Campus Only Senior Thesis
Bachelor of Arts
© 2020 Molly R Ferguson
This paper constructs the retailfashion data set from the top grossing online retail fashion parent companies in the United States in 2018. Open source facial attribute recognition python libraries face-recognition and face-classification are used to detect facial encodings in the collected images, and to assign gender and racial labels to each facial encoding. Analyzing the labeled portion of the data set, this paper demonstrates that brand web pages publish a disproportionate number of images of White models, while under- representing Black and Asian models, and hire disproportionate numbers of Female models to Male models. The brands and parent companies included in the retailfashion data set all exhibit either racially biased composition or model function and contribution, and in many cases show both biases clearly. This paper answers the question of whether White models and non-White models are portrayed differently on major online fashion retail websites in the United States, and compares biases across brands for trends explaining these problems within the industry.
Ferguson, Molly, "Using Facial Attribute Recognition Libraries to Examine Gender and Racial Bias in Top-grossing Online Fashion Retailers" (2020). Scripps Senior Theses. 1551.
This thesis is restricted to the Claremont Colleges current faculty, students, and staff.