Interview with 2024-2025 Greenberg Research Fellow Christopher J. Anderson

Interview conducted by Grace Shaffer (PhD student in History, University of Southern California, and Center Graduate Assistant)

How did you come to the field of Holocaust Studies? And which came first: your interest in GIS or Holocaust rescue?

I guess I’ve always been an amateur historian. In terms of the Holocaust, it is a period of history that has interested me, both personally and professionally. More directly, in 2014 I was invited to come to a conference at Yad Vashem in Jerusalem, and I was there for 10 days, give or take. And so that led to more of a scholarly interest. I was fortunate to be in the company of some fantastic scholars at Yad Vashem. They covered a broad range of topics, including Holocaust rescue, which is where my research lies. But it took a number of years, this marriage of GIS and Holocaust studies.

I worked in geography for a long time. Academically, it was through my time at USC beginning five years ago. But also, way before that, it was in industry that I worked in geography, mostly in the real estate appraisal field in the commercial, industrial, and agricultural realms. I’ve been a geographer and using geographic information systems going back to 1997 really, before I even knew what GIS was. I recall using those systems even then, but as I progressed, I became interested in GIS as more than just a tool to be used with already prepared data sets and interrogating those data sets for relationships. I wanted to be like behind the scenes, to understand how these systems work, and from a programming standpoint, from a theoretical standpoint. So 2018 is when I started to look into graduate programs in GIS, and then I started at USC in the Spatial Science Institute in their graduate program. Immediately when I started, my first class, the very first GIS project I did was dealing with Holocaust rescue. And it’s still the same plateau that I’m researching in France.

So, to answer your question, these interests developed together. As I went more on the academic route of GIS, my interest wasn’t necessarily in all the things I’ve been doing in my career in terms of the real estate valuation side, but my true passion is more in history. And this subject matter, in terms of the Holocaust, has great meaning for me. So that’s why immediately I went in that direction.

It also ended up being something rare in the GIS field. I was invited to present at ESRI User Conference in San Diego in July. ESRI is the largest, by far, dominant software firm in GIS and they create ArcGIS Pro. If I had to look at the thousands of people at that conference and the metrics of the people doing qualitative GIS or even historical GIS, we were just a tiny, tiny little fraction compared to the professionals that are in government, managing infrastructure, and environment, so GIS is typically very much a quantitative realm. You can look at historical GIS very quantitatively, and I have done that even in the course of looking at Holocaust rescue. But the interest to me is more of the qualitative side and mixed methods really. I am looking at both — the quantitative and qualitative — to look at the data spatially, geographically, and try to understand it in that way. So long answer, but that’s how I got into this marriage, if you will, of Holocaust Studies and GIS.

You have already touched on this a little bit, but how did you developed your current project and how do testimonies play a role within the larger study?

When I was a student at USC, in their graduate program at the Spatial Science Institute, after that first class, when I had done more of a quantitative look at the plateau in France that I’m studying, I was at a conference that June in Washington, DC for UCGIS, which is the University Consortium for Geographic Information Science. There was a seminar being given prior to the start of the General Conference, and the speaker was a person by the name of Alberto Giordano, and he was at Texas State. He was presenting his research of using GIS to better understand the geographies of the Holocaust. I went to his seminar, and afterwards, I approached Alberto, and I told him about what I was currently doing at USC and that I had this idea about research into Holocaust rescue on the plateau, and what would he think about that as a possible dissertation topic? And he was interested. And within a year and a half after that, or whatever it was, maybe a year later, I saw that Texas State was accepting applicants into the PhD program for GIS. And so I called up Alberto immediately and said, “Hey, remember me? We met in DC, and we talked about this possible topic.” Maybe this is unusual, but I came into my PhD program already knowing what I wanted to do. I don’t want to say exactly, because certainly I’ve learned a ton since then about the Holocaust, about qualitative methods, about all of the different challenges that come along with the marriage of these two things. But at least it gave me that primary seed, and I was thinking along those lines immediately from the beginning.

