing the encyclopedia, quite apart from the issues already mentioned about readership and does your phone support it and do you have the font? Does it even occur to you to look for knowledge in that language? So these contexts are complex in each country. I could give other examples of the situation. I'll just give one example. In Ukraine, most of the country speaks Ukrainian as of these days, speaks Ukrainian as a first language and has a passive understanding of Russian and yet many people, and in the Eastern part people speak Russian more, but yet many people consume knowledge on the Russian Wikipedia and not on Ukrainian Wikipedia and the reason is depressingly simple. That's what comes up when they look for stuff on their computer, on their phone. That's what Google serves up, the Russian Wikipedia first. Very often it's because their operating system, their windows or whatever is configured for Russian because for historical reasons, Microsoft didn't produce a Ukrainian version or they did not have one installed. So it's a depressingly technical reason that's keeping people from consuming knowledge that is available on the fairly mature Ukrainian Wikipedia keeping them from consuming it. So I just wanted to give that other example of the landscape. So something that we are doing about this, our language engineering team, Amir is on the hangout and will be able to take questions later, has around 2012 deployed the ULS, the Universal Language Selector, which for the first time enabled contribution and readership in any language independent of how your computer or browser happened to be configured. That is huge, especially in the global South. And as you can see in this graph on the left, the Punjabi Wikipedia, Punjabi to those who don't know is a major, major world language with about 100 million speakers. The Punjabi Wikipedia was next to non-existent, was very tiny, very few people editing it until ULS was enabled, finally allowing people to contribute in Punjabi. So you can see that bump in the graph is exactly when ULS was deployed. It was not problem free, those of you who remember it was a difficult deployment, but you can see the effects very clearly on the Punjabi Wikipedia. And now they have about, this graph ends at about 15 active editors, but in fact just last month they had a very successful contribution drive and are now up to 40 active editors and 16,000 articles on the Punjabi Wikipedia, which is very much on its way up. The other major contribution, again just now, we heard it up top in the announcement section, is content translation, a huge, huge thing for all Wikipedia's that do a significant amount of translation. And in the Global South Quantitative Survey that we have just recently run with 47,000 respondents from the Global South, an average of 18% of contributors said they frequently translate articles from other languages into their language Wikipedia. And in some geographies in Egypt, the highest rate was 55% said they do translate into Arabic from other languages. So a software affordance facilitating massive, convenient software translation, including translation memory and other advanced features, could have a significant impact on generating that local language content converting from high quality, mature articles on other Wikipedia's. This is of course early days we have not felt yet this impact, but I really look forward to seeing it happen. It's going to be huge. So let's move on and zoom in a little to the experience of contribution. So if you want to contribute to Wikipedia, like my friend here, user Vishwa Prabha from India, from the Malayalam Wikipedia. He's a veteran editor. He has more than 36,000 edits on Malayalam Wikipedia. But he told me I could have made at least double that if only when I click edit, not visual editor, just plain old Wikitext took less than 30 seconds to get to the editable box so that I can start typing. That is his experience on his very low bandwidth connection down in Kerala and Southern India. And while he waits, he switches over to the tab with Facebook open just until it loads, but then two hours later, oh wait, I fell into the Facebook rabbit hole and Facebook ate my edits. I don't even remember what I wanted to do and stuff. So this is a very real lived experience of this already committed, already motivated veteran editor that is keeping him from contributing much, much more. Low bandwidth friendly user experience, some kind of solution to offer people who are through no fault of their own in a low bandwidth situation that we all can do very little about is a need that we may be able to solve technologically hint hint. Another issue is what happens when you do contribute particularly on the major mature Wikipedia's like English, French, et cetera. The existing policies of those Wikipedias, the existing policies of notability and reliable sources may create without intention a bias against global south contributions and against content about the global south whether or not it comes from a global south editor. So first let me spell out, the policy is actually fine. If you take the time to read English Wikipedia policy on reliable sources on notability, they're actually fine. The policy is not the problem. The problem is that good faith policy compliant edits get rejected because the patrollers, the enforcers