Talking Tech with Mona Elswah & Emna Mizouni on Moderating Magrabe Arabic Content on Social Media
In today’s episode, we dive into an often overlooked but incredibly important issue: content moderation in non-dominant languages, specifically Maghrebi Arabic dialects. Recently, CDT released a groundbreaking report titled "Moderating Maghrebi Arabic Content on Social Media," which explores the challenges, gaps, and opportunities in ensuring fair and effective moderation of Arabic content on social platforms.<br><br>To unpack these insights, we’ll be joined by two leading experts in the field: Mona Elswah, a researcher in digital rights and content moderation, and Fellow of the Content Moderation for Languages in the Global South Project at CDT; and Emna Mizouni, a Tunisian digital rights advocate, community leader, and founder-CEO of Digital Citizenship.<br><br>We’ll discuss the unique hurdles faced by Maghrebi Arabic speakers online, why it's critical for platforms to address these gaps, and what needs to change to make content moderation more inclusive and equitable for all users.
Top Keywords
- region 0.024
- mona 0.022
- maghreb 0.017
- maghreb region 0.015
- arabic 0.014
- content 0.012
- companies 0.009
- report 0.008
- moderators 0.008
- dialects 0.007
- moderation 0.007
- tech 0.007
Transcript
Welcome to CDT's Tech Talks, where we dish on tech and Internet policy while also explaining what these policies mean to our daily lives. I'm Jamal Magby, and it's time to talk tech.
Welcome to Tech Talk. Bye.
Today, we're exploring a fascinating and often overlooked issue, content moderation and nondominant languages, with a focus on Maghrebi Arabic dialects. CDT recently released a report titled moderating Maghrebi Arabic content on social media, which sheds light on the challenges, gaps, and opportunities in ensuring that social media platforms handle Arabic content fairly and effectively. Joining me today are two incredible experts, Mona Elswa, a researcher specializing in digital rights and content moderation and fellow of the content moderation for languages in the Global South project for CDT, and Emna Mizoni, a Tunisian digital rights advocate, community leader, and cofounder and CEO of Digital Citizenship. Mona and Emna, so great to have you on today. Hi. Thank you. Thanks for having us. Of of course. We're happy that you could be here. So before we get started and before we jump into, like, the meat and potatoes of all of this, Emna, would you mind giving us a little about your background and what brought you to this specific issue? So I think,
my interest, not my work at the time, my interest brought me to talk to people, brilliant people like Mona. So I I am a Wikimedian. If you know Wikipedia, you know that it's part of a big movement and then more than the Wikipedia the Wikipedia project. I contribute to those projects, Wiki Commons, getting people together in conferences, regional conferences across Africa and the Arab world. And it got me started to understand what's exactly content and how to add to it, how to moderate it, what's good, what's bad, and who decides what's good and what's bad. And, in my interest, I I grew into wanting to know more and contribute more with my background and my languages. And that's how I met Mona, and I became actually more interested from the questions that she first asked me in person.
That is fantastic. Thank you so much for sharing. Mona, so let's let's just jump into it. Can you give us a brief overview of this new CDT report and why this topic was chosen as a focus language?
Of course. So the Malaga region is despite having lots of people, like, more than 100,000,000 people in this region, it is very understudied, very underexamined, especially in the conglomeration field. The use the usual focus is on the Eastern part of the Arab world, like in Palestine, on The Gulf, in Egypt maybe. But the Maghrib region is absolutely absent in this field from this field, and we wanted to be the voice of this region on of these various dialects of the Maghribi Arabic in the region. And we are the first report to actually focus on this, and I'm so proud of this. And, also, region is Maghribi Arabic is a, a low resource language. So it also have an additional layer of its, challenges to automate the collaboration unlike the Masriki Arabic, which is not a low resource language. So this is why we picked the Maghrebi Arabic dialects compared to other dialects in the region.
