Q&A with HRC Investigations Lab alumnus Devon Lum

Maggie Andresen
Human Rights Center
20 min readNov 1, 2023

Exploring HRW’s recent investigation into mass killings of Ethiopian migrants at Yemen’s border by Saudi border guards

A video by Human Rights Watch depicting their recent investigation.

The following Q&A is between HRC Co-Faculty Director and Investigations Program Director Alexa Koenig and HRC Investigation Lab alumnus Devon Lum. During his time at the Human Rights Center, Lum worked on investigations such as “Myanmar’s Junta using bodies to terrorize” and “Violence Against Health Care: Attacks During a Pandemic.” After graduating, Lum worked as the Lab’s coordinator for one year. Currently, Lum is working as an open source research assistant at Human Rights Watch. He was deeply involved in their recent investigation into the murder and abuse of Ethiopian migrants by Saudi Arabian border guards, which was covered in every major media outlet and has led to statements from the governments of Ethiopia, the United States, Germany, and France, and United Nations High Commissioner for Human Rights Volker Türk. This interview has been edited for length and clarity.

Alexa Koenig: Thanks so much for joining us today, Devon. I was wondering if you could start by introducing yourself and your current role?

Devon Lum: My name is Devon Lum, I’m an open source research assistant at Human Rights Watch, where I’ve been for two years. I’m housed within the Digital Investigations Lab. I handle a lot of visual investigative work with videos and photos, but I also work on other types of open source research. I’m kind of a catch-all researcher, and I’m currently based in Hawaii.

Portrait of a man smiling with glasses and brown hair (left), next to a portrait of a woman smiling with blonde hair (right).
Portraits of Devon Lum (left) and Alexa Koenig (right).

Alexa Koenig: While we’re on the thread of Human Rights Watch, can you tell us a little bit more about your team and what it’s like to work there?

Devon Lum: We are made up of six researchers: two geospatial analysts, a data analyst, a spatial visual researcher, a head open source researcher, and then me. In addition to those researchers, there’s also managing director Sam Dubberley, and an associate. Both of them help us with research, but their main focus is more on managing projects and helping coordinate our interaction with the rest of HRW.

More generally, HRW has different divisions that focus either on thematic or regional topics of interest. So for instance, we have the Asia division that’s obviously focused on human rights abuses that happen in Asia. But we also have a separate refugees and migrants rights division that works across a breadth of different locations, but is specifically focused on the rights of refugees and migrants.

We’re most similar to those thematic divisions, but with research strategies as our “focus.” Remote research is our overarching theme, and because of how broad remote research is, we work with most divisions across HRW. We get in touch through requests with researchers who are interested in incorporating either visual, data, or geospatial components into the research, and are guided by those regional or thematic experts throughout a project. We then add our insights, and merge our findings into a final product.

This manifests in short-term projects or longer-term projects that can take several months, or even years. Recently we’ve been leaning into longer-term projects, so less focus on rapid-response verification of videos and photos, and instead digging deeper into not only visuals, but lots of information types we find that relate to human rights abuses.

Alexa Koenig: I would imagine that there must be a need for both rapid-response team members that can verify videos as they come out and become relevant, but also these longer deep-dives. How is that balanced within your team?

Devon Lum: Right now we’re responding by doing everything that we can, taking on as many projects as we can fit into our schedules, and handling both the rapid-response and the longer-term projects. But other divisions have started hiring their own open source researchers as well. For instance, the Crisis and Conflict Division — which is generally more focused on those rapid response projects — just hired a research assistant who carries out duties that are almost identical to mine, but is primarily focused on supporting their team. There’s expanding interest across the organization in building out open source skills, which will hopefully reduce the amount of hyper-specific rapid-response projects that we need to work on. I doubt we’ll ever be fully detached from those.

But I do think that those trends, in addition to our team running trainings on open source research within the organization, will reduce our workload and give us more opportunities to focus on long-term investigations. We also have a weekly team meeting that started out as just being for the Digital Investigations Lab, but now incorporates people from other teams who work on open source as well. So we have a spreading open source community within HRW, which has been great so far.

Alexa Koenig: That’s really interesting. I think most large institutions at this point are really grappling with solutions to this: Do you create one person with open source expertise on each team, or a couple of people within each team, or do you centralize? I do think that kind of hybrid centralized/decentralized model with teams having a person who at least knows the basics, but then having a centralized team that can go deeper on projects, probably makes a lot of sense.