of those policies interpret them a little too narrowly or a little too superficially. And so they do not recognize, for example, the media outlet that is cited in a citation. It's not the New York Times, I never heard of it. Not reliable, bang, gone, reverted. Additionally and compounding that, and a bigger challenge, is that non-Western traditions of knowledge are much harder to cite according to our generally Western citation standards, making it again difficult to bring some very high quality knowledge that does exist but is differently organized, was never published in a peer-reviewed journal, et cetera, making it very hard to bring that knowledge to Wikipedia. And on the far end of the difficulty scale, we have oral knowledge. There's plenty of knowledge in the world that is oral and was never written down anywhere and we currently have no way to cite it. There's just no means of relying on that knowledge and bringing it into Wikipedia. That is certainly a difficult problem to solve. I wanted to give you a quick example. The story of A. A is an educated Indian, in fact, a Rhodes scholar, no less, who has read the rules and has offered this, this edit, this is the first revision of this article on the right about B.C. Adelia Fahemi, a women's activist philanthropist from Nigeria, who is also currently the first lady of Akiti State in Nigeria. Certainly a notable person. Again, this is the first revision, okay? We're complete with what, 11 references and internal links and everything you could wish for in a newbie contribution. This is what was submitted and literally five minutes later, a patroller using page curation, a technology that we created, has slapped this speedy deletion notice on that page. This friendly old message that says basically, the upshot of all this, if you are not a very quick reader or not familiar with this is, this isn't important. You have not made the case for why this is a notable article. Now, I want to stress, this decision by that patroller is just not correct. According to Wikipedia standards, it was a bad patrolling decision. That happens, of course. But think about that contributor who has done everything by the rules, far more than we would expect most newbies to be able to do with citations and everything and just got this, this bad patrolling decision slapped in their face and what do you do about it? Well, what A did about it was turn to her veteran Wikipedia buddies because A happens to be Anasuya, my boss. And she turned to me and said, hey, what the hell? What happened here? And I took a look and I was like, wait, what? No, that's completely wrong. That's crazy. And she also happened to share a room at the time with Florence, our former chair of the board, who is also a veteran and feisty Wikipedian Florence looked at this and without a second thought, removed this template. So she undid, she reverted this call for speedy deletion as against policy. So that was a happy end. This article still lives and has been improved and that's a happy end. But what if A was not Anasuya? What if she did not have veteran, confident Wikipedians at her side to sort of undo the damage of this bad patrolling decision? And again, bad patrolling decisions do happen. That user, by the way, who made that bad patrolling decision is still using page curation today. I hope their decisions have improved. My point was, even when you do everything by the rules, if you're documenting a Nigerian woman using sources that do not begin with nytimes.com, you may run into bad patrolling decisions. People in the Global South, if you talk to people who have tried contributing on English Wikipedia, will tell you these stories again and again. So I deliberately chose a story that sort of had the best starting conditions. Obviously many contributions are lower quality. All right. So moving from the individual to the community level, communities are geographically dispersed in and out of country in the Global South. Many Global South countries have a large diaspora, frequently fairly active in that language, that region's language Wikipedia. And we can see that editorship does not necessarily follow viewership. If you look in the left graph, you can see that the Swahili Wikipedia is surprisingly significantly edited from Europe. Do you see that gray piece in the center? This was originally all numbers, but someone turned it into a graph, so now it's less legible, I think. But that's just me. Yes. Yeah, anyway, it is significantly, it is significantly, I'm heckling from the stage. So the Swahili Wikipedia is significantly edited from Europe. It's not significantly read in Europe. It is significantly read in the USA, actually. Lots of people are consuming Swahili Wikipedia in the USA. Of course, views, readership tracks population size. Edits track a much smaller group, naturally. But we can see that very few people edit from Kenya. Many more people edit from Tanzania, where Swahili, remember complex language landscapes? Swahili has a much stronger status in Tanzania. It is the language of education in Tanzania. And so we have more edits from Tanzania. By the way, depressing fact, of the four most active editors on Swahili Wikipedia, three are Europeans. We can also look at the Bengali Wikipedia. As we can see, it's vastly edited from Bangladesh. Again, the language landscape there