The research has evolved a lot since, especially the more I looked into the gaps in scholarship that are there, both within geography, GIS, and Holocaust Studies. Anybody that’s familiar with Holocaust Studies knows that initially, we looked at a lot of perpetrator documentation. I definitely didn’t want to just look at numbers, and numbers tend to be perpetrator documents. I wanted to look at it from the victim’s perspective, and because I’m also dealing with rescue, I wanted to look at the rescuer’s perspective as well. Rescue isn’t usually part of the three typical categories of perpetrator, bystander, and victim, which is, of course, a problematic categorization in any case. In that whole tripartite division of people, rescue sits somewhere in an unknown category of people. So I wanted to look at the rescuers’ perspectives and understand them geographically, how people understood and created their own sense of place. As the project evolved, I also wanted to look at the perspectives of victims. All of this naturally led me to Holocaust survivor testimonies. Alberto had been working extensively with testimonies, as well as Tim Cole and some other scholars, studying how survivors would construct place. Place is a geographic concept that’s different than space. Space is like a point on a map. It’s got latitude and longitude and altitude. You can physically say, “Alright, it’s this intersection,” or it’s this mountain, or it’s this village, or this whatever. But “place” is more nebulous. It’s more of a construct. It’s an intersection between people and the earth, and how we look at a given environment and how we perceive it.

So I’m very interested in how survivors perceived this place, this place of rescue, if you will. And rescue is defined in various ways, and I don’t really have time to get into my definition of rescue at this point. But I want to know how they would have understood it, emotionally and semantically. Did they perceive it as a place of safety, a place of isolation, loneliness, or what? What was this place? On the plateau that is the focus of my research, the primary village is Le Chambon-sur-Lignon, and it’s a famous location. But sometimes the legend, the myth, whatever is is, becomes more important than the truth. I just really want to try to understand what this place was for survivors and rescuers. We have a limited number of testimonies of the actual rescuers, but we do have some, and I want to know if there are differences between how they would have described Le Chambon-sur-Lignon versus the survivors. I would expect that, but that’s just an assumption at this point.

The project, then, is taking testimonies, and they are my primary documents. And if there are existing transcripts that have already been transcribed from the testimonies, great. If not, we’re going through the process of transcribing them. After that process is done, it will take a lot of scrubbing and editing just to make sure that the transcripts are correct. And then from there, editing those transcripts down to just the period in those transcripts where the survivors were talking about the plateau, the Plateau Vivarais Lignon, which is southwest of Leon, northwest of Marseille, in south central France. Obviously, the other parts of these testimonies are extremely important and crucial to the lives of those survivors, to those victims. But I am focused on just this particular location and how it was understood and perceived. So I’m editing the testimonies down to just those parts dealing with plateau, and then using the tools of natural language processing and AI to understand all the things that I mentioned earlier, such as: what topics were discussed about this place; what sentiments existed: positive or negative; what emotions were felt: happiness, fear, sadness, loneliness, etc.

I use natural language processing. What is that? Well, that’s a series of computer tools that digitally “read” text and tell us something about it, like the topics, sentiments, and emotions. From that, there is also a tool called geoparsing. That’s another part of natural language processing, and if you have a given text, and there are locations referenced in the text, it can do “named entity recognition” to detect those locations, or toponyms. Both AI and traditional natural language processing generally are pretty good at that. But then comes the disambiguation step, also called the resolution step, which attempts to decide which toponym is it referring to. Is it Paris, France or Paris Texas? Is it Le Chambon-sur-Lignon? Or is it another Le Chambon, because there are several Le Chambons in France, and that’s just the luck of the draw. There are a lot of villages up in the plateau which have unique names in France, but unfortunately, Le Chambon-sur-Lignon is one of many Le Chambons in France. And often times, when survivors talk about their experience in the testimonies, they don’t necessarily always say, “Le Chambon-sur-Lignon.” In fact, quite frequently they just say, “Le Chambon,” or just “Chambon,” or, maybe just “Chambon-sur-Lignon,” without the definite article. And that’s where the models begin to struggle. The out of the box geoparsing model in natural language processing was giving me pretty low accuracy. So you train the model, you help it to understand. You give it examples. And teach it that “Le Chambon-sur-Lignon” is the one that they’re referring to, and I know that since I’ve listened to the testimony, read the transcript.

