Signaled or Suppressed? How Gender Informs Women’s Undergraduate Applications in Biology and Engineering

Abstract

How does gender inform initial academic commitments and narrative self-presentation in science, technology, engineering, and mathematics (STEM) fields during the college application process? Analyzing 60,000 undergraduate applications to the University of California, the authors surface two key findings. First, extant gender segregation of academic disciplines also manifests in intended major choice. Additionally, gender and SAT Math scores together strongly predict intent to major in biology and engineering, the most popular and gender-segregated majors. Second, using natural language processing, the investigators find that author gender is more predictive of essay topics written by prospective engineers than prospective biologists. Specifically, women intending to major in engineering write about essay topics that signal their gender identity to a greater degree than women intending to major in biology, perhaps to mitigate gender-transgressive academic commitments. The authors subsequently argue that prescriptive and proscriptive ideas about men and women’s academic choices remain highly salient in a moment of imagining future academic and professional selves.

The BTS Effect: How K-Pop Fandoms Reshape Sociopolitical Messaging on Twitter

“…authorship characteristics) and how topic prevalence varies along a given dimension (Roberts et al, 2014) in a linear regression framework. This allows STM to be used for highly controlled causal analyses as well as for more predictive and exploratory analyses that do not depend on as much specification (Egami et al, 2022;Giebel et al, 2022). Our use of STM is more aligned with the latter because we do not have access to a full battery of controls about the tweet authors, though we do have a discontinuity (before and after the BTS tweet) and temporal data that we exploit.…”

Large Language Models, Social Demography, and Hegemony: Comparing Authorship in Human and Synthetic Text

“…For each application essay, we have a variety of metadata reflecting sociodemographic attributes and contexts of the applicant. Following previous work, the sociodemographic attributes we focus on include "first-gen" status (whether or not the applicant has as least one parent who completed a college degree) and gender (binarized as Male or Female) Alvero et al (2020); Giebel et al (2022); Lee et al (2023). These data are important given longstanding barriers for women entering into engineering and the underrepresentation of lower income students at selective universities.…”

“…But it was also the case that women were at times more similar to the AI than men. It is worth noting that although women are not typically associated with processes of hegemony, they tend to submit stronger overall applications than men Giebel et al (2022), pointing to ways that context could shape answers to the question of which students are writing most like LLMs. This, paired with the low variation in these same features for the AI, show how standardization and homogenization is likely but also likely to become associated in the writing styles of certain groups of people in hyperfocused ways.…”

Sociolinguistic Perspectives on Machine Learning with Text Data

“…The category is often used in sociology as a racial category condensed into a single binary variable, similar to other so-called Census categories like Black and White. However, there is important variation within the Latinx category, including income, educational attainment, and linguistic expression and repression (Giebel et al, 2022;Portes & MacLeod, 1996;Mendoza-Denton, 1999). Latinx communities have also been subjected to policies at the subgroup level, such as the World War II era Bracero Program that brought Mexican laborers to US farms or the Cuban Adjustment Act which, among other things, streamlined the immigration process for Cubans coming to the United States.…”

“…Intracategorical frameworks have been used in other studies using machine learning to analyze college admissions essays and text from other evaluative settings. For example, strong patterns have been shown in college admissions essays broken down by applicant gender and declared major (Giebel et al, 2022), academic background of transfer applicants , and by multiple social categories for parole candidates in hearing transcripts (Bell et al, 2021). These and other studies use a variety of machine learning techniques for text, such as topic modeling, word embeddings, named entity recognition, and document classification all to show how documents vary along important social dimensions.…”