#MindTheAdjective Newsletter (03.04.20)

#MindTheAdjective refers to language used, verbal or written, because our word choices have implications on humans and… technology.

Dr. Augenstein (University of Copenhagen) pointed this out in a study where she examined adjectives used in English-language literature to describe men and women. The study, which sought to understand the effects of word choices on gender perceptions, found differences across gender descriptions: women were typically described by physical appearance (i.e. “sexy”) whereas men were described by behavioral and personal qualities (i.e. “brave).

The tl:dr repercussions of gendered language on machine intelligence: machines use algorithms that understand human language. If human language is biased, the machine’s results will also recreate that original bias - much faster and at scale. One example of where this is particularly harmful is in employee recommendations, as companies increasingly turn to AI to sort through job applications.

Adjectives lead to stereotypes; stereotypes lead to real-world outcomes. Researchers in Germany looked at the effects of the stereotypes around boys and reading comprehension, and found a self-fulfillment of the "boys can't read" stereotype among young boys. Perceiving this stereotype as true, young boys in this study lacked motivation and achieved lower reading levels.

The work of #MindTheAdjective belongs to us all. Parents can teach children healthy and accurate self-perception. HR professionals, managers, and employers can reduce disparities in hiring and performance evaluation processes, and therefore level the playing field for hiring and promotion outcomes. Computer scientists can take special care to use unbiased texts, and develop unbiased algorithms. Educators can remove gender stereotypes from their communications and lesson plans to effect neutral learning.

Regardless of the role(s) we hold, we all communicate, daily. Let’s make our communications eliminate gender bias.