makes it so. But it is consumed in many other parts of the world. Within countries, we have communities that are in larger countries, like Turkey and Brazil. You have the numbers before you. We have communities that are very geographically disparate and unconnected. And sometimes they don't meet up. They don't have a national Wiki conference. They don't have regular meetups. They don't have a lot of off Wiki get-togethers, which are not necessarily a precondition for doing programmatic activity, but very often before you've met people, had a drink with them, had lunch, whatever, it's hard to sort of plan the museum partnership or the other off Wiki kind of activity. And so that's what we see in Turkey and Brazil, for example, large, active editing communities, very little off Wiki activity, very little programmatic activity, partly because of the self-organization difficulty. And that is something we can help with from the ground up, not through catalysts. Moving on, here's a confusing graph. Okay, so bear with me. Every dot, every dot is a user. The lines that you absolutely cannot see from where you are, connecting those dots are interactions between users. The red dots are people, users who are only connected to one other user, who have only ever interacted with one other user. The interactions I'm talking about are only on the user talk pages space. So not article talk pages, not policy discussions. User to user talk page messages is what is mapped here. And as you can see, there's, you can see an evolving form of on Wiki communication that is easy to chart in this way without speaking a word of any of those languages. You can see that that Swahili Wikipedia, you can actually see those editors that I mentioned, each of them with a cloud of red dots. Those red dots are newbies. They're newbies, of course. They are people who were either welcomed on their user talk page or warned on their user talk page. But there was that one interaction with one active user. That's what the Swahili talk page interaction looks like. But you can see the Tamo Wikipedia, which is larger, more mature, has a lot more interaction between members, between that core of other colors, yellow, orange, green, et cetera. The bluer, the more connected a person is. This is, by the way, is a tool that we have up online. We can, someone can probably share a link on IRC. Yeah, you can actually play with this diagram live and kind of find out who that dot is and how many people are they connected to, et cetera. Fun for the whole family. But we can see that as you go to a larger and larger Wikipedia, that core tends to become denser and to have those interactions. Very broadly, it probably correlates to community health, in a way. But there's a lot more research we could do with this kind of mapping. We could, for example, start mapping what is happening on article talk, where people argue about facts and content and stuff. We could map what's going on on the project space, on the Wikipedia namespace, where people argue about policies and elect admins and do that kind of thing, which we haven't yet, but we certainly could. But even when you compare similar communities, similar size communities in terms of active editors, in terms of the editing community size, like Tamil, Wikipedia, and Hindi, Wikipedia, you see very different patterns. As you can see, and I've provided the actual numbers of outgoing communication and ingoing communication for those obvious dots. So if you look at the Hindi community graph, you see that those three clouds are centered around three users, one of whom has sent out 1,000 messages to other users, warnings, welcomes, whatever, and has not had a single person write on their user talk page. Another one sent 1,200 warnings or welcomes and had 22 people write in. Another sent 4,000 messages, had seven people write on their own talk page. So it's a very, very distinct communication pattern on Hindi, Wikipedia. We need to interpret it carefully, but it's clearly different than the pattern on Tamil, Wikipedia, where some of those central dots, this is a very imprecise legend that I made there, but you can see that some person had 1,000 messages going out. So again, he's one of those people who welcome people or one of those people who monitor recent changes and sends warnings and still had 100 people connect to them back on their talk page. Not necessarily those same newbies, of course, but it's a different communication level in the community. We also see that these similar sized communities have different challenges because of their environment, because of where they are. The Tamil Wiki is closely in close relation with the government of Tamil Nadu state in India and has already affected some very interesting partnerships, including content partnerships with the Tamil Nadu state government, which has an interest in cultivating Tamil language. Hindi Wikipedia is far, far more diverse geographically across all of Northern India and many second level speakers, but has a higher potential for national level partnerships, particularly with the current government. Their challenges are sometimes different too. A prominent Hindi Wikipedia told me, for example, one big challenge they're facing is the huge turnover among Hindi editors. So people swing by, do a bunch of editing for a year or two and then