Applications of Graph Theory in Understanding Gender Bias
In the modern-day workforce, gender inequality has been a prevalent and unfortunate reality. Reports outline that women only comprise 28% of science, technology, engineering, and math-related positions, ranging from a slight minority of 46% within biological sciences to an extreme 16.5% within the field of engineering and architecture [1]. While there exists a clear gender gap at the professional level, data regarding earlier levels of education is not as clear.
A recent 2022 article published by researchers from the University of Milan investigated whether or not gender inequality had roots within the secondary education / high school level. Researchers ran a network study on ten co-ed high school classes in an Italian high school, testing whether perceptions of classmates’ reading, math, and science performances were affected by gender biases. Since the Italian school system assigns students to a classroom for a full school course of five years and allows for students to have access to each other’s grades, students are already quite familiar with their classmates’ academic abilities. Students were asked to nominate their top four candidates for the respective three academic categories. Researchers hypothesized that students would follow gender biases promoting more female students to be nominated for the reading category than their male counterparts but vice versa for the science and math categories [2].
To assess their hypothesis, the researchers created a 195-node graph with nodes representing students and edges corresponding to a nomination that a student gave to another. Then, a directed Exponential Random Graph Model (ERGM), a type of model that allows for the probability that a pair of nodes has an edge between them to be modeled using graph theory, was created and analyzed. By using an ERGM, the researchers could calculate a parameter whose coefficient “indicated the likelihood that a female rather than a male student was nominated,” and could implement control parameters such as friendship bias and actual academic ability which could limit potential erroneous causal relationships [2]. The so-called “gender receiver” parameter interestingly showed that there were no statistically significant results within the math category, but supported the expected hypothesis that female students were more likely to be nominated for the reading category but less likely to be nominated for the science category [2].
Within the framework of INFO 2040, graph theory was highlighted as a tool to create mathematical models of various network structures such as communication, social, or information networks. In this study, researchers were able to represent a population of 195 high school students using a more complex graph theory model which revealed clear statistical data to further our understanding of the effects of gender bias. By taking the simple nodes and edges foundation of graph theory and amplifying its effects, we have the ability to both study and model the behaviors of a large population, allowing us to broaden our understanding of the prevailing issues in the world.
[1] https://www.aauw.org/resources/research/the-stem-gap/
[2] De Gioannis, E., Bianchi, F., & Squazzoni, F. (2022). Gender bias in the classroom: A network study on self and peer ability attribution. Social Networks, 72, 44–51. https://doi.org/10.1016/j.socnet.2022.09.001