With geoparsing, you pull all the named places out of the text. I should also add, I’m not just looking at named places, or toponyms, but I’m also looking at the unnamed places. When they’re talking about the plateau, they may mention the woods or the river, or the barn or the attic, a hiding place, or whatever may be. There are all kinds of other locations mentioned, but they don’t necessarily have coordinates that we can assign to them. For example, “the woods behind the house.” I also care about those references. So once I get the toponyms that I want, and they’re correctly resolved to the correct coordinates, and also the non-named places like the woods, apartment, house, or farm, then I put it though the natural language toolbox via what is called aspect based sentiment analysis. In my case, the aspect that I care about is geography. What is the sentiment, then, around that location, and was it described positively or negatively? I should add that algorithms always have accuracy challenges, and you have to measure your results. Beyond sentiments, I am also trying to detect the topics being talked about. And again, the specific emotions being expressed. Lastly, geographically analyzing and visualizing the findings. Perhaps interpolating between how survivors are describing Fay-sur-Lignon, and then over at this other village, maybe they’re speaking of it this way, or the woods are talked about that way. There are all kinds of possibilities in how we visualize that and how we understand that. One of those is spatial interpolation, on which Alberto and I just published a paper in July. That was a little bit different, because it was one person, and it was his emotional journey across southern France. This is multiple people, and we’re not necessarily tracking direction, but, to make a long story short, there are a lot of ways that we can visualize this, and that’s yet to be determined, because I think, yeah, it has different possibilities.

But then the second part of this, which as far as I know I don’t believe anyone has really done, certainly not in Holocaust studies, is that I want to use historic aerial photography and compare an analysis of them to how survivors are describing place. The French took these aerial photographs of the plateau right after the war, in 1947, which are freely available, and being just a few years after the time in question, they should be more or less accurate. I will combine all these aerial photos through remote sensing techniques, creating an orthomosaic, which is like a stitched together compendium of all of these aerials about this area. And then doing land use and land cover segmentation and classification, resulting in land cover statistics, such as this much of the area was woods, this much was cropland, this much was urban or village or whatever. And using landscape metrics that measure things like connectivity and fragmentation, homogeneity and adjacency. And then compare and analyze that to how places are described by the survivors. Given the metrics about this place, and perhaps it was rural and had low measures of connectivity, how does that compare to the description being given by our survivors, how does that fit into the social networks that were at play on the plateau as well? What role did the landscape play in the social networks that existed on the plateau, say on a given farm or in a village.

Overall, you begin to look at the data in terms of addressing the “why” questions. Why was this place a place of rescue, if you will? The why questions are always interesting, but also problematic. You look at what the data is telling us, and perhaps draw some inferences, perhaps conclusions to the why questions. The last part of the research will be looking at those more quantitative things about this plateau, which I’ve already done in a small scale when I was at USC, to try to understand what was spatially unique about this place in terms of topography, economy, infrastructure, politics, and then compare those findings, again, with how the testimonies describe it. The plateau became a place of rescue, and not only a place where rescue occurred, but occurred in large numbers and occurred successfully. It seems, in a sense, easy to answer the “why” question, but I think when you really look at the testimonies, the experiences of survivors and rescuers, there’s a lot more nuance to it, and the simple answers are never satisfactory. I’m trying to look at it as a geographer, as a historical geographer, a Holocaust geographer, if you will, to really understand what happened there, and then perhaps, try to advance scholarship in some small way about this plateau.

You have served in the military as a chaplain. Does this experience inform your research?