they move on, they disappear. There's very few old timers who kind of stick around and can own the norms, the values, the traditions of the editing community. Apparently not so much a problem on Tamil Wikipedia. So this is really just the tip of the iceberg on what we might learn about these communities. We are complimenting this very bird's eye, mechanized look at communities with qualitative research as well. We're conducting a series of interviews with GlobalSouth communities to learn, for example, what does it mean? What do you talk about on user talk space versus article talk space on different Wikipedia's to compliment what we can learn from quantitative approaches. I'll just mention briefly because I talked about institutional partnerships, those in general are much harder to effect in the GlobalSouth. It's much harder to get a GlobalSouth museum director, university director, et cetera, to partner with you. People are generally with far lower budgets for this kind of not quite curricular activity. They are sort of less advanced along the path of embracing free and open culture, et cetera, as compared to the Global North. All right, so these have been some very, very brief tastes of things we've learned and seen in the GlobalSouth. We want to offer some suggested takeaways and allow some time for discussion. So suggested takeaways. Access is not enough, not enough, not enough access, but it is not enough. We need to have access plus generate awareness and there's a variety of interesting ways we have not yet really explored like mass awareness campaigns. Get in touch with radio stations. There's all kinds of things we could do to create awareness to the existence of Wikipedia, to the existence of knowledge in Wikipedia and to the editability of Wikipedia. And the availability of local language content. This also came up again and again in our quantitative GlobalSouth survey where people were asked, among other things, what would make Wikipedia better for you? What are you missing in your Wikipedia experience? And again and again, people said specific content, local content, people in Bangladesh said we want specific Bangla content, we want content about Bangladesh, et cetera. As Jimmy said, native language contribution is crucial for us to achieve our mission. It's right there in the vision statement. Every single human implies it cannot be just English, it cannot be just French. The vast majority of the world, remember, an absolute majority of the world does not speak English at all. But it's complex. Like I said. And there is no one context to rule them all. Every time we say the GlobalSouth, people in the GlobalSouth would tell you kittens die. I will tell you, we are just using shorthand, but as we have said repeatedly in this talk, context differs across the GlobalSouth. It's just shorthand for a bunch of different contexts in a bunch of different regions. There is no GlobalSouth approach, GlobalSouth solution. Another takeaway is that we need to scale and customize carefully. I'm thinking in this regard in particular about the catalyst programs that were started as a result of our previous old strategic plan from 2010. They were very ambitious. They were at scale from the get-go, and they did not achieve what they set out to do. They did not work as planned. The India catalyst was successfully transitioned into a much more successful grant partnership that is building things from the ground up. The Brazil catalyst was terminated without a continuation. But both of them had serious design flaws in terms of the ambitiousness and scalability that we embarked on them. A lesson I hope we have definitely, definitely learned. And finally, remember Vishwa and the low bandwidth experience. That's just one example of why we need diversity of languages and contexts, including what is your bandwidth in our product thinking, which I don't think we have had sufficiently so far. We need far more presence of those very different experiences of Wikipedia's out there in our product design, not just our user studies, not just our UX design, our basic prioritization of features. Do we need a low bandwidth experience? Is there some other way we can solve this? I don't know, but I want our product people to have that voice in the room with them. Finally, if you felt that this was all a little rushed and a little superficial, I agree. Here's where you can learn much more about everything here. So Wikipedia Zero will be giving a tech talk sometime this month, TBD. I'm sure the Zero team will follow up on WMF All and update us about this. The Global South Survey I have mentioned a number of times with lots and lots of interesting and occasionally surprising graphs and results will be discussed in detail at the research showcase toward the middle of the month, February 18th. And we will be screening a documentary called People Are Knowledge, made in 2011 by Achau Prabala, who's a member of our advisory board and a longtime friend to the Wikimedia movement. He's made a film about oral citations, about the prospect and the challenge of citing or finding a way to cite oral knowledge on Wikipedia. Some of the old-timers may have seen this before, but I think the majority have not. And we get to screen it because it's free content and it's on comments. So we will have a screening on March 4th and Achau