The report highlights,
a few key challenges. Could you walk us through some of the main findings? Of course. So the first thing that we found to be super interesting is that condom moderation approach there are different conglomeration approaches by tech companies. So we all think of collaboration that, you know, tech companies apply, all the same methods, all the same policies, and, like, they are exactly the same. But, no, actually, companies have different ways in the way you think of quantum moderation for the majority world, for the global South. And we found that US based social media companies are more likely to apply what we called as global quantum moderation approach, which is basically assuming that all users are the same and therefore applying all policies all over the world. No exceptions. So a user in The US will have the same treatment as a user in Tunisia, for example. And that has its own pros and cons. If you compare this to what we the other method would be found is a local collaboration, which is employed so far as we found by TikTok, which is a a bit of personalizing and localizing some policies to meet the cultural context of the Maghreb region. And we found this approach, again, has its own pros and cons. But these different schools of thinking of collaboration, we we found to be very, very interesting. And for us as in the in the field of tech policy that we also need to think when we approach tech companies and make recommendations, we need to know that they think, you know, differently. So this is the first finding. The other finding that conno moderators for Arabic language are assigned content from across the region no matter where are they where where are they from. So, a moderator from Egypt will be assigned content from Tunisia and vice versa, which is, as you know, Amna, we we speak in English because we don't understand each other's dialects, especially, like, I struggle to understand the Malarabi, different Malarabi, dialects.
Yes. We struggle.
I've been it's very good with the Egyptian dialect. I'm I'm so bad. There was a Tunisian Algerian dialect. I will never get it. But although I wouldn't, an Egyptian moderator will be as could be assigned content from Tunisia, and they have to make decision, which is really we found to lead to mistakes and errors. And this is what moderators told us that sometimes they did not understand what's going on, and they just have to do it to to, you know, to action the ticket as they they called it. So that's another finding. And, also, we found that tech companies don't diversify when they when they when they hire NLP teams. And this is something that is common in the NLP field, but still does not lead to fair and equitable models. And we can do better in, you know, hiring, diversifying the NLP field in tech tech companies.
If, if I may add something, the assumption of applying the same rule to everyone everywhere led to the fact that the whole Arab region, North Africa and The Middle East are as seen as one cluster instead of a cluster that is full of diverse backgrounds, culture. And so, for example, if we talk about, what just, Munna just said right now, like, an Egyptian would be assigned a ticket from, the Maghreb region, and they are supposed to understand the cultural background, the colonial background, the indigenous background, the Amazigh, and everything, every aspect of the language and dialect and culture, all embedded in a comment sometimes that it's hard for even the Maghreb region, speakers themselves to understand when it comes, to differentiate between content from Maghreb from Morocco to Tunisia. And so this clustering regions without really understanding the background and making it look like it's the same, they all speak Arabic or they all have Arabic as the first official language, it would work. That's a huge mistake and that we saw lots of repercussions of such a policy, basically, in the Maghreb region.
And I think, I wanna keep walking us down this track. And you've worked extensively with digital communities in North Africa and and beyond. What are some of the main content moderation challenges, especially for Magrabi Arabic speaking users? There is not enough, content moderators from the region.
