What are our ethical responsibilities as HPE scholars?

Timeline showing the total number of slaves embarked by year from the SlaveVoyages.org dataset

Edited by Emily Sellars from contributions by the Broadstreet editorial team

“As we learn from new data and new methods, it is paramount that we keep the truth of [enslaved people’s] essential humanity at the forefront of our efforts…It is possible, after all, to disturb a grave without ever touching the soil.” -Jamelle Bouie, “We Still Can’t See American Slavery for What It Was

This post grew out of a conversation between Broadstreet co-editors last week about a column by Jamelle Bouie for The New York Times on some of the ethical and moral dimensions of quantifying the scale of the slave trade (the column is linked here and was published in the print version of the Sunday Review on January 30th under the title, “Quantifying the Pain of Slavery”).

For those of us who work on HPE, there is a lot in Bouie’s column to think about: the extent to which quantifying history can be “dehumanizing,” the practical and ethical consequences of relying on data generated by immoral historical processes like the slave trade, how our academic projects or datasets might be misused or misinterpreted, and, most importantly, what our ethical responsibilities are to those we study, to their descendants, and to society as a whole. What follows is a lightly edited version of our conversation, which we hope will be the first of several posts on related topics.

As you’ll see, we found many broad areas of agreement and some areas of disagreement with Bouie (and also within the editorial team). One thing that is clear to me is that these sorts of conversations should be happening more frequently. As scholars who work on historical settings, often using quantitative data, HPE researchers may be shielded from directly confronting some ethical and moral dimensions of social-scientific research. Unlike those who work on political economy in contemporary settings, we cannot usually interact with those we study, much less obtain their consent to be included in our research. The typical methods that we use, like archival research, are not usually viewed as “human subjects research” by university Institutional Review Boards. Especially for those of us who study a time and place that is very different from our own, there may be more physical and emotional distance between us and our research subjects (for better or worse, as we discuss below).

Many of the issues brought up in this post and in Bouie’s column have been considered by others, including the professional organization of archivists and historians to some degree. Though HPE scholars are not usually studying living human subjects, some of the issues raised in APSA’s Principles and Guidance for Human Subjects Research arguably apply to our work as well. Finally, we see this short discussion as the beginning of a longer conversation about these topics on Broadstreet. If you’d like to contribute to this conversation through a guest post, please let us know! You can reach us by email at info@broadstreet.blog  or by getting in touch with one of the editors directly.

Map of the Bight of Benin, 1845. Source: J.J. Lopes de Lima, “Ensaios sobre a Statística das Possessões Portuguezas…” Source: Emory University via SlaveVoyages.org

“Dehumanizing” aspects of quantitative research

Emily Sellars: Bouie’s column brings up several important questions, but a central thread is that something crucial can be lost in the process of quantifying and systematizing historical data into large, searchable datasets. What do we make of the argument that quantification of processes like the Atlantic slave trade can be “dehumanizing”? 

Ali Cirone: I wrote about this a bit in my last postancestry.com has put a Holocaust database completely online (everything from passenger lists to video narratives), and it caused a similar discussion.. Academics in the digital humanities have written about using these. But more broadly, people taking data/infographics/databases out of context, or people not engaging with history, is a threat with all data, though maybe it’s more visible with sensitive data. The Naomi Wolf “death recorded” goof is a good case in point – she was supposedly studying a persecuted population, and shedding light on this group in history, but she got the history completely wrong. But the easier it is to download a dataset, maybe the lower incentive to engage with any details.

Adam Slez: I think it is worth comparing the norms in historical research to the norms in ethnographic research, where it is very common to consider that academics routinely make careers out of data collected from marginalized populations. This raises questions about what our obligations are to the people or groups that we study. I don’t think that qualitative methods are inherently less prone to doing violence, so much as it is that qualitative researchers are forced to confront the prospect of harm on a more regular basis. This is due in large part to the role of IRBs in the research process. As a quantitative social scientist who uses historical materials, much of what I do is traditionally exempt from the perspective of an IRB, which focuses on a very specific kind of harm. This is not to say that there are no ethical concerns associated with the type of work that I do or that an IRB necessarily solves the problem associated with other types of work. Quite the opposite. Any time we study people, there is a risk of doing harm, regardless of method. Bouie’s article challenges us to think about what it means to do harm and how this works in the context of quantitative history.

Tracy Dennison: Bouie’s piece is actually more balanced on this question than many things I’ve heard/read in academic contexts in recent years. The impulse to turn these into “either/or” questions is so frustrating. Of course both quantitative and qualitative research are necessary for understanding the past. And Adam makes a good point about qualitative evidence – the violence is just of a different nature.