has graciously agreed to actually participate and take questions, even though it would be like 4 a.m. for him in India. Finally, sometime in March, we will have a panel discussion on our activities to date to answer all of your questions like, what have we done so far? Did you try this? Did you try that? And can I have links? So we'll have this panel with me, Amir from the language engineering team, Caroline from Zero, Yan from the LCA team, and Kim from community engineering, where we will try to both provide a lot of context on some of those past activities, but also brainstorm with you and answer any questions you have. That precise date is also TBD. Queue discussion. Thank you. Questions, wait for the mic. And James is playing James on IRC. So nice work, you guys, that was a great presentation. I think from the product perspective, we are interested in becoming more diverse. I think we've been primarily talking with Caroline and the communications team about sort of sharpening our pencils around where we wanna focus and sort of what are the key features that that market needs. I think Caroline's gonna talk a lot more about that, but Dan has been thinking about stripped down apps that would be better for low bandwidth. So I hear you, and we're definitely interested in looking forward to partnering with you guys more. Basically starting now, so. Great, thank you. That's the best kind of question, you just say thank you. Yeah. About those pie charts. On IRC, lots of people were very impressed by the widgets that they could click on and look at different communities and they thought that was awesome. And there was consensus in the room about putting new data centers into the global south, so success on that front. The question was, the first question came out was from Erin who asked, do we have an idea of how large the kind of, the size of the prize is? So is it, what's the size of the current internet population versus how many we're currently surfing? And how does that cut down in the global south? Are we at 80% market penetration in the west and 10% in the global south? Is it 50-50? Do we know? We have pretty rough data there. I think those global numbers about internet users in our penetration all kind of give or take half a billion, you know? So very hard to give precise answers on this, but we are far from saturated in the global south, definitely. I would like to dig a little more into those numbers, but I don't feel I have anything intelligent to say right now. Yeah, so great presentation. I like to hear you called out the question of governance and moderation and patrolling, because that's an aspect that we don't talk enough, I think, and it's great to see there are initiatives both on the product side with content translation and on the outreach side on increased diversity, but it strikes me that the problem of increased diversity at the patrolling governance moderation layer is actually way more critical because we can bring in all these people just to be killed by machine guns as soon as they start becoming more engaged. So there's that problem of engaging more diverse population of moderators and people involved in governance. There's also a problem like giving negative feedback to people who flag something for deletion right after something is created, that flag is removed and there's no negative feedback for these people. So they can keep doing this over and over and we need to think seriously about how to address these two problems. Maybe that's a product question. It's not just like an issue of socializing, changing policies. It's something we can do as scale with product. So I want to hear your thoughts if anything's happening in these two directions. I can't speak to what is or isn't happening. Of course, you're perfectly correct and I tried to stress that the policy is actually not the problem here. We have done some two years ago about two years ago when Mariana and Steven were doing that research into templates, you'll remember. We've done some small things on that. The thank feature, all that stuff to nudge people into being more civil and to nudge people into providing positive feedback as well as negative feedback. So that's all good. I don't know that we have a lot of thrust around that question anymore but I think people are better qualified than me to talk about it. I agree with you, we need to do more and this patrolling behavior and again I'm not picking on that one user, lots of patrollers make mistakes. I have made mistakes as a patroller. Suggest to me that some kind of training, some kind of maybe positive feedback, not so much negative feedback on patrolling but positive feedback for good patrolling. I don't know, I think we do need to spend some time thinking about this, what can we do that is not about changing policy, that is about enabling and what we've done in the past like page curation, the tool that was used to shoot down this article is an asset. It's a tool that was very useful for some patrollers and not so useful for others because they claimed to create a huge backlog. But remember patrollers are doing this insane, thankless job of monitoring recent changes at 3 a.m. while we sleep and making sure Wikipedia is free of porn and vandalism. So it's complicated. We need complicated solutions, we need creative solutions. What I'm saying