Start Date
10-12-2017 12:00 AM
Description
The digitization of commerce has provided both a mechanism by which goods and services are exchanged, as well as an efficient way for consumers to voice concerns, in the form of online rating systems. Yet, recent work has begun to uncover biases that manifest during the feedback process. In this work, we build upon research that has identified race and gender biases in online transactions, and probe the mechanisms by which such biases may manifest. Using an experimental methodology, we find little evidence of bias across race and gender when information about historical quality is available. Moreover, in the presence of a high quality experience, we find no differences in ratings based on race or gender. However, when quality is poor, worse ratings accrue for African Americans and females, notably when they are rated by white men, suggesting that bias may manifest because of errors of attribution.
Recommended Citation
Greenwood, Brad; Adjerid, Idris; and Angst, Corey M., "Race and Gender Bias in Online Ratings: An Origins Story" (2017). ICIS 2017 Proceedings. 5.
https://aisel.aisnet.org/icis2017/IT-and-Social/Presentations/5
Race and Gender Bias in Online Ratings: An Origins Story
The digitization of commerce has provided both a mechanism by which goods and services are exchanged, as well as an efficient way for consumers to voice concerns, in the form of online rating systems. Yet, recent work has begun to uncover biases that manifest during the feedback process. In this work, we build upon research that has identified race and gender biases in online transactions, and probe the mechanisms by which such biases may manifest. Using an experimental methodology, we find little evidence of bias across race and gender when information about historical quality is available. Moreover, in the presence of a high quality experience, we find no differences in ratings based on race or gender. However, when quality is poor, worse ratings accrue for African Americans and females, notably when they are rated by white men, suggesting that bias may manifest because of errors of attribution.