Data Science and Information Visualization for the Detection of Gender Bias

2023年2月3日更新

Project Summary

Principal Investigator	ITOH Takayuki (Faculty of Core Research Natural Science Division)
Co-investigators	NAGASAWA Natsuko （Faculty of Core Research Natural Science Division）
Research period	FY 2022 -
SDGs Goals	5（Gender Equality）

Outline of Research

Numerical methods based on data science are essential for detecting gender bias. Additionally, information visualization is helpful for the qualitative discovery of gender bias; it is a useful tool for qualitatively discovering important trends and problems hidden in various data of professional work and daily life through visual representation.
This study develops a method for identifying gender bias using data science and information visualization. Specifically, we introduce visualization methods to compare statistical survey results of male and female participants along various attributes, thereby facilitating the discovery of gender difference latent in specific attributes in diverse data. For example, in 2022, open data on warmth to air conditioning showed localized gender differences depending on various attributes such as building type, season, clothing, and body type.

Research Reports

Gender bias is latent in a wide variety of data derived from daily life and corporate operations, and its discovery and interpretation are not always easy. This study aimed to elucidate the latent gender bias in data through information visualization. In this research case study, at first, we worked on the visualization of bias in the results of movie recommendations. Our results revealed that women belonging to a minority group in the original data, showed a larger bias in the movie recommendation results. Second, we visualized gender differences in air-conditioning temperature perception and observed that a complex combination of factors such as season, constitution, clothing, and building and air-conditioning types caused differences in temperature perception between men and women.

REPORTS
- Reports（FY2023）
- Reports（FY2022）