I want to congratulate you on your recent investigation into abuses of Ethiopian migrants by Saudi Arabian border guards. I know this has garnered a lot of attention, and that it has potential for significant impact. I was wondering if you could tell us a little bit about how that project came about, and how you started working on that issue?

Devon Lum: Human Rights Watch has focused on the “Eastern Route,” which is the migration route from the Horn of Africa across the Gulf and into Yemen, and then subsequently into Saudi Arabia, since at least 2014. Prior to this most recent report, we had identified instances of abuses against Ethiopian migrants in Yemeni and Saudi detention facilities, and even killings of migrants that had happened along the border as well. But in 2022, my colleague Nadia Hardman started to hear from her sources that the number and frequency of these killings of migrants by Saudi border guards on the border had risen rapidly, and that those killings appeared to be deliberate, both in escalation and the manner of the targeted killings themselves.

We had started hearing that Saudi border guards were using explosives against migrants at the border, in addition to small arms we’d been seeing. We heard that groups of several hundred would be reduced to tens in a single crossing attempt. So those details, combined with a U.N. report that came out in October 2022 that noted the same systematic pattern and increase in killings, motivated Nadia to start the project in late 2022. Nadia reached out to our team because she had worked with one of our members in the past on a similar topic in the same area, and she knew the value of visually substantiating whatever interviews we could gather. She filled us in on the project in November or December 2022, and we got started the following month on searching for visual substantiating evidence.

Alexa Koenig: And have you been working on this project in parallel to others? Or has this been a deep dive on only that investigation?

Devon Lum: Yeah, definitely. Earlier this year, I worked primarily on a visual investigation of state crackdowns against Peruvian protesters. In that project, I collected and verified over 37 hours of video footage and 663 photographs of the protests. I substantiated testimony from multiple different areas across the country where protesters had been killed or had been fired at inappropriately by security forces, and either rebuilt those specific cases that we had testimony for by connecting the names that we collected to the videos themselves, or I rebuilt the entire day’s event. That project was happening from January to April and at the same time, I started collecting content for the Saudi project around late February or March.

Alexa Koenig: Can you talk a little bit about how you conducted the investigation, like what kinds of methods went into this?

Devon Lum: It was very HRC Lab-inspired. Once we gathered the first batch of interviews, my team and I started reading through them and identifying key locations that kept occurring in the testimony. We learned the names of two of the camps where migrants would be transported to before crossing the border. The survivors we interviewed also provided the general location of where they were brought to, and a rough idea of where some of the killings were happening. Because we knew where along the border the abuses were allegedly happening we were able to develop a general idea of the landscape that would be shown in videos of interest. However, we didn’t have the correct spellings of the migrant camps, as there was quite a bit of variation between how people pronounced the names of the camps based on linguistic differences or variations in experience.

In order to start the visual investigation, we started writing down potential spellings for the camp names in different languages, and then used them to search on various social media platforms for any visual content that represented what the survivors had told us. We were looking for the migrant camps themselves, visuals of the crossing, the use of explosive weapons, that type of thing. After searching with a variety of different spellings, we found one page that very clearly was posting videos from one of the camps. Within the comment section of some of those videos, there were comments from other people that we looked into who also posted videos from those and other camps. They gave us a starting point of exactly what we should be looking for visually and the correct spellings for each location, and allowed us to build up this network of sources who were living in the area. Those sources included smugglers of migrants, border residents who did not appear to be involved in the smuggling, and then migrants themselves who were posting these videos.

Once we found the videos, we archived them, because we didn’t know how stable these accounts were. To do that we used an in-house archiving tool that downloads the content, saves the HTML file, takes a screenshot, and makes sure that if we ever need to prove where this piece of media came from, we have all the requisite information.Then we organized that archive into a spreadsheet with tags for what was depicted, so we could very quickly filter through and try to identify overlapping events or people of interest. After the content discovery, we tried to establish “where” and “when” each piece of content was recorded as well as “who” and “what” was shown.