is we need to put some resources into experimentation on that front. One, I'll give a shout out to my friends from grant making, one way to experiment is through grants. So through individual engagement grants, experimental grants where we can say, hey, why don't you sketch out this extension, this model, this thing and see how it goes. Yeah, other questions? Damon. I was just gonna ask real quick, the Swahili example I think is really interesting because you have three central editors, right? And a whole bunch of single interactions with those editors. Just I know we're still getting into data on this, but is there a general trend in the type of feedback that we're seeing there and is it positive, negative? Because I really agree with Dario's assessment there where we need to give feedback mechanisms either positive or negative from some form to those editors. And imagine there are certain locales that this type of investigation is easier. And Swahili sort of stood out. Yeah, I think the problem we were discussing is actually much more acute in the very large Wikipedia's that are much less personal and much less, and you're drinking from the fire hose of recent changes and it's overwhelming, it makes people jaded as well because they've seen so much crap. On Swahili Wikipedia, I don't think there are three, I think there are more like five active editors are Swahili Wikipedia to a large extent. I mean, literally they are the ones who keep coming back week after week, day after day and make more than 100 edits a month. So they know each other. They engage maybe not on each other's talk page so much. Maybe they engage on the Village Pump because the Village Pump on Swahili Wikipedia is a pretty quiet place. And they just coordinate, coordinating between four people even though they don't live in the same country is fairly easy. So I wouldn't, I don't think feedback to those Swahili admins is actually anything we need to worry about. So those very tiny communities are doing okay in terms of their governance. Other questions, comments, provocations? Eric, I'm very disappointed in you. I know. No heckling. No heckling. Louis over here? You got the light for me too. And then Jimmy. Hello. Or not me. Yes, Emily. Yeah, real quick, so these patrollers, like I know some of them are kind of automated, some of them are people, are you evaluating their authority to decide whether that particular article was okay to squash? How are they qualified, interviewed for a particular subject matter, if at all? Short answer, short answer. No, we're not evaluating it. What qualifies them to do it is 80% their willingness to do it and 20% one single admin has given them the rollbacker bit maybe. You can do it even without the rollbacker bit. So it's maybe 99%. They're willingness to do it. They self-select and unless they are very egregiously wrong and someone goes through the motions of actually dragging them to the admin notice board or going through some other community process, they will also keep on doing it. Yeah. Well let me, I just noticed you went real quick at the end of that last response to Damon. You went from, this is a small community, they talk to each other, they have good governance which is not where I expected you to go there, right? Do we have, because I find that to be somewhat different from they are civil to each other, right? Like that there's a good atmosphere of conviviality or whatever you want to call it, right? And I was wondering if you just wanted to explain. No, they're a very good catch. Always be careful with the word governance around a lawyer. Good catch. I did not mean to assert that they have no governance issues but I did mean to assert that the governance of patrolling, like how patrollers are given feedback which was his question is not really an issue. I mean, if someone goes rogue on Swahili Wikipedia and starts reverting good edits, he will be stopped very easily. Yeah, Jimmy. I was intrigued by the concept that in the Ukraine people Google and they find just Wikipedia. Not the Ukraine. It's not the Ukraine, Jimmy. It's just Ukraine, sorry. They're really upset by that. Yeah, go. I'm sure, okay, fine. And now I'm scared. It's colonial, colonial. Anyway, okay, I'm sorry. I know the president of the Ukraine. Yeah. All right, so anyway. I was intrigued by Virginia Carl. And I assume, yeah, I assume that this is a problem in many places, that people aren't finding their local language. Has there been any thought about being more sophisticated, geo-targeting banners to people, you know, when, if I'm in Kiev and I Google and I get the Russian site, when I go to the Russian site, shouldn't it have a banner at the top saying, by the way, do you know this article also exists in Ukrainian and we really need your help to edit Ukrainian Wikipedia? I think it should. You would need to convince Russian Wikipedia admins of this fact to let you run that banner and that, in this particular example, may prove tricky, but yes, theoretically, yes, that's one obvious approach to trying to get people to be aware. Hey, you're here because Google brought you here, but hey, there's a local language version you may want to check out. And the same, I mean, the same could be true, I mean, in English Wikipedia. Yeah. If you're in Delhi and you come to English Wikipedia, we could tell