That's number one. Not enough. If I may name for example, when it comes to TikTok, we know that TikTok, they are very responsive, and they they understand the difference between the regions between Maghreb and Mashrakh, for example. But in different platforms, it's not the same. I know that some of them, they are trying their best to put the context and understand the religious minorities, the cultural minorities, all of that. They try to understand it, but it's not enough. They don't have enough bandwidth to do it within the platforms themselves, and they do not outsource enough. And even as here I speak with the head of digital citizenship that partnered with, Mona and CDT to do the work in the Maghreb region, we try to approach those platforms. Because like you said, we have a knowledge of the Maghreb region, and we try to use the platform and our connections to them, but they don't have enough capacity to respond to the cases that we bring. We are only one entity. We're not bringing all of the cases from Morocco or Algeria. We're just like our main focus is Tunisia and the most imminent and dangerous cases from the other countries we bring them in, and they don't get picked up easily. And so I think the the challenge of, like, not having enough human resources and not having enough attention, because let's speak honestly. It's, it's not profitable for those platforms to allocate, human resources for those countries because they don't generate enough revenue comparing to other markets. I mean, if we put all of our countries together and compare it to the Egyptian market, maybe we could be balanced. Maybe. And so it's, it's not only a culture dominance also. It's an economic
decision. Mona, I wanna turn to you and kinda build on a point that, just made and something that that you touched on in your report, and that's the role of human moderators. What are some of the challenges they face? And I know you both have kinda touched on different points throughout our conversation, but I'd just like to dig a little bit deeper in here, you know, besides dialect. Are there any additional,
challenges that these human moderators face, and why are they important? Talk generally on human concentration. In general, this is the most horrific job that a human have to do, to be honest. The the majority, of the common moderators I spoke with, if not all, in all regions so far, are either having PTSD or had they are or sometimes even, they are on medications. Like, it is so bad. Like, it is so bad. Even in this region, like, in the Arab world, I didn't know that, you know, the commudoration is, like, a huge business in this region. It is very it is what also makes it very difficult, as you said, Jamat, like, one of the most challenging is it, like, the secrecy around it. Like, they have to work in the dark. They can't tell and share, you know, with their spouses, with their friends, what they do and what who did they work for. They are not even sure who do they work for. Do they work for the vendor or the tech company? It is very, you know, it's it's it's very, the secrecy culture that was built on this job is really one of the challenges and makes it even harder on the human moderator. And, also, the exploitation of the work is not just, you know, the bad bad payments, which is true, but also that they are exposed a lot of traumatizing work without enough training. Like, they are trained in just two weeks to understand a lot of the policy that these companies have to employ. And we're not talking about just, you know, 10 policies and that's it. We're talking, like, a lot of a lot of policies, a lot of interpretations. And even some of these interpretations are not very clear, so they start the job and they have to, you know, to pick it up later. But we're talking about two weeks of training. And during the training, they are showing, like, you know, celebrities, pages and accounts as examples. This is not the kind of content they review. It's way more traumatizing. And, again, they deal with this fact later in the job. And once they start the job, they they are either thinking of leaving, which many of them do in this region, or they think like, okay. Let's just, you know, stick to it. It's not the worst job I can do it, and then eventually they leave. The the also, it impacts the collaboration, you know, cycle because they they like, in the Maghreb region, I found them to be, like, you know, leaving the job alone. Like, a lot of them quit and resign and because they can't continue. And and then these vendors have to rehire new people, retrain, and then, you know, etcetera. And it it leads to further slowing, and the the the the error the the error cycle continues. So there are many challenges, but the the worst part of it is exploitation of workers all over the world of commodoreters, I guess. But this region is I I I didn't know it were it even exists, to be honest. And it exists to that larger scale that exists now.
Wow. That is that is intense, and a little scary, if if I'm being frank. And and based on your experience, what are steps social media platforms can take to address some of the issues that Mona raised and you raised throughout this conversation effectively? Oh, wow. Effectively, I don't think anybody has a recipe for
efficiency in this, in this world. It's it's swallowing us in a very scary way. I just wanna go back to what Mona said and link it to my response. It's thanks to the report that Mona led and issued that we can see the other side of content moderation in the Maghreb region. Nobody talks about it. Nobody talks about nobody saw the Maghreb region as a region that is understaffed and they're equipped and that our people are suffocating. And I know some people, even if they get, like, through the cycle, with, being outsourced, they get through the cycle and become managers, say, of the content moderators, they decide to quit because it's an inhumane job. It's it's not a joke when you see in one of the cases, somebody, one of the those content moderators said and, of course, we can't name the person or name the platform because they are signing, NDAs and stuff. He said they saw in more like, a week time, maybe, they saw more than 200 pictures of genitals, human genitals, in a week time. Any human being. I'm not talking about the violence they're exposed to. There is so much violence. Mhmm. And we've seen with leaks in, for example, I speak of Tunisia with leaks of videos, child, sexual exploitation, women violence, and, like, videos of men raping women that goes on those platforms. Not everybody sees it publicly, but they also moderate, in some cases, them direct messages. Right? So this is too much to carry as a normal person, and I don't think they are giving them enough, psychological support, if any. So to be effective, maybe they need to to be more open about this role. Our content moderator give them more time to study, the policies, automate more in some cases. And with the automation comes the other, the the appeal possibilities. We don't have enough, appeal possibilities. So that's why we try to push back on, the automation of content removal. But maybe that would spare these people, actually, some horrific content. So more time, psychological support. They need to get out of the shadow, basically.