We tried to identify specifically where the migrants were crossing. To do that, we looked for videos that showed large groups that were in the mountains or on a well-defined trail, and then used the landmarks like foot trails, trees, and mountain ridges that were visible in the videos to geolocate where those videos were recorded. Once we had an idea of where migrants were walking and had geolocated a number of these videos that were all in the same general area, we identified foot trails visible on satellite imagery that connected these videos. Through that process of geolocation and geospatial analysis we were able to reconstruct a route that migrants were taking from one of the camps. In addition to just the trail itself, we needed to verify that each video was actually from the time period that we were interested in. So we geolocated as many videos as we could, and then did reverse image searching and general searching of each video to identify when it had first surfaced online to the best of our abilities. We also compared the same landmarks we used to geolocate the videos to dated satellite imagery, in order to identify when the videos could have been taken. For instance, there were regular changes in the number of tents on the border, because of the changes in migrant flow. Whenever we saw a video with tents, we could quickly cross reference to when they looked that way on satellite imagery, and get a better idea of when the videos were filmed.

A map of the border between Saudi Arabia (left) and Yemen (right) in black and white with two foot routes commonly taken by migrants highlighted in orange and pink.
3D model of likely Saudi border guard posts and patrol roads near fences identified with satellite imagery near the migration route from the migrant camp of Al Thabit in Saada Governorate, Yemen, into Saudi Arabia. Graphic © Human Rights Watch.

In addition to the “where” and the “when,” we wanted to confirm that the people in the videos were Ethiopian, because that was a key part of our testimony. To do that we hired external translators to translate the videos and identify what languages were being spoken. The spoken languages depended on who posted the videos. Border residents posted videos with spoken Arabic, and people from Ethiopia were speaking Afaan Oromoo if they were Oromo, and Tigrinya if they were Tigrayan, confirming those in the videos were Ethiopians. We also counted to the best of our ability the presenting genders of the migrants as well, because we had received information that the majority of those being targeted were women and children. What we found in the videos was similar to the testimony, there was a very large portion of women present in the migrant groups.

We often received images of wounds and of victims from those we interviewed, and we sent those images to forensic experts who identified to the best of their ability what could have caused those wounds. Through their analysis, they were able to determine that the wounds exhibited by migrants in a number of photos showed clear patterns consistent with the explosion of munitions and other characteristics consistent with gunshot wounds. This further substantiated the information given to us by survivors about how Saudi border guards were targeting them.

On top of the visual evidence that we collected, we also conducted geospatial analysis of the border itself, using satellite imagery to look for increases in security structures in positions of Saudi border guard posts, and any armored vehicles or weapons that were present nearby the border. We also looked for burial sites, and found eight of them, four of which had expanded since the start of 2022.

Alexa Koenig: Thank you, that’s really helpful. I find that there are so many organizations that are starting to try to assess whether they should have open source investigation teams, and having this kind of detailed explanation can really make it tangible how you can start approaching some of this work and the value added for humanitarian purposes in other contexts, so thank you.

I know Sam Dubberley has always taken the psychosocial resilience piece of this very seriously. Given the range of materials to which you and others on the team were exposed, what were some of the ways that you all as a team either tried to take care of yourselves or each other? Was that even possible, given the pressures that you were facing?

Devon Lum: It’s always possible to take care of yourself and your teammates during projects like these, to some extent. I don’t think it always happens, but in this instance we had really solid communication between all the team members on how we were doing with the material.

There’s a few key instances I remember that reinforced team resilience during the Saudi project. The first time I ever spoke to our lead researcher Nadia Hardman, she was talking to me about interviews she just conducted with survivors who had been injured, lost limbs, or seen extremely horrific things. As she was walking me through this information, I had to stop her to check in and ask, how are you doing with this? I wasn’t myself talking to the survivors, but from listening to Nadia for a few minutes I already felt the gravity of what they were saying and the weight their stories would put on her or anyone who interacted with the material. That interaction really jump-started resilience for the rest of the project. Sam also told me at one point that he was aware of how much I was working on the project, and how intense the content was. He told me if I wasn’t feeling up to it at any point that I should take a break and not push through. It’s not worth exposing myself to that much graphic content if I’m not going to be okay, as a person.

As a team, the Digital Investigations Lab has had several sessions with an external consultant who walked us through the basics of vicarious trauma, second-hand trauma, different types of reactions that can be experienced while watching content like this, and strategies for dealing with those reactions. Our team has also held open spaces for discussion throughout the year. So we have a formal understanding of resilience through those sessions, but during the work itself we prioritize flexibility and consistency with when we check in with each other. For instance, incorporating casual wellness checks during meetings as much as possible and acknowledging when someone on the team needs to step away to process what they’ve seen before continuing working.You can have as many protocols about resilience formalized and written down as you want, but when it comes down to it you won’t always be able to implement those protocols. Having flexibility is really key to protecting everyone’s mental well being. A shorter way to answer the question is that there was really good communication throughout the project, and a really open understanding of where everyone was with the material.