people, hey, you might be interested in these languages, which are common in your area. Yeah, absolutely, and that's a feature, the geo-located central notice is a feature that many of our communities don't really know, which is huge. Actually, a lot of people don't know that content in Wikipedia is different, so one thing I love to do when I travel is to translate on native Wikipedia, and then see the difference between English version and their version because it gives such a great differences in points of view and things like that, so I think there's something to think about. But to answer the community-based question or comment on that, I think coming in was like the iron sword or whatever, isn't the right approach, right? Because the communities that exist there are self-governing. What we find really working well is good communication and getting people on the same page about what are we trying to achieve here because we're all, at the end of the day, want to build the best encyclopedia in the world, and sometimes we may not have a bigger perspective or bigger picture, and bringing people together and talking about what changes do we need to make given the trends that we're seeing has been showing to be one of the best methods for us to do so, and we've started to do that. Jeff's team has actually been doing that and so has the grant making, and I think that's the kind of path that we need to continue on. I hope that we need to wrap up for lunch. Are there any more questions? Can I just, as an Ukrainian, I feel like I have to jump in here. So I think we've been bashing a little bit on the access point a little bit, but what I wanted to point out about Ukrainian as a speaker of both Russian and Ukrainian is that there's another reason why a lot of people who are in Ukraine and speak Ukrainian go to Russian Wikipedia and that it's, for a long time, it was just a higher quality encyclopedia. There were more contributors, more people speak Russian overall globally than Ukrainian. And what I'm wondering, I guess, is to what extent our vision statement actually holds us back from reaching our mission in the sense that it's every human rather than every machine-translated piece of content that we're actually after. So we have this vast treasure trove of knowledge that we've collected over all of the languages that exist in the community projects. But we haven't really touched so much is taking that information and just translating it by our machines as good as it can get, slapping it up there to provide something, some kind of stub for all of the languages that are underrepresented. I'm wondering to what extent we should be pushing people to create original content to their own language and to what extent we should be making use of translation libraries that exist and machines and algorithms to do this work for us. Eric, do you want to answer that? Yeah, fair enough. All right, so that's a very good point. And I was actually just looking around on Swahili Wikipedia to understand and answer today's question, like how do you use this talk to each other? And if you look at the sort of most common message that's being posted, it's a typically long welcome blob in Swahili with two sentences in English and the last sentences. And by the way, never involved post-computer translated texts like Google translated, et cetera, to this site. Like it's literally such a strong averse reaction against machine translated content that they put it in their welcome template for every new user asset. Don't do this, or you'll get banned, kind of. Yeah, and this has a lot to do with many global south languages where there's a large group of semi-fluent or semi-literate people who can contribute but in kind of sub-standard language according to the, I don't know, higher level speakers and it would be rejected because they set the tone on that wiki. And so the sort of interesting middle ground here are these bot-created articles that we see in many languages that are grammatically correct and carefully prepared for specific topics like cities or species, et cetera. And we see in Resonator, Magnus's wiki data tool automatically generated article abstracts in specific, like right now English but it could support any language that are grammatically correct. So we could potentially populate wikipedia with auto-generated grammatically correct content or at least not necessarily populate but show such content. So I think that's a very intriguing possibility. Yeah, and Hindi wikipedia, for example, specifically requested Magnus to add Hindi to the terminator. So there are also positive examples of welcoming that kind of intervention. I was wondering though, like on the Chinese wikipedia for example, like we have the concept of variants where the content is automatically transliterated but not translated. And it seems like for certain languages like maybe for example, Ukrainian and Russian, there's enough similarity that maybe you could build something that is a fairly good machine translation service that might even pass muster or people who are very skeptical of machine translation but just in very particular language pair cases. Amir, are you with us? Do you want to take that? I really need to cut it off. Yeah, I mean basically content translation may get us there. All right. Hello, can anybody hear me? Thank you. Okay.