Emily: I think that there is something to the argument that it is easier to be aloof/distanced when dealing with large-scale data than with specific narratives or case studies, even if one is very thoughtful about thinking through where the data come from, how information was generated, etc. Something like the transatlantic slave is just more abstract/less personal at the level of millions of people than when thinking about specific families/individuals. This is not to say that qualitative research can’t “dehumanize the reality” in its own way, or that the less abstract approach is inherently better or more ethical. I agree with Tracy that when done well qualitative research and quantitative research are complements not substitutes.

Pavi Suryanarayan: Most studies working with quantitative data will try to square their findings with the qualitative evidence from a particular era. The individual stories or narratives are what inspire us to think harder about how to create useful, scalable  measures to capture the same phenomena on a broader level. It’s what led so many of us to think about how people interact with institutions (like the slave census for instance) to think about how that interaction then shaped development broadly (state, parties, regimes). I think the piece is nuanced but also underestimates how much quant scholars grapple with the “people” in their stories. I also want to point out that the norms of publishing here make some of us deal with this explicitly. Reviewers will pushback on everything from the appropriate language (please say enslaved not slaves), to attribution (Roediger and Du Bois said this first not you!), to measurement.

Adam: This all goes along with the risk of abstracting away from the lived experience of people—a risk that is compounded when using data that is itself the product of an effort to rationalize large organizations created for the purpose of profiting off of dehumanization. As Vicky points out, these types of problems aren’t necessarily unknown to researchers. For example, when I teach graduate statistics now, I close the first lecture with a discussion of how the collection of statistics by the state has historically contributed to the exercise of power, as highlighted by Foucault’s work on governmentality and Bourdieu’s work on classification and symbolic violence. The question is what we should do to address or confront these issues in an ethical way. At the bare minimum, we need to be reflexive about where our numbers come from.

Tracy: It makes sense that these questions arise in the context of the most violent and painful episodes in history, but they are relevant to history more generally. The voices of people who had the most miserable experiences are rarely captured in sources (for all sorts of reasons: they couldn’t write, they lived their lives outside the formal institutions that left records, etc.). There are all kinds of potential representativeness issues in working with these kinds of sources, which tend to be glossed over in criticisms of quantitative studies. When used together, qualitative and quantitative sources can inform one another and help balance out the problems of each. 

Emily: Historical research in general, quant or qual, is also a bit more distanced from the people under study than e.g., contemporary ethnographic research (again, not to say that the latter can’t be misleading/dehumanizing/exploitative). Adam raises a good point that we aren’t always forced to confront these ethical issues as directly in our written work as scholars working on settings that they can directly observe or influence. On the other hand, some of the criticisms of quantitative data as being more abstracted or distanced from the people under study would apply to scholarship on the contemporary era as well, for example large-N studies of political violence.

Vicky: I don’t really think that any method is inherently humanizing or dehumanizing. What matters is who uses it, how well and with what end in mind. I agree it is easier to be distanced from individual human stories as a quantitative researcher, but I don’t believe this is necessarily always bad. Sometimes it is necessary to zoom out of specific stories to get more accurate answers to our questions (and, by the way, these questions often have important implications for people’s lives, so there is an ethical dimension there too; we need to combine methods to get to the truth). And the focus on cases, narratives and stories can also sometimes strengthen the effect of researcher’s biases, as they interpret, report and transform those stories through their own lens of viewing reality. I agree with everyone that qualitative and quantitative research are complements and each has its own challenges, both in finding out the truth and in doing justice to the populations studied.

Emily: These are great points, building on something that Pavi brought up in our chat. There are some ways in which quantitative research may be less exploitative than qualitative research. It may be easier to preserve the confidentiality of specific people, for example, which is an important consideration given that our research subjects can’t usually consent to having their stories shared. Vicky’s point about how cases/narratives/stories strengthening the biases of researchers is a good one, too. An issue we get into below is how research might be subsequently misinterpreted or misused. This is a problem with qualitative as well as quantitative research. It is easy to think of shameful examples of people circulating wildly unrepresentative stories along the lines of, “Look! Here’s an enslaved person who didn’t want to be freed.” or “Look! Here is a narrative of why slavery/colonial rule/mass violence or whatever wasn’t so bad.” One thing that quantitative research can contribute is to illustrate how distorted and misleading those narratives are. 