Alexa Koenig: Yeah, I think it can be so hard to carve out the space. And different things help people in different ways; everyone has a different kind of cocktail of what is supportive and what isn’t. Many people use breathing and meditation or mindfulness exercises, but I get panic attacks when I do them. That really underscored for me that we have totally different ways of dealing with the stress of human rights work. So thank you for sharing all of that.

What has the reaction and response been to this report coming out?

Devon Lum: It’s been crazy. I have never worked on something that has received the attention that this has. I knew it was going to get attention, but didn’t expect the amount that it received. In a media summation report that was sent around inside of Human Rights Watch, it was mentioned that HRW got almost 18,000 media mentions in the following week after the report was released, which was probably the largest number we’ve ever had. There were follow-up reports that were carried out by The Washington Post and The Guardian that exposed the involvement of the U.S. and German governments in the potential training of Saudi border guards, which we had not reported and our own research. So additional findings and additional pressure were placed on these states and the international community as a whole, and those states responded.

The U.S. State Department called the report’s findings concerning, and raised their concerns with the Saudi government. The German ministers said the same, and subsequently ended their training program with the Saudi border guards. The State Department didn’t mention why they ended the U.S. program which was providing training to Saudi forces this June, but we were carrying out discussions with them internally about the findings of our report before it came out and believe that we may have influenced their decision on continuing that program. So there’s been tangible changes, but we’re still hoping for a U.N.-backed investigation into the killings. We know how quickly something like this can slip from the attention of states and the public, and are really hoping to continue putting pressure on international bodies until there is an established U.N.-backed investigation.

Alexa Koenig: Well, first, congratulations! That’s huge, I think it’s such progress for the human rights space to get even a single government to acknowledge the work, and to acknowledge what people have been suffering. What do you think it was about this investigation that prompted so much attention from the media and governments?

Devon Lum: I think it was a combination of a lot of things. The scale of killings happening at the border is so shocking and so underreported that it took a lot of people by surprise, and that led to a massive dissemination of the work. The rigorous way we conducted the research, and the way we presented it through various reports, op-eds and videos, also encouraged its spread. Our communication and collaboration with different news outlets before the release gave them an opportunity to conduct their own investigations and craft their own stories for simultaneous release. And,this is conjecture obviously, but we released the report in August and Europeans are usually on vacation in August, so there is generally less reporting like this published during that period. But that’s less about the actual work itself, and more about circumstances it was published in.

Alexa Koenig: Thank you. Maybe we could go kind of back in time a little bit to sort of how you came to work in digital open source investigations in the first place?

Devon Lum: Yeah, I’d love to. So in 2018, a housemate of mine started telling me about this group that she had just joined on Berkeley’s campus. She told me about these amazing guest speakers, the really interesting skills that she’d been learning, and all of these real-world projects that she’d been involved in. I hadn’t heard of anything like that on Berkeley’s campus, or really anything that had such a level of potential impact that was accessible to students. It turned out that she was talking about the Investigations Lab at HRC. So I started looking into it and applied the following spring, and luckily I was accepted and started working there in February 2019.

I immediately took to it. I started with the Digital Verification Corps through Amnesty International, and after a couple of months I knew that I wanted to do this professionally. Although it was compelling to investigate and rewarding to dig up information that could change opinions or even change the status of people’s lives, it just really hit me that I had a route into improving the lives of people that I was nowhere close to physically. It was a very powerful first few months in the Lab. I was a team lead for a year, then I was lucky enough to take on the coordinator position. I eventually got the Human Rights Watch position and have been going strong since then.

Alexa Koenig: One question I always love to ask alums of the Lab is how do you feel the HRC Lab experience prepared you for what you do now professionally? What would have helped prepare you even more?

Devon Lum: In terms of how it prepared me, it’s very clear. I use the skills that I learned in the Lab every day in my work. I use geolocation, discovery, and analysis of online data — they’re all very key parts of my work. I also think the focus on resilience early on was really important. I’ve met a lot of researchers who don’t work with visual information as their primary mode of research, but either interact with graphic imagery through our team, see it online, or are sent it directly from people that they interview. They are really shocked when they see it, and haven’t been taught skills on how to manage their reactions to that type of content. So I’m really happy that I have that background and am able to take a step back, take breaks, and meditate. I felt very prepared not only for the skills that I would need for this profession, but just knowing what it takes to manage the mental side of things through the Lab.

