Andrew Piper is Professor and William Dawson Scholar in the Department of Languages, Literatures, and Cultures at McGill University. He is the director of .txtLAB, where he uses data science to better understand literature and culture, and the author most recently of _Enumerations: Data and Literary Study_ (Chicago 2018). His work has been profiled in The Guardian, The Atlantic, The New Republic, Le Devoir, Real Simple, and on the CBC.
Andrew Piper will join us at Tech Forum 2020 to deliver a talk called Why are women always talking to men in novels? A conversation about debiasing books using data analytics.
What would it mean to "debias" a book? Is this something authors, editors, and publishers want to be doing? And whose job would it be anyway?
Bias is a notoriously tricky concept. I'm going to try to explain it here in a simple and straightforward way. By bias, I mean anything about books that is predictable in a way you perceive to be negative. Notice how this makes bias something that is both measurable — it keeps happening — and subjective. It depends on how you perceive it. When it comes to predictable behaviour, bias is easy to find. Agreeing on what it means, and what to do about it, isn’t quite as easy.
Here's an example: In recent research coming out of our lab, we looked at more than 26,000 characters in contemporary novels using machine learning and found that male characters considerably outnumber female ones, by a ratio of about 60:40 (a surprisingly constant ratio across a whole range of domains). Thus, we would say there is a bias towards women in novels. When authors are inventing imaginary worlds, when editors are selecting manuscripts, and when publishers are promoting books, for whatever reason they consistently reproduce this bias across all kinds of books (from prizewinners to YA).
This is just one example. Bias can obviously take a number of different forms. It can creep into the language authors use to portray characters or the dialogue they construct for them. It can occur around setting or types of plots or subject matter or the authors we choose to promote. Sometimes the two are connected, so that certain types of authors get shoehorned into telling certain types of stories.
But, you might ask, what if readers and writers simply prefer books this way? Maybe some readers really want tales of men convening together (on ships or in battle) or, as in the title of my talk in March, men and women talking to each other all the time. For sure, some readers definitely do prefer these kinds of stories, just as some prefer to see books that focus mostly on the stories of women, as in the recent success Mrs. Everything by Jennifer Weiner.
The problem starts when most books do these things and do them consistently. That's when we may inadvertently be reproducing inequality and stereotypes on the page that might run counter to what we believe in. Most of all, we are unintentionally limiting the creativity and novelty of our storytellers. Debiasing is really about unlocking the power of new stories.
So what can you do about bias? Here are a few short tips that I'll elaborate on in my talk:
Have a conversation. Bias is a complicated topic and will mean different things to different people at your organization. Arrive at a shared understanding of what you mean by bias and also what your goals might be to reduce it. What predictable things do you want to change about your lists and how can you get there?
Get the right tools. A lot of bias is driven by unconscious behaviour. And when it comes to keeping an eye on a lot of books and the complex aspects of those books, even our best intuitions need a reality check. Data can be very helpful in testing our intuitions and seeing if there are things we are missing. The new world of data science has a lot to offer for understanding the contents of your books. My advice here is to get some help.
Aim for unpredictability. Sometimes the goal of addressing bias can feel like a checklist. “Did I do x,y,z?” I prefer to frame it as a quest for producing more surprising outcomes. When you address bias, what you are really doing is making your lists less predictable. This can open up new audiences, markets, and inspiring stories. Don't think of it as a way of limiting your choices but the opposite: It's a way to break out of old habits and make some noise!
Andrew Piper will be talking more about bias in books at Tech Forum on March 25, 2020 in Toronto. You can find more details about the conference here, or sign up for the mailing list to get all of the conference updates.
An EDItEUR presentation on the European Deforestation Regulation.