What We Teach About Race and Gender: Representation in Images and Text of Children’s Books
利用人工智能分析百年间广泛阅读的儿童书籍中的图像和文字，发现黑人和拉丁裔人群及女性在最具影响力的书籍中代表性不足，且角色肤色普遍偏浅，并探讨了供需两侧的经济因素。
Abstract Books shape how children learn about society and norms, in part through representation of different characters. We use computational tools to characterize representation in children’s books widely read in homes, classrooms, and libraries over the past century and describe economic forces that may contribute to these patterns. We introduce new artificial intelligence methods for systematically converting images into data. We apply these tools, alongside text analysis methods, to measure skin color, race, gender, and age in the content of these books, documenting what has changed and what has endured over time. We find underrepresentation of Black and Latinx people in the most influential books, relative to their population shares, though representation of Black individuals increases over time. Females are also increasingly present but appear less often in text than in images, suggesting greater symbolic inclusion in pictures than substantive inclusion in stories. Characters in these influential books have lighter average skin color than in other books, even after conditioning on race, and children are depicted with lighter skin color than adults on average. We present empirical analysis of related economic behavior to better understand the representation we find in these books. On the demand side, we show that people consume books that center their own identities and that the types of children’s books purchased correlate with local political beliefs. On the supply side, we document higher prices for books that center nondominant social identities and fewer copies of these books in libraries that serve predominantly White communities.