I was an Army Reserve chaplain, and that came about after 9/11, really, and I served a fairly long time, and it is how and why I obtained my Masters of Divinity. To get personal for a brief moment, we lost a family member on 9/11, and that experience is what spurred me to serve. I felt called to serve in a way that was not in a combat role, but rather in a helping role. And one thing led to another. My experiences in my Army career with trauma, and of people and place, have had a large impact. I have been working, both in my military role and in my civilian roles, including in industry, in a way that was inherently geographical as well. So in the last several years, it’s just been this intersection of my experiences with people and with place and with trauma, and combine that with my interest in the Holocaust and the experience I had in Yad Vashem, and, so, yeah, I think my military experience is probably key. It’s certainly a big piece of the puzzle. I think that there are important things we can choose to do in our lives. And we can make a choice to be a part of something significant. I certainly felt that when I was in the military. So you could ask, “Why aren’t you still in the military?” Well, it’s hard on families, and it’s hard on yourself. I was at a point in my life after 10 or 12 years where I was ready to step out. But I still want to have a life of meaning and significance, and so I think that the same reasons why I was an Army chaplain are the same reasons why I’m interested in Holocaust geographies.

That being said, as I look at Holocaust rescue and at people who are helping, I’m not looking at Holocaust rescue because I consider it a redemptive story. Because to say that there is redemption, or redemptive stories, within the Holocaust leads down a path that has serious ethical questions. So I’m not in this because it’s a redemptive story. In fact, quite the opposite. I’m interested in the Holocaust because it’s clearly not a redemptive story. But within the midst of that, there were people who made different choices, and that’s what interests me. It is not that they made these amazing choices, and it’s a Hollywood ending, and “Isn’t this wonderful?” Because even for the people that found refuge on the plateau, many of them, many of whom were children, oftentimes they had lost their parents or other loved ones. And after they left the plateau, they still faced incredible hardships, even if they got to Switzerland. Or perhaps they made it on the plateau to the time the Allies completely gained control of that part of France. My point is that they were still victims. So there is no redemptive happy ending. Some would say up to 5,000 lives were saved on the plateau. I think that number is too high based on quite a bit of research that’s been done since that number was proposed or stated. Regardless, it’s interesting to me because there are choices that people can make even in the midst of the worst circumstance. And I think we need, as scholars, to know and not be afraid of that research either, even in the Holocaust. I think sometimes we shy away from Holocaust rescue because we want to avoid even the perception that we’re somehow furthering this “Hollywood Ending” type of thing. And so I guess I’m saying, right off, that’s not why I’m doing this. I’m not doing this because it’s redemptive. I’m doing this because it’s not, and that even in the midst of this awful circumstance, people, some people, a very, very, very small minority of people, chose to go a different way, and that’s interesting. And as a geographer, I want to know about that.

Are there any conferences or workshops that have shaped your research?

Yes, the conference at Yad Vashem that I mentioned earlier was probably the most shaping, even just for broad awareness. And then the UCGIS Symposium in 2019 in DC, where I met Alberto, that was transformative, just in terms of what came next. It was important because I was exposed to others doing what I was already trying to figure out how to do. When I went to Alberto’s seminar, I was like, wow, this is pretty amazing. And I found out what he and Tim Cole, Ann Knowles, and others are researching. And then three years ago, I did the European Holocaust Research Summer Institute through Royal Holloway in London. Because of Covid, it was not in London, it was online, but I still did it, and it was fantastic. Just incredible scholars. It was a week of Zoom, but that was fine, because of the quality of the research and the information being presented, and just the opportunity to interact with these amazing Holocaust scholars. And then I applied for the next year in 2022, when it was no longer Zoom, and I was accepted, thankfully, to go to London in person. It was extremely important to be there with scholars such as Dan Stone, Paris Chronakis, and the many other scholars that were there, in terms of how that shaped my use of testimonies. In speaking with Dan Stone, he helped me to see the importance of being sensitive to the nuances of using Holocaust testimonies. It’s easy in GIS to be positivist, because GIS is typically quantitative, it’s producing this map, and here’s the data, and boom, you know, the truth in a map. “A map doesn’t lie.” It would thus be very easy to present my findings in a very positivist kind of presentation of my findings. Dan Stone really helped me to see how I have to be sensitive in my visualizations and how the research is presented to avoid that kind of simplistic spatial explanation. The other instance that comes to mind was just a short one-hour mentoring program through the Holocaust Education Foundation of Northwest University, which has a program to have an hour with a renowned Holocaust scholar. I was able to speak with Deborah Dwork, and she was interested in my research, and she also gave me some good advice too. She told me to remember that even as you’re doing your geographic research, that I realize how important people are, their agency, and their choices. And I think that certainly resonates as a geographic approach rather than just a spatial approach. Geography always must include this variable of people. With people, it is how we interact with an environment and with a space to make it into a place.

