Slave Sales on Twitter

In his 1975 critique of Time on the Cross, historian Herbert Gutman noted that,

about two million slaves (men, women, and children) were sold in local, interstate, and interregional markets between 1820 and 1860, and that of this number perhaps as many as 260,000 were married men and women and another 186,000 were children under the age of thirteen. If we assume that slave sales did not occur on Sundays and holidays and that such selling went on for ten hours on working days, a slave was sold on average every 3.6 minutes between 1820 and 1860.

Every 3.6 minutes. I first encountered that figure not in Gutman’s book, but in Steven Deyle’s Carry Me Back: The Domestic Slave Trade in American Life, which argued in a lengthy appendix that 2 million remains a good estimate of the number of slave sales in the antebellum American South. But Deyle added, chillingly, that “it is quite possible that the average frequency of sale was even greater.”¹

A few weeks ago, in my class on the Civil War era, I was trying to think of a way to communicate these enormities to my students. I was preparing a lecture about some of the ways that enslaved people in the antebellum South used their laboring power and their family units to find points of leverage in their ongoing conflicts with slaveholders. Drawing on scholarship by Steven Hahn, Stephanie Camp, Dylan Penningroth, and others, I wanted to help students understand what some white Northerners found so surprising when they encountered slavery up close before and during the Civil War. Some enslaved people seemed to have property that they considered theirs—houses with keyed doors; garden patches and livestock they controlled; money they earned by selling goods and services. These customary privileges, though spread unevenly across the South and never recognized by the law, represented hard-won but not insignificant victories in the never-ending war between the enslaved and their enslavers.

Yet I also wanted students to realize how fragile these privileges were—a fragility born of the ever-present danger of being sold. I decided that while lecturing, I would open a small terminal window on the screen at the front of the class and run a shell script that looked something like this:

#!/bin/bash

COUNT=20
while [[ $COUNT -gt 0 ]]; do
    echo "A slave was just sold."
    sleep 180
    let COUNT=COUNT-1
done

Every three minutes, that code printed A slave was just sold to the screen—a subtle reminder of the Damoclean sword that hung over every cornpatch, Sunday wage, or home that an enslaved man or woman had won.

Reflecting on that experiment after the class, I started to wonder about a similar experiment using social media. In a previous collaboration with my digital history students, I had built a “Twitter bot” called TexasRunawayAds, which automatically tweets an excerpt from a runaway slave advertisement about twice a day. By slightly modifying the script for that bot, could I create a new Twitter bot to emulate the shell script above?

The result of those musings is Every3Minutes, which tweets a reminder of an antebellum slave sale once every three minutes.

Why Twitter?

My decision to create this bot deserves some explanation, especially since one worry I have about the experiment is the risk of trivializing the history that it recalls. I also wondered if fellow historians would raise an eyebrow about such a use of Twitter. Scholars may rightly be skeptical about whether a serious scholarly argument can be made in 140 characters or less, especially since it took both Gutman and Deyle many more words—and much more work—to come up with the estimate of a slave sale taking place every 3.6 minutes.²

Those concerns are fair. The Twitter bot is not making a new historical argument so much as it is publicizing an old historical argument in a new forum. My thinking here is similar to that of Roy Rosenzweig in his 2006 article on historians and Wikipedia. Rosenzweig’s appeal to historians to take Wikipedia seriously had implications for new media beyond that platform, because it was premised on much broader rationales. “One reason professional historians need to pay attention to Wikipedia is because our students do,” he wrote, and the same can now be said for Twitter. And “if historians believe that what is available free on the Web is low quality,” Rosenzweig added, “then we have a responsibility to make better information sources available online.”

There are challenges that come with any attempt to meet that responsibility. Not least is the fact that, as Rosenzweig explained, historians may find themselves subject to conventions of citation and conversation quite different from those in the academy. That’s true for Twitter, as well. But whether historians like it or not, conversation about history, and specifically the history of slavery, is already happening on social media. Several times in recent months, Ta-Nehisi Coates and Jeet Heer, who both write and tweet for large online audiences, have asked questions about the history of slavery aimed directly at academic historians. Not to mention the wide-ranging conversations that took place on Twitter after The Economist published an inflammatory review of Edward Baptist’s new book.