Regarding what I wish I had learned, it’s hard to say because I do feel like I’ve been set up so well for this career path. I think just improving on a specific skill set, probably more data analysis or more coding skills, which I’ve been starting to pick up through the work itself. I had very little exposure to those types of command line tools before going into my current job, and projects at the HRC that incorporated them often required prior knowledge or expertise to join So more exposure to that would have been helpful.

Alexa Koenig: Great, that’s really helpful. And what advice would you give to students in the Lab who are ideally going to Human Rights Watch someday, or have that as their goal?

Devon Lum: Keep working, keep doing what you’re doing, and keep the passion that you feel about the work as well. Looking through a lot of videos or documents and trying to find one word or one moment can be really grating, but you never know what you’re going to find. I also think it’s helpful to remind yourself what could happen if you do find that one smoking gun. So keeping your passion, your focus, and talking to as many people as possible while you’re in the Lab is all important. As a student, you have the opportunity to meet your project partners, whether they’re at Human Rights Watch or elsewhere. I initially got my position because of a relationship like that, to some degree. I needed to apply and interview of course, but I knew people at HRW through the Lab. The students are in such a unique situation at HRC where they have access to these people and can really get to know them, and pick their brain. So make use of that as much as you can.

Alexa Koenig: Great, thank you. What do you see as the future of digital open source investigations? Where do you think the field needs to go or do next?

Devon Lum: I think that there needs to be a continued democratization of the skill sets that I use, and that the Lab uses. The skills that we use are in some ways challenging, but in a lot of ways very accessible once you’re given the tools needed to do them. The more people are taught how to do things like verify videos, the greater literacy the public will have with things like fake videos, and the more protection we’ll collectively have against misinformation or disinformation. We’ll also have more people around the world to document human rights abuses.

And that kind of pans into another point which is harder to crystallize for me, but one that I’ve been thinking about for a while, which is that governments really need to start using this information in a productive way. In response to our Peru Project, Peruvian President Dina Boluarte tried to minimize our findings by calling us an NGO then saying the events that took place following the start of protests in December 2022 were complicated and required further investigation. Two weeks later, Peruvian state officials directly contradicted visual evidence we had verified showing Peruvian security forces killing civilians inside and outside the grounds of an airport. Their statements came months after the killings took place and demonstrate an intentional rejection of widely available, easily verifiable evidence which had been available for analysis since the day the abuses took place. I think it’s the responsibility of governments and international actors to stop feigning ignorance of this material, and to establish measures that will allow them to use it to improve the human rights realm and hold perpetrators of human rights abuses accountable. This information is available. Our team at the Digital Investigations Lab is tiny, so our resources compared to governments across the world are completely negligible. I think there’s a huge amount of room for governments to use their resources to act in a beneficial way.

Now, the reason that I’m torn on pushing for this is because I don’t trust governments to use these methods in a good or responsible way. I don’t want even more open source intelligence operations to be used to target individuals or to be used maliciously, but I do think that there is so much room for this type of work to benefit human rights through its use or acknowledgement by governments around the world.

Alexa Koenig: It’s such an important point that you’ve just made. I think for all of us in the human rights space, our job is to hold governments to account. And it’s fair to say some of them probably are using a lot of these methods to surveil and to spy, but your point about not being able to feign ignorance is so important. I remember saying back in the spring of 2017 that in five to ten years, it should probably be considered legal malpractice if you’re a lawyer who doesn’t have some basic understanding of how to do this research. I still stand by that, it felt like it was a bold claim at the time, but I think you’re taking that point and even just broadening it out to the responsibility of governments. They shouldn’t be able to say, “how should I know,” when you can find it with a very small team at Human Rights Watch. Is there anything else you’d like to add?

Devon Lum: I just want to personally thank you and the Lab for my current employment, my experience, and all the opportunities that I’ve gotten over the years. I don’t know if I would be able to work in a field that’s as rewarding as this one without the experiences that I’ve had through the Lab. So thank you for that.

Alexa Koenig: Thank you. The Lab is such a collective effort and I’m really grateful to individuals like you who’ve both done the work and then played the leadership role to make the Lab what it is, and constantly pushed to make the whole space better. So thank you for all that you’re doing at Human Rights Watch now, and have done at the Human Rights Center.

--

--

Maggie Andresen
Human Rights Center

Maggie Andresen is a freelance journalist and runs communications for @humanrightscenter