The call for more diversity in children’s literature began decades ago. Educators know that representation, among authors and within the pages of each book, matters. In recent years, the landscape is changing with more diverse and #ownvoices books published, and those titles receiving recognition with some of children’s literature’s biggest awards. Recent Newbery winners such as Tae Keller's When You Trap a Tiger, Jerry Craft's New Kid, Meg Medina's Merci Suarez Changes Gears, and Erin Entrada Kelly's Hello, Universe proves it, right?

A recent study, however, shows any progress has been limited.

Researchers looked at the skin colors in Newbery and Caldecott winners (Mainstream) vs. the winners of awards based on identity (Diversity), such as the Coretta Scott King Awards. (See box below for full list.)

“In very broad sum, despite growing awareness in recent decades, children’s books generally skew toward lighter skin and male representation,” concluded the authors of “What we teach about race and gender: Representation in images and text of children’s books,” an analysis of more than 1,100 award-winning children’s books by the Becker Friedman Institute at the University of Chicago.

Historically, representation has been difficult to measure at scale. Conventional efforts to systematically measure representation in books have relied on time-consuming content analysis techniques. These methods require researchers to examine each page by hand to get an in-depth look at the characters and stories. By necessity, studies done that way often look at only a small selection of books because this process requires so much time and resources. The five authors of the study—Anjali Adukia, assistant professor at the University of Chicago’s Harris School of Public Policy; Alex Eble, assistant professor at Teachers College Columbia University; Emileigh Harrison, Ph.D. student at University of Chicago’s Harris School of Public Policy; Hakizumwami Birali Runesha, director of the University of Chicago’s Research Computing Center; and Teodora Szasz, University of Chicago computational scientist—developed and applied innovative artificial intelligence technologies to examine the racial and gender makeup of characters in the images and text of each book. The AI tools detected characters in photos and illustrations and classified their race, gender, and age.

“While AI tools also reflect bias in their training data and algorithms, they can be more replicable, can be standardized, and can be applied to a much larger sample than manual content analysis,” the authors said in the paper.

They analyzed the images and text in the books using different computer models. For images, they created a model that could detect faces in illustrations and then classified skin color and, separately, predicted the race, gender, and age of the faces. They also used Natural Language Processing tools to count the number of gendered words and the race, birthplace, and gender of characters.

Adukia noted that the binarization of gender in the study and said, “Future work really should account for the inclusion of nonbinary or genderqueer, or gender fluid identities.”

The researchers chose books that have won prestigious awards, creating a “Mainstream” collection of Newbery and Caldecott winners of the years and comparing that to a “Diversity” collection of books honored by awards that “highlight experiences of specific identity groups,” such as the Coretta Scott King and Stonewall awards.

“They're highly influential,” said Adukia, whose mother was a school librarian. “They're very likely to show up in school libraries, classrooms, and on people's bookshelves.”

In total, the 1,133 books from 19 different award categories were studied. It included more than 160,000 pages of content published over the last 100 years.

The analysis of images revealed the following about race in children’s books:

•  Books in the Mainstream collection are more likely to depict lighter-skinned characters than those in the Diversity collection.  Specifically, books in the Mainstream collection are much more likely to depict characters who are racially ambiguous in terms of skin color, disproportionately using skin colors that cannot be classified either as that of light-skinned characters nor as that of dark-skinned characters, a technique the authors call “butterscotching.”
• Over the last two decades, representation of lighter skin tones in Mainstream books has actually increased.
• Children are more likely than adults to be shown with lighter skin, in both Mainstream and Diversity collections.
• Females have always appeared in pictures over time, but they are predominantly white females and still average less than 50 percent of pictures and text.

The authors also compared the female appearances in images to female mentions in text and found:

• Females are more consistently seen in images than written about in the text, except in the collection of books specifically selected to highlight females. The authors deduced that this suggests “symbolic inclusion of females in pictures without their substantive inclusion in the actual story.”
• Males, especially white males, are persistently more likely to be represented by every measure, with little change over time.