To participate helpfully in such conversations requires not only that historians be on Twitter, though that is an important first step, but also that we answer in ways sensitive to the particular kind of discourse that occurs there. Critics of The Economist, for example, used a clever hashtag to critique its review and to amplify individual critiques. When Twitter skeptics worry that 140 characters may not be enough to make a complex point, they misunderstand the ways that hashtags or numbered essays can track a conversation unfolding over multiple 140-character tweets. Similarly, while it is not easy for a tweeter to footnote a historian like Gutman or Deyle, it could be easy—and arguably more effective, for a tweeter to @ mention a feed like Every3Minutes and thus direct readers to Gutman in a way that is internal to, rather than outside of, the social media environment.

In short, although this Twitter bot differs from the usual sorts of history found on Twitter, my main rationales for making it were similar to those of other tweeting historians. My largest ambition is to help combat, in a new medium, the same myths about slavery that concerned Gutman in 1975 and that still persist today.

The Messages in the Medium

The process of making the bot has also suggested new rationales and implications, however. For conveying Gutman’s message in this way puts questions to the audience that may be unique to the medium. First, Twitter is by its nature social media; its currencies are “followers” and “favorites.” But this bot shifts attention from those metrics to the raw number of tweets it emits, a number that will increase rapidly and serve, I hope, as an arresting reminder of slavery’s magnitude.

Of course, the bot does not shift attention entirely away from the follow-or-favorite economy of social media. Indeed, I suspect it will make those who encounter the feed acutely conscious of dilemmas that an original reader of Gutman might not have faced: Should I follow a feed about slave sales? Am I annoyed that these reminders come so often? What would it mean not to follow or, once having followed, to unfollow or mute the tweets? Do those deliberate or implicit acts of silencing resemble, in microcosm, our nation’s larger inability to come to terms with slavery’s history despite evidence of its continued relevance all around us?

Even if those questions never occur to anyone but me, writing the code for the Twitter bot also raised different questions that I expect to wrestle with for some time.

For example, I learned at the very beginning that Twitter’s API prevented me from realizing my original intention, which was to tweet the same line of text every three minutes. Duplicate tweets are not allowed by Twitter’s code. So the medium itself forced me to generate, randomly, a wider range of strings. I decided to do this partly by varying the order of the phrases in each tweet and the delimiters between them. But then I began to vary combinations of words using Python lists like these:

people   = ['a slave', 'a person', 'an enslaved person', 'someone', 'a black person', 'an African American']
roles    = ['child', 'parent', 'grandparent', 'grandchild', 'friend']
verbs    = ['sold', 'bought', 'purchased', 'traded']
people   = people + [p + r for p in [p + '\'s ' for p in people] for r in roles]

Such variables forced me to attend to “an enslaved person” as someone bearing multiple relationships to other persons. The code also soon involved me, unwittingly, in a troubling objectification of the human individuals whose stories I was attempting to conjure. (Python is, after all, an object-oriented language.)

These were the sorts of discomforting reflections I hoped my digital history students might have while working on an assignment involving JSON and runaway slave ads. But though I designed that very assignment, the ways in which I found myself “encoding” people still crept up on me in the writing of my code. For me, the project has already become more than a way to change history on Twitter. It is also a way to think through how putting history in code shapes or changes history.³ What’s more, it has prompted me to think harder about what Twitter bots are or could be for, particularly in the realm of digital history. As Mark Sample is reported to have said in a recent talk, “coding Twitterbots provides a way, through making, coding, and tinkering, of exploring what could be (but isn’t) and what shouldn’t be (but is).” Yet what does that mean for a Twitterbot-coding historian?

People should not have been chattel (but were). They can be tweeted (but, until recently, weren’t). Yet should they be comprehended by lists?

Notes