AI capability seems to be growing every day. Do you think is can help support researchers as they do this important Holocaust research?

Whether it’s AI or even natural language processing, I think we have to approach all of these things with a lot of caution, with a grain of salt for everything. As I was saying earlier regarding geoparsing, I had to train a model. Even after four or five trainings, the ability of the model to precisely identify Le Chambon-sur-Lignon was about 35%, and that’s pretty awful, right? If you’re trying to do research and care about the geographical concepts of space and place, and if the space is off by miles or kilometers, that’s not acceptable. In the geoparsing world, accurate is considered acceptable if it’s within 161 kilometers, which is 100 miles. I just don’t think that’s acceptable, either. In the Holocaust, the difference in less than a kilometer is the difference between being interned in a camp or not. Or the difference between safety and danger, between life and death. All these tools are very, very imperfect. AI is very good with toponym recognition. So in other words, oh, here’s the place, here’s Le Chambon. But then in the toponym resolution or disambiguation step, AI does really awfully, according to all the research that just came out this year. Whether it’s Chat GPT or whatever, it just does a really poor job of disambiguating these places. Thus whether it’s as a geographer or historian or an anthropologist or a sociologist or whatever, AI is this tool, one of many. But I think we need to always be good scholars, and need to be checking accuracy, precision, recall, and F1, all of these statistical measures. Not just taking the results of AI at face value. And I think as scholars, we wouldn’t do that anyways, but people get so excited about it sometimes that we forget to do the measurement of results or maybe we do it, but it’s not as visible. We thus need to be very transparent as scholars, if we’re using AI, in our methods, and then how we measured things like precision and recall, how we did the science. And then take everything with a grain of salt and cross check. Now, on the other hand, both AI and traditional natural language processing tools provide the ability to scale up. The testimonies are a fantastic resource that are there in very large numbers, tens of thousands. So what insights can we gain, even given the inaccuracies? With geoparsing, I got it up to 85% precision by training the model. But that still means that 15% isn’t right. So that’s concerning, right? That should concern us. But given all of that, what can we do with 10,000 testimonies instead of 10? What insights can we gain about the experiences of survivors, not just as geographers, but in all the other scholarly fields. And there’s a whole group of scholars around the world that are working on this, in Holocaust studies and many other areas such as literature. Digital Humanities isn’t a brand new thing. It’s been around. But as the tools get better, the ability to have more valuable insights from those tools increase. And so I think as scholars we just have to be open to what those tools can give us, given all of their limitations and errors, and just always be transparent about our methods and measurement of results. And the other thing ethically, even as we do analyze 10,000 testimonies someday, we always have to preserve room for the individuality of those testimonies, because every single one of those testimonies is a person’s experience, and they all had different experiences. There may be commonalities, but we don’t want to give up that individual. We don’t want to do any more violence to that individual than was already done. We must be careful that we remember the individual even in those efforts to look at things at scale.

Learn more about Christopher J. Anderson here.

Read a summary of his lecture here and watch it below.