 knowledge through languages while making Wikimedia more multilingual and multimodal matters. I am very happy to be holding this conversation. My name is Claudia. I am a brown feminist techie and activist from Bolivia. And I am the coordinator of the language justice program at Who's Knowledge, which is a global multilingual campaign working to center the knowledges of the global majority world online. To start us off, I want to introduce our wonderful panelists. First we have Bhaskar Dharmawan. They are a researcher and extraordinary language justice activists from Indonesia whose research interests include AI, ethics, digital activism, queer internet and information access in the global south. I also want to welcome Ishan Chakraborty. Ishan is assistant professor in the department of English at Jadapur University in India, where he is pursuing a PhD with a dissertation just around the corner. And he is an inspiring language justice advocate and disability rights activist. Last but not least, we are happy to welcome Dumisane Dubane joining us today. Dumisane is a polyglot person from South Africa. He is an amazing language justice strategist and activist Wikipedia and convener of the first week in Lava for African Wikimedia. Welcome to all three of you and thank you for joining us today. To set the table for the panel, I want to bring some anchors to our conversation and offer some context about why we came together in this panel. At whose knowledge we think deeply about historical and ongoing systems of power and privilege that define our lives today. Within this power analysis, language is a key entry point to understand the wider picture of how these structures of power operate, what impact they have on the internet and beyond and what are some of the ways to change that. Multilinguality and multimodality are at the core of knowledge, equity and language justice on the internet and on Wikimedia, right? In fact, Wikimedia is a global movement that is rooted in languages and because Wikipedia is among the most visited websites and is arguably the largest open and free knowledge repository, what happens on Wikipedia has great impact on the larger internet. Another anchor for this conversation and I will finish with this is intersectionality at whose knowledge we recognize that systems of power and privilege are so hard to dismantle because they encompass multiple and interlocking forms of oppression. So to build a fair, just and equitable knowledge movement, we need to factor these intersections into our ways of thinking and doing within the Wikimedia movement. So today we're definitely going to be bringing an intersectional lens in today's conversation. Okay, that's enough of me. I'm really looking forward to start this sharing and I will open up the space with you, Pasca. You have been part of our very first research action process on languages, which was the state of the internet's language report and we have learned so much from you in that process. I remember you shared with us how difficult it was for you at a younger age to find online content in your language that was affirming and positive to your identities. Fast forwarding time today as a Wikimedia user, what can you say about your experience looking for content that is valuable to you in your language of choice on Wikimedia? All right. Thank you, Claudia. Hello everyone. Good morning, good evening, good evening, good evening to all of you. My name is Josie Pasca-Dormon from Indonesia and it is an honor for me to share my experience and perspective in this session with other wonderful panelists in here. To answer your question, as a queer Indonesian, I'm quite privileged to be able to communicate to understand English content. However, many other queer Indonesians, especially those who are marginalized either economically, socially, geographically, or any combinations of them, are not able to understand English. Therefore, having queer content in our native language becomes very important for the queer community in Indonesia. Unfortunately, there is limited number of queer articles in Indonesian Wikipedia. There are not many articles about activists, organizations, public figures, and social movements of queer Indonesians on Wikipedia. And many of the existing articles are also a stub or article vindicant in Indonesian, which require further extension and elaboration as they are missing a lot of important details and information. Some articles were also subjected to vandalism and also online hate speech. For example, the article on Gede Utomo, who is a prominent LGBTQI plus activist in Indonesia, is currently vandalized with an added stating that he is, quote unquote, activist sart. This can be translated as preferred or defined activist in English. I believe you can still see this vandalism on his article in Baza Indonesia right now if you go to a page about Gede Utomo in Baza Indonesia because I just checked it a few hours ago and I found this vandalism and this makes me really sad as a queer person, to be honest. All of these issues regarding Indonesian queer content on Wikipedia is quite unfortunate because the government actually imposed heavy censorship on queer representation, positive queer representation on Indonesian mainstream media, and a big platform such as Wikipedia often becomes the go-to place for many Indonesians, especially queer Indonesians, to learn more about LGBTQI plus topics. When people are googling something related to queer topics, for example, Wikipedia often comes up on top of the search results and that just shows how feasible Wikipedia is as a gateway to information for many marginalized communities, especially for queer people in Indonesia. So for this reason, I believe that it is important for us to improve the quality and quantity of queer Indonesian articles in Wikipedia and we still have a lot of rooms for us to grow and develop this queer content on Wikipedia. Thank you so much for that, Pasca. That seems to me really fundamental, the point you are making about how the content that is on Wikipedia impacts search results and therefore the content that is easily accessible to people through search engines, right? That's a really key point. And I