There is also something to the idea that quantifying the scope of processes like the slave trade makes it difficult to dismiss them as historical blips. This was a huge, centuries-long phenomenon. On a related note, the news this week that the U.S. has now seen more than 900,000 deaths from COVID-19 since early 2020 (900,000!) I think illustrates both the power and limitations of quantification. On one hand, it emphasizes just how staggering the toll of this pandemic has been, an idea that somehow keeps getting lost. On the other hand, this shocking fact is more abstract than thinking about the specific people who we’ve lost.

Copied excerpt from the relación geográfica of the town of Teocaltiche (“Tequaltiche”) from 1585 describing the destruction of the cocoliztli (“cocoliztle”) epidemic. Source: LLILAS Benson Digital Scholarship, UT-Austin

The ethics and limitations of using historical data generated by immoral/coercive practices

Emily: Another key set of ideas in Bouie’s column relates to the moral dimensions of using data that were generated by coercive systems and practices like slavery. As he points out, a lot of what we know about the scope of the slave trade comes from records that were generated through its practice: ship manifests, economic records from plantations, legal documents, etc. The content of these records reflects this bias and shapes both what can be learned from this information and, he argues, potentially how we think about slavery in general. What do we think?

Vicky: Bouie argues convincingly why it’s important to know the context in which the data is collected, what the data contains and whose interests it represents. But from this it doesn’t follow that the researchers who actually use the data are unaware of these limitations or know them and ignore them.

Pavi: Anecdotally, this is exactly what a typical audience asks us to explain. Who collected this? What were their incentives? What does it mean for how much we should trust the data? Whose agenda is this pushing? Can we think through what bias would mean for the interpretations of our findings? 

Ali: Agreed. This is basic research ethics – across any of our disciplines, researchers should not misrepresent (and be transparent). 

Emily: I think it is worth separating two concerns. One is the issue of whether data are being “correctly” interpreted, which might include factors like whether we’ve adequately thought about bias or what the data actually contain. HPE scholars are used to fielding questions about these things. A separate issue is whether reliance on these types of data sources distorts the entire academic conversation about certain subjects. This is something that I discuss in my post on archival silences in HPE research. Even if we are careful in how we interpret archival data generated through processes like the slave trade, certain perspectives are systematically absent. This naturally shifts the sorts of questions that we pursue in a way that is worth reflecting on. It also shifts the tone with which we discuss specific topics. 

This brings up the contested idea of “epistemic violence” developed by Gayatri C. Spivak, which develops how the generation and dissemination of “knowledge” itself can do harm to people. Bouie raises this point in his column, quoting from Marisa J. Fuentes’s Dispossessed Lives: Enslaved Women, Violence, and the Archive. As Fuentes argues and Bouie quotes, “The violence [of slavery] is transferred from the enslaved bodies to the documents that count, condemn, assess, and evoke them….Epistemic violence originates from the knowledge produced about enslaved women by white men and women in this society, and that knowledge is what survives in archival form.” I am frequently asked about general inference/interpretation issues with my data, but seminar audiences in political economy have basically never pushed me on this dimension.

It’s not clear what HPE scholars ought to do about this problem, besides keeping these ideas in mind. The answer is almost certainly not to avoid studying topics like the slave trade or to completely ignore and set aside data sources that were generated through its practice. An important point that Harvard historian Drew Faust brings up in her short response to Bouie’s column is that these documents may also reveal “truths that they did not intend to record” and that illustrate something about the actions, ideas, and agency of those who were enslaved. Edgar Franco-Vivanco wrote a great post for Broadstreet about some of his work analyzing judicial claims by Indigenous communities to illustrate something about “the repertoires of contention” that these communities could use to contest colonial injustices. All this is to say that this is a difficult topic to navigate. I was excited to see that the theme for this year’s Economic History Association meeting is “hidden figures” emphasizing work that incorporates commonly excluded groups in economic history (thanks for the tip, Vicky!).

The first page of the court register of “Liberated Africans” from the Brigantine “Semiramis” in 1835. Source: SlaveVoyages.org and the British National Archives.

On the ethics of making our data available or public, sometimes with minimal context

Vicky: A big part of Bouie’s piece was not so much about data being used in research, but more about data being made public, displayed in newspapers, etc. in ways that could be an affront to people. It seems like a big part of his criticism was not about relying on these data to study slavery, but on making them searchable online, displaying them as posters or infographics, without any context. I would definitely agree that this is a problem and see it as distinct from the qual vs quant question in research. 

Scott Gehlbach: This point seems to relate to one quant vs. qual dimension, which is the scale of the data. Going public with quant data, and especially making the data searchable online, potentially affects so many more people than going public with qual data (though the impact on any particular individual may be smaller, so the net effect may be unclear). 

