“Twitter is an ok place to be a little annoyed, right?” asked historian of slavery Elise A. Mitchell in January. She then reported in a thread that Grammarly, the freemium writing-assistant app very familiar to consumers of YouTube videos, had started to suggest changes in language around slavery. If a person were to type the word slave, Grammarly would now suggest the replacement enslaved person. The phrases fugitive slave and runaway slave now triggered the suggested replacement freedom seeker. If you were to write slaveowner or master (in places where the context indicates you’re talking about slavery), the app would suggest enslaver instead. The use of master or slave in nonslavery contexts (as sometimes occurs in engineering, or when describing that one big, fancy bedroom in a house) provokes a suggestion to consider an alternate. “I’m wary (and weary) of who gets to be an authority over the language we use to talk about our past,” Mitchell wrote.
These suggestions appeared in the app as of Jan. 18, Grammarly spokesperson Senka Hadzimuratovic confirmed to me in an email. Asked how Grammarly decided to add “suggestions” about slavery, Hadzimuratovic wrote: “We prioritized these particular suggestions based on a combination of the curiosity of our users for these topics as ascertained through user research, the prevalence of the language in question in writing, and the potential hurt the language can cause.” (Hadzimuratovic added that the company included “subject matter experts on race and ethnicity” in its research process; I asked to speak to one, and she declined, citing nondisclosure agreements signed when the experts agreed to work with the company.)
Hadzimuratovic also sent me a list of 11 resources the company had used when researching these changes: guides produced by organizations like the Underground Railroad Education Center, the Southern Poverty Law Center’s Teaching Tolerance project, and the NAACP. The suggestions Grammarly provided are, so to speak, in the air—you can find support for every one of them in these resources. So why did the many historians of slavery who replied to Mitchell’s thread agree that the company had misstepped?
This is what happens when experts in a complicated subject matter see their work translated into normative prescriptions. First, on freedom seeker. As Mitchell pointed out in her tweets, “Not everyone who fled was seeking ‘freedom,’ ” and “running away is not the only way to ‘seek freedom.’ ” There is a growing body of academic work on how some enslaved people in the United States who could not run away altogether for whatever reason stayed and “sought freedom” in other ways: negotiation, temporary absences, smaller (and less permanent and risky) acts of resistance. (Historian Daina Ramey Berry wrote about enslaved people’s negotiations around freedom here, in an essay referring to, and synthesizing, some of this work.)
Enslaver as a substitute for slave owner doesn’t quite work, either. Sometimes enslavers didn’t themselves own people. Scholar Nicholas Rinehart wrote a piece about slavery and language for the online journal Commonplace, which he linked in a reply to Mitchell. In it, he has a list of the different roles that could fall into the bucket enslaver:
Merchants who purchased and sold slaves; ship captains and crew who held them captive; auctioneers who ran slave markets; slave traders who acted on behalf of wealthy planters; overseers on plantations throughout the Americas; managers who ran plantations for absentee estate holders; banks and firms that financed those plantations; institutions and other corporate bodies that owned slaves; and more.
With so many specific roles falling into the enslaver category, an automatic correction in this language—whether executed by habit or by software—might make everything about the history you’re describing less precise. And since one of the major lessons historians of slavery try to emphasize is that the institution could be found woven into every part of the economy and society, the loss of specificity could be significant.
Additional issues arise when you consider the word enslaver in a global context. In replies to Mitchell’s thread, historians specializing in slavery in Africa, indigenous North America, and South America added that enslavers and slave owners were two different groups in these contexts. “People were enslaved so that they could be sold, not owned” in Africa during the time of the slave trade, tweeted Rebecca Shumway. “To conflate [enslavers and slaveowners] would make the history unintelligible.” “That distinction is critical to understanding indigenous North American forms of slavery, too,” added Brett Rushforth. “One can’t understand the Jesuits in Brazil without it,” Andrew Dial said.