want to follow up with a similar question with you, Ishan, as a person who also thinks deeply about intersectionality, you have been an ally of whose knowledge for some time now and you are presently co-designing with us a research action process where we are exploring the intersection of accessibility and languages in South Asia. What can you tell us about your experience as a Wikipedia user from that crossroads where accessibility and languages meet? Greetings and lots of love. This is Ishan. I prefer to use the pronoun he, pronounce he, him, his day for myself. I'm a cis gay man and I'm a person with deaf blindness. So thank you so much, Claudia, for that question. I'm just going to sort of point wise, you know, addressing your question point wise would make things easier for me because also we have certain time limits. The first thing that I would want to focus on is definitely content, something which Pasca was also focusing on that Pasca was also mentioning. Same thing applies to Bangla, Wikipedia as well, but I'm talking about my intersecting identities, particularly when I'm trying to find out materials on Wikipedia in Bangla, positive, affirmative articles and materials in Bangla on Wikipedia vis-a-vis the intersections of disability and queerness. There is very little material available on Wikipedia and this actually speaks volumes about the socio-political economic condition of conditions in which the Bengali, Wikipedia users they are having to live, they're having to survive. Firstly, the socio-economic political conditions, they do not allow, they do not allow us, often they do not allow us to speak about these quote-unquote sensitive issues, these quote-unquote issues, issues which are in a way also tabooed in our society in certain ways. It becomes very risky at times, safety issue becomes something very important to actually come up with to contribute articles in Bangla, on Wikipedia in these areas. So creating content becomes very difficult, but also since materials and contents are not available on Wikipedia in Bangla and as Pasca was mentioning, same thing applies to disabled queer Bengali internet users, many of them cannot access internet in English or they cannot access internet in English in any meaningful useful manner and the English articles often remain beyond their reach and these positive affirmative articles, they do not reach them and therefore they do not get to know about their rights, they do not get to know about their, they do not get to access the language which they need to affirm themselves, to assert themselves, their rights, their identities in the society. So that is the first point regarding content, but what I would also want to add to it is form. It becomes extremely as a person with profound visual disability, I can say this with a lot of, there is a lot of complaint, I would say in my voice that I would want to express that Bengali in Bengali accessing Wikipedia is particularly difficult. For example, if I want to do a search in Bangla to find out materials on Google, it becomes very difficult, some of the websites are actually they do not read out these Bengali texts, but anyway using your touch screen mobile phones or desktops to type out Bangla is anyway a cumbersome process, is anyway very difficult. At least I find it very difficult, many persons who belong to the visually disabled community of Bengal, they find it very difficult as well. So accessing as in terms of usability as well, I would say there is a lot of barriers, there are a lot of barriers. And the third point that I would want to really focus on is the question of what you were saying intersections, yes intersecting and I would here want to reflect on my position a little bit. Today I can access internet in English, I have certain cultural capital, I have certain privileges therefore probably I can afford to disregard the lack of materials in Bangla on Wikipedia on the intersections of intersections of queer identities and disability, but there are many visually disabled persons that also we need to keep in mind that disability and poverty in a country like India are inextricably related. Disability and lack of resources in a country like India are reciprocally related as well and therefore we must understand that the cultural capital which is needed to access Wikipedia in English might not be available with many disabled persons, many queer persons or many disabled queer persons. So we also need to be careful of that power angle which comes in over there. So these are the three points which I would really want to raise. Yes the point that I want to pull out from what you were saying is how the marginalization of some languages have a greater impact other than simply accessing certain content but it really reinforces and amplifies other kinds of injustices as well. And I want to bring one other angle into this conversation that relates to new language technologies since we are also speaking about accessibility and I want to tie that with structured data and mention Wikipedia which is another very important Wikimedia project. Wikipedia is arguably one of the most important datasets used to train large language models today and it plays a key role in the development of new language technologies which is exciting and worrying at the same time because gaps, inaccuracies, biases that you were already talking about from Wikipedia are transferred to Wikidata and then imprinted in technologies relying on large language models. Dumisani, you are currently part of a research action process at whose knowledge through which with our amazing partners at the Distributed Artificial Intelligence Research Institute we are exploring different ways of building tech for African languages. More specifically we are exploring what does it mean to build these technologies centring community values, ensuring that language communities have power and safety over what and how they share. Okay that was a very long preamble to my question to you Dumisani which is as an African languages advocate and Wikipedian what do you think are some of the challenges and opportunities that new language technologies bring to the Wikimedia community? Thank you Cloudy. That