Emily: There are again a couple of separate concerns here. One might be the privacy/confidentiality issues related to making online databases on the slave trade, for example, searchable to specific individuals. Another is the issue of making data available to those who might not interpret them correctly or who are too distanced from the process through which the data were created to understand the implications.

Ali: Arthur Spirling and I wrote a bit about this in our JHPE paper on “Turning History into Data.” Increased data accessibility might reduce the focus on the context. As we wrote there, “Advances in the availability of digital collections have also made it easier to access and even directly download data, which could very easily reduce engagement with the case and its particulars.” A related point by Lara Putnam (2016) is that information access is “radically diminishing the role of place-specific prior expertise as a prerequisite to discovery.” A lot of empirical mistakes can stem from a lack of context-specific knowledge, as we discuss in our paper.

Emily: This brings up a bigger question of who should be held responsible for the misuse of data: the person or organization who generated the data, those who made it available, or the outside individuals who then misuse or misinterpret what’s presented. I think that at least some responsibility lies with those who generate the data to provide appropriate context and to consider the implications of making the data available. This is especially the case when what is being made public is something of a sensitive nature (like Ali’s example of James Joyce’s NSFW love letters from the chat). There are benefits to getting people interested in a topic through making this information available, but how do we balance this against the interest of those under study, who might prefer not to have their information publicized. Unfortunately, we generally can’t ask those we study about this.

Ali: This ironically relates to a larger debate about technology, social media, and how we consume information. We are now in an age where we are quick to see something, make a conclusion, and retweet our conclusion – all without ever clicking through. Sensitive topics (in history, or otherwise) are also more likely to go viral, and suffer this fate. Any type of content can be interpreted out of context (or deliberately altered to leave the context out). This does not mean we should shy away from creating and sharing any content, but we should recognize the incentives and try to combat them, to get people interested and to engage with the history. That is what these infographics and online databases are trying to do, to share these (often) victim’s stories. So if we think these resources should exist, then should there be minimum standards for presenting them? 

Adam: We don’t tend to see our work as being tied to public projects of preservations and/or commemoration the way some folks in history or anthropology do, which means that we aren’t forced to think as much about reception or the importance of constantly confronting the significance of human agency.

Emily: Yes, scholars in the digital humanities especially have some experience navigating these sorts of questions, and there is a lot that we can probably learn from them on this. At minimum, I think, we should be clear about what our data represent and make it as easy as possible for those who use these resources to interpret them appropriately by providing additional context and information when necessary. The difficult thing is that we can’t fully control how our work will be used or interpreted by others. Bouie has a nice discussion about this, related to the popular infographic that he and his colleagues at Slate put together using the Slave Voyages dataset. It is not always easy to foresee how others might view or adapt your work. 

Bouie’s discussion highlights some of the benefits of digitizing and publicizing data on the slave trade, too. I liked the example of how the “Oceans of Kinfolk” database can be used to connect enslaved people with their descendants. This highlights the question of access. It might be easier to curate or manage how others respond to our scholarship when we keep it behind paywalls or in academic repositories, but this keeps information from those who lack an institutional affiliation or technical training, some of whom may have a close personal connection to the question under study. What are our responsibilities to them?

Moving forward

It is difficult to close this conversation as there are no clear answers to these questions and a lot more than could be said about all of these issues. We are thus keeping the conversation open and hope to develop these ideas further in future posts. Please let us know if you’d like to write one. Thanks to editors Ali Cirone, Tracy Dennison, Vicky Fouka, Scott Gehlbach, Adam Slez, and Pavi Suryanarayan for contributing and to editor-in-chief Jeff Jenkins for getting this discussion started. 

Header image: Screenshot of the Timeline of the Total Number of Slaves Embarked by Year from SlaveVoyages.org


  • Sellars is an Assistant Professor in Political Science at Yale University, where she is affiliated with the MacMillan Center for International and Area Studies, the Leitner Program in Political Economy, and the Institution for Social and Policy Studies. Her research interests are at the intersection of comparative political economy, development economics, and economic history. She is particularly interested in the politics of emigration and in the historical political economy in Mexico and Central America. Her work has been published or is forthcoming in the Journal of Politics, the Journal of Development Economics, the Journal of Urban Economics, and the Journal of Political Institutions and Political Economy. Sellars received her Ph.D. jointly in Political Science and Agricultural and Applied Economics from University of Wisconsin-Madison.

One thought on “What are our ethical responsibilities as HPE scholars?

Leave a Reply