The substitution of the construction enslaved person for slave is the language change in this set that’s probably the most widely practiced, and most natural to a lot of people, me included. As a group of scholars explains in a crowdsourced Google document titled “Writing About ‘Slavery’? This Might Help,” the choice of enslaved person separates “the condition of being enslaved with the status of ‘being’ a slave.” Since I made the switch in my own writing and speech, I have not been able to encounter the word slave without feeling a small jolt of wrong-ness. But the consensus on that one is not total, either. When my former colleague Jamelle Bouie and I put together our History of American Slavery podcast back in 2015, our former colleague Katy Waldman wrote about the question of enslaved person vs. slave in a piece for the website. Historian Eric Foner, answering her question, explained why he had not discarded the word slave:
I do not think that slave suggests that this is the essence of a person’s being. … It is a condition in which people find themselves and that severely limits their opportunities and options, but it does not mean, as some claim, that the word means they are nothing but slaves. Slaves are human beings and can be husbands, wives (in fact if not in law), fathers and mothers, members of religious groups, skilled craftsmen. … All people have multiple identities, including slaves.
In his article about slavery and language, Rinehart argues that this substitution doesn’t always do what we think it does. The idea of moving to enslaved person is to reverse the process by which the institution of slavery supposedly, as Rinehart terms it, “reduced persons to nonpersons.” But this references a historical process that is, in scholarly terms, contested. In places outside of the United States, where other laws prevailed, enslavement did not necessarily render someone a legal nonperson; even in the United States, scholars have begun to identify instances in which enslaved people did assert legal personhood. What’s more, the assumption that slaves were seen as nonpersons, some have argued, lets the people who perpetrated slavery off the hook by allowing them the fiction of historical ignorance. “Rather than seeking to extinguish the humanity of its victims, slavery invests in, and relies upon, their human capacity for suffering,” Rinehart wrote. The cruelty, in other words, was the point.
I called Rinehart, who is currently a postdoc in the department of English and creative writing at Dartmouth, after the conversation about Grammarly unfolded online and asked him how he explained all of these fine points of language to students—one key audience for Grammarly’s services. He told me, “One thing I want to caution readers and students about in general is this idea that there are good words and bad words. Language is really too context-specific, too complicated, too variable over time” for specific recommendations to be made. “We can adjust our language in any number of situations, but it doesn’t necessarily solve a problem, so much as it introduces a new set of problems.”
While I imagined how frustrating his students might find this, I also know what he means. The problem is that the history of slavery is huge: It spans centuries, continents, and contexts. And the need to be sensitive—not to multiply harms by reusing the language that enslavers favored, for example—is acute. For a person writing about slavery without a deep level of knowledge, there really isn’t any substitute for the investment of time and attention.
As a reviewer of Grammarly Premium, the version of the app that includes the new suggestions around slavery, put it in Fast Company in 2021, the software, in theory, “has the skill set of an actual editor,” checking for tone, clarity, and word choice. But of course, actual flesh-and-blood editors misstep in this matter all the time. That crowdsourced Google doc I mentioned above points to several instances where the media has messed up in recent years. When the New York Times called civil rights activist Julian Bond’s great-grandmother a “slave mistress” in his obituary in 2015—not that long ago!—the choice was so obviously wrong that the paper’s public editor acknowledged it in a column. “No simple phrase can capture the whole of Jane Bond’s humanity,” historian Martha S. Jones wrote in her response to the situation. “To tell her story, however briefly, requires honoring the entirety of her life. No shorthand phrase, like slave mistress, can suffice.” Yet the mistake had been made.
In January, Grammarly responded to feedback like Mitchell’s by pulling back on at least one suggestion—the one flagging fugitive slave and suggesting freedom seeker instead. On Jan. 21, the app began flagging only the word slave in that phrase and suggesting a replacement accordingly. Grammarly let me test out the premium version of the software while writing this story. And so, as I write, a little tab with a G on it is hovering to the side, ready to flag my infelicities of expression. It’s following me around my Mac, appearing whenever I make words in Word, Slack, Twitter, TextEdit, or Gmail.
The interface uses a group of emojis to note when things are going wrong, and when I write in the word slave, here in my Word document, the little yellow face with the mouth that’s a line appears. The effect is unsettling, and a little alarming—a friendly robot, inserting itself where it cannot belong. “Maybe I’m wrong,” Mitchell ended her thread, “but I just don’t think we can automate this.”