was my way of introducing myself in my language. It's one of the smallest languages in South Africa and you touch on a very important part on the intersection between knowledge production and language activism. On top of that you add another layer which is artificial intelligence where at least in this sense large language models and how they play a role in the way we are producing knowledge in the Wikimedia ecosystem. I think for me it's said that 20 years later we are now talking about artificial intelligence and large language models when we haven't even responded to some of the early calls about oral citation which has been therefore almost 20 years from the work that Ashal did from India to Brazil to the work that Bobby did here in South Africa and in Namibia it's a connection of communities trying to say that knowledge production and curation does not only happen in a written format and you know here we are we still don't have any solution in terms of oral citation so that they can be included in the written platforms that we have that such as Wikipedia. But let's talk about these large language models that are coming in. If you look at the Quentin translation tool which is enabled on most of the Wikipedia's you will see that a lot of those translations are going to languages in the Global South. We are representing almost 50 percent more than 50 percent of the whole population in the world and yet in terms of technology that is serving us and our needs we lack behind we really seriously lack behind. My challenge or my problem with large language models deployment on Wikipedia is that it is being done and leaving the community behind. Which kind of language models are being deployed? Who are the people behind the technologies? What's the interaction between me editing and translating in my language to their model becoming better with each and every edit? Do I get told about about this? Do I have a say in how these language models are deployed and improved in my language wiki? You said it correctly that the wiki data in particular plays an important role in training large language models but wiki data didn't happen on its own. It is a collaborative effort of all contributors around the world in different languages where we've painstakingly worked to incorporate and to localize our own languages into the interface and now that work is being taken on and fed into a machine translation with all the mistakes that we have made as volunteers with all the mistakes of lack of curation and lack of better tools in small languages to make sure that the quality that we're producing on our language wiki is the quality that we need. So there's a lot of problems that we need to sort out as we as we go on to roll these different language models but centered to that through those problems is how do we bring our communities along and make sure that their pattern parcel of this development of these language models. Thank you for that Dumisani. Yes, absolutely right. I think it's coming to clear focus how important it is to put at the center community knowledges, communities values, communities ways of doing and being. Yeah and I want us to move into a slightly different direction now. We have pointed out some of the critical ways in which issues of language justice manifest in wikipedia and wikimedia projects and I want us to think about actions and imaginations to make wikimedia projects more meaningfully multilingual and multimodal and for the next set of questions I will go backwards and start with you Dumisani. I know you are a person who is constantly seeing the world and inhabiting it through the lens of power, languages, knowledge and justice. So from your positionality what is the one barrier or challenge you would like to see addressed in the movement to advance towards more multilingual and multimodal community? I think the one barrier is building on what we were talking about in terms of the deployment of large language models is how do we center communities in that deployment? How do we bring them along in the development of different machine translation technologies and how they get deployed into the projects? It's not enough to say we've put in a talk page notification that this is going to be deployed and here's the meta page for you to go and read a little bit more about this machine translation technology that we're deploying. That's not enough. Taking our communities along is making sure that they're part of each and every design aspect of that rolling out, including explaining to our communities where or who these large language models are being developed by. Google, do we want to improve the language models that they have? META, how do they tie in with the values of our communities and of our ecosystem? But if we just simply roll out without asking those questions, then we are doing a disservice to our communities who are continuing to work hard to improve. With every translation that I'm doing on my little Wikipedia, I'm improving a language model that's sitting somewhere out there that I don't even know about. So that's a problem that we need to tackle and we need to do a little bit more visualization of our communities to these models that we are deploying and take them along because they are the greatest asset we have in this movement. Those are such key questions to Misani that you are raising absolutely. I think that one of the issues as we need to tackle building up on what you were saying is that we don't need to understand profoundly how these technologies work. We just need to be aware that these are processes, these are models that are not including us, the majority of the world, and therefore they need to change. We need to change them. With all those thoughts and reflections, I will go next with you, Ishan. For you, what is the one action or issue area you hope to see move forward? Okay, so number one is definitely in terms of the content translation of articles which contain positive affirmative representation of disability and queerness, the intersections, the various kinds of intersections. And of course here we are thinking of two spectrums, disability and queerness. They are not homogeneous ideas, but two spectrums that we