Computational appraisal of gender representativeness in popular movies

Supplemental material of the research published in HSSCOMMS (2021) by Antoine Mazieres, Telmo Menezes and Camille Roth.
More info on https://mazieres.gitlab.io/gender-movies

ABSTRACT

Gender representation in mass media has long been mainly studied by qualitatively analyzing content. This article illustrates how automated computational methods may be used in this context to scale up such empirical observations and increase their resolution and significance. We specifically apply a face and gender detection algorithm on a broad set of popular movies spanning more than three decades to carry out a large-scale appraisal of the on-screen presence of women and men. Beyond the confirmation of a strong under-representation of women, we exhibit a clear temporal trend towards a fairer representativeness. We further contrast our findings with respect to movie genre, budget, and various audience-related features such as movie gross and user ratings. We lastly propose a fine description of significant asymmetries in the mise-en-scène and mise-en-cadre of characters in relation to their gender and the spatial composition of a given frame.

DATA

- facialfeatures.csv
Raw inferences from the face and gender detection models.
- Movie's IMDb ID
- Timestamp of the frame
- Face's gender (0 is female, 1 is male)
- Face's bounding box coordinates : xmin, ymin, xmax, ymax

- metadata.csv
Movies metadata.

- human_evaluation.csv
Results from the human evaluation of the detection models.

- model_correction.csv
FFR_corrected = a + b * FFR_uncorrected