So it sounds like we're calling for some transparency as well. Absolutely. Absolutely. Turning Mona, turning back to you, the report outlines several recommendations for platforms, policymakers, and civil society. What is one key action point you'd like to emphasize?
Better communication with users. And, I mean, I'm not an an digital citizenship or doing great job, and they have taken the burden of being the middleman, you know, connecting users impacted with tech companies. And this is a job that many civil society organizations have taken on them. You know? It's it's it's a lot of work on them, and they're doing their best. And, you know, this is a job that should be done. You know, tech companies should, first, appreciate this job. The second part is, they should do better communication, direct communication with users, better, you know, provide better reporting channels. Like, users can actually report and actually more transparency on what what happens with these reports, not just the automated boring messages. Like, there should be a better listening to the users and better feedback, and this could be easily done. And this could get back the users need to retrust these companies. These companies harmed lots of users in this region. And they you know, this impacts these their business as well because they are, you know, turning away from Western social media companies and moving to non Western ones. So this trust needs to return, and to return, you need to enable better communication. And, with better communication, these companies could learn more about the region and could hopefully change, the way they think about cogent ratio when it comes to the global South and the mother of region more specifically. And this is the one recommendation I will pick among the others. But
The I I know it's a hard choice because there there are a couple good ones that I think we could we could spend all day on. Before we end this conversation, and and I wanna turn this to you first, and then, Mona, I'd like to hear your thoughts. I would just like any final thoughts or one key point you would like any listener to leave with, after this conversation?
A key thing would be massively report things. Like, use the report button as much as you could, but be judicious. Not everything that you dislike is something that should be taken down or reviewed. That's one. But the most important item to say that we need resources to basically counter all of the hatred and misinformation and disinformation and all of those atrocities that are happening on social media. All of the platforms here, we're talking about like, when I said the westerns and non westerns platforms, all combined, we need more resources. And if there is a way, maybe, if there is a way to limit, the usage of people in certain way to reduce the amount of, content. I mean, this is radical kind of, but it could help certain people to limit them from, producing content if they showed abuse of the policies, but then make the appeal more and more, accessible. Because appealing on decisions, from those platforms has always been unclear and not granted in most of the time. So if you take radical measures, we're okay as long as you give appeal options and, like, properly appeal, and listen to those cases.
And, Mona, any key point you'd like our listeners to leave? I think we need more people like Amna in the region,
and more digital citizenship. Like, we need more pressure on these organizations. We need more voice. We need more research, on this region. Like, it's it's it's just so sad. Like, the amount of research on Arabic accommodation in general is just so sad. And on the Maghreb region, it is almost doesn't exist. So we need more research, more pressure, and users should keep pressuring, social media companies to do better moderation.
I think those are good good stopping points for us, but Mona and Emna, we have to have you both back soon. I don't. There is another report, a similar report that, that Mona authored that I would love to have you back and and talk about. And Emna has been such a pleasure. I think we need, as Mona just mentioned, more people like you, more people pressuring, the companies, and and making helping to to make better decisions for for all users. So thank you. Thank you both. Thank you so much for having me. The pleasure is mine. Thank you so much, Jamal. It was amazing,
to be with you today.
Of course. And to all of our listeners, if you'd like to take a look at the report or keep up with any of the work that CDT is doing, please visit us at cdt.org and follow us on Facebook, Mastodon, LinkedIn, and the social media platform formerly known as Twitter at SendDemTech. I'm Jamal Magby, and thank you all for talking tech.
