Why understanding and responding to search engine bias matters to language educators

Abstract

1 INTRODUCTION

Computer science research has increasingly documented social group bias in artificial intelligence (AI). Examples include bias against African Americans in software used by courts to determine bail and sentencing decisions (Angwin et al., 2016), facial recognition systems performing better on people with lighter skin (Buolamwini & Gebru, 2018), and a hiring algorithm penalizing graduates of women's colleges (Silberg & Manyika, 2019). Sources for AI bias are complex. They include bias in the initial data used by the AI along with the role of AI algorithms themselves in “amplifying” such initial biases (Ntoutsi et al., 2020). Effects of AI bias are commonly seen in search engine results that an AI application many use on a daily basis (Noble, 2018).