Racial Bias in Customer Service: Evidence from Twitter

Abstract

Detecting and reporting systemic racial bias is an essential step toward the eradication of racial discrimination in our society. Doing so not only requires society members to voice and share their anecdotal experiences, but also relies on researchers to document systematic statistical evidence of racial bias. This paper documents the first large-scale evidence of business-to-customer racial bias on digital platforms on which the perpetrators are individual employees who act on behalf of a company and the victims are customers. This is in contrast to existing studies of racial bias on digital platforms that focus on peer-to-peer marketplaces in which both the perpetrators and the victims are individuals acting independently and on their own behalf. By analyzing more than 57,000 social media customer complaints to U.S. airlines and leveraging a variety of analytics techniques, including text mining and facial recognition, we present quantitative evidence that African American customers are less likely to receive a response when they complain than otherwise similar White customers. Furthermore, our deep learning–based falsification test shows that the bias is absent without the race-revealing visual cue. This study offers a practical yet powerful recommendation for companies: conceal all customer profile pictures from their employees while delivering social media customer service.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 43 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.