are talking about and they are intersecting with each other to really produce this iridescent identities, iridescent layers of identities. So really to actually translate those articles which are available in English and other languages into Bangla to start a conversation there so that other Bengali users who identify with these causes, who identify with these issues, they can also come up and contribute articles to Wikipedia on these issues. That is definitely one. But the second point is as far as the form is concerned, I would want to uphold, I would want to highlight the need for multi-modality. It's just not the written text which is out there, which I'll have to use my text-to-speech software to read, but also other forms of ways in which materials can be or articles can be made available to visually disabled participants, to disabled participants, deafblind participants, for instance. Audio material can be one, audio versions of the articles can be definitely one that would actually talk about, that would also bring into question the accents, the accents in which actually these articles are being read out. The human touch to the articles as it were, just not a robotic voice reading out the entire article, but a human voice reading it out to me, that that can be one, but the other can also be sign languages bringing in sign languages into the entire picture. We do not have a Bengali sign language as such, but definitely acknowledging Indian sign language and actually translating these articles which are relevant to persons with disabilities, deafblindness and queerness, making articles available in sign language, I think that can also be an important tool, important way to approach it. So these are the two things which I would want to see. Thank you so much for that Ishan. Yeah, I think one of the things we all should have in the back of our minds as we think about multilinguality and multimodality is exactly that what you said, right, that knowledge is so textured, is so rich and so plural and it gets reduced to really its simple form when we just rely on written text. And Pasca for you, what is the one thing you would change to improve content creation in Bahasa, Indonesia or Wikipedia that would be your joy? Right. First of all, I would like to say that I'm grateful that Wikipedia has provided a lot of knowledge and visibility for marginalized communities all across the globe. But I think it would be very great if we can bring more native queer content to Wikipedia, especially those coming from global South countries. I noticed that many of the LGBTQI Bella's articles on Indonesian Wikipedia are translation of the English ones. And most of these articles focus on social events and perspectives from the northern countries, which might not be representative of the perspectives and social dynamics and live experience of queer community in other regions, especially those living in global South. Therefore, I would love to have more queer Indonesians and allies to contribute to this knowledge. However, one thing that I experienced when I first created an account on Wikipedia is how intimidating and overwhelming it feels to start carrying out the first tasks on Wikipedia, let alone to write the first article. There are so many terms and technical requirements and procedures that you need to learn and get used to, which makes sense because we have standards to keep up. And I think it only makes sense for the contributors to try to keep the standards. However, at the same time, this might discourage people, especially those coming from marginalized communities, to contribute to Wikipedia. And they might be less inclined to spend their time and to volunteer to write articles and try to build this knowledge on Wikipedia when they are already facing a lot of difficulties in real life. Therefore, I believe that it is important to address these issues. And I believe that events such as Editorton would help facilitate and increase the participation of local queer community in creating native queer content on Wikipedia. And I would love to see more of these kind of events. I believe that through this type of events, we can provide free training and workshop for new editors on the procedures and standards of Wikipedia, and we can jumpstart their contribution to knowledge building in Wikipedia. And this can also facilitate networking and knowledge sharing between queer individuals and their allies, which is important as queer people, especially in Indonesia, rarely have this opportunity to socialize and collaborate, openly with each other, to do the scrutiny from the government and the extremist groups. So I think this would be a great idea to organize more Editorton and queer Wikipedia events, especially in the global suck, for queuing on populating Wikipedia with more native queer content that are relevant with the local queer community. And hopefully, these kind of initiatives would lead to more active and sustainable contribution and content creation from queer Indonesian on Wikipedia. That's also a very important point, Pasca, in the sense that we can tie this issue on increasing participation on Wikipedia, tie that with the different reflections we have been having around accessibility and around also new language technologies. So I think these are all great points to have in mind. For me, I would have loved that this conversation would be accessible in other languages than English. Unfortunately, that was not possible. And that perhaps is yet another point of reflection as we close our panel on how important practice is. Okay, I think our time is almost up. Clearly 30 minutes is not enough for this conversation. We have a lot to say. But we are grateful to Wikimania for providing us this space. And we are very sorry that there isn't a way to interact with all of you who are listening and watching us, because we also care deeply about what you have to say. Yeah, so please reach out to us via our social media channels. You can find us with a handle at who's knowledge. And let's continue the conversation there. And to Pasca Dumisani and Ishan, once again, thank you so much. It has been an honor to be in conversation with you. Thank you and goodbye.