Text Mining Undergraduate Engineering Programs' Applications: the Role of Gender, Nationality, and Socio-economic Status

Bo Lin, Bissan Ghaddar, Ada HurstarXiv 2021 · OpenReview

TL;DR

This paper introduces a novel text mining approach combining LDA and word embeddings to analyze the motivations of over 43,000 engineering program applicants. The study reveals that while interest in technology and social impact are top motivators, gender has the most significant and robust influence on why students choose engineering, even more so than nationality or socioeconomic status.

What Problem Does This Paper Address?

Despite decades of efforts, women and other underrepresented groups remain significantly underrepresented in STEM, particularly in engineering. To effectively address this, we need to understand the underlying motivations that drive students to pursue engineering. Previous research often relied on smaller sample sizes and surveys, limiting generalizability. This paper aims to bridge this gap by analyzing a large dataset of university applications to uncover the nuanced motivational factors and how demographic characteristics like gender, nationality, and socioeconomic status shape these motivations. This understanding is crucial for designing targeted and effective recruitment strategies to promote equity, diversity, and inclusion (EDI) in STEM.

Methods

The authors developed a novel text mining approach that integrates two powerful techniques:

  • Latent Dirichlet Allocation (LDA): This is a standard topic modeling algorithm used to discover abstract "topics" that occur in a collection of documents. In this context, LDA helps identify common motivational themes within the applicants' essays.
  • Word Embeddings (specifically, pre-trained Word2Vec): Word embeddings represent words as dense vectors in a continuous space, capturing semantic relationships between words. By using pre-trained embeddings, the approach can understand the meaning of words and phrases beyond simple co-occurrence, allowing for a more nuanced interpretation of topics and their relationships.

The proposed method combines these by first using LDA to extract topics (motivations) from the applicant essays. Then, a pre-trained Word2Vec model is used to represent these topics and individual sentences in a semantic space. This allows for a more robust assignment of motivations to each applicant, capturing semantic similarity rather than just keyword matching. This hybrid approach aims to reduce subjectivity and the need for extensive manual intervention compared to traditional methods.

Datasets

The primary dataset used in this study is:

  • University Application Data: A collection of 43,645 applications to 13 direct-entry engineering programs at a large Canadian university, spanning the years 2013-2017. This dataset includes free-text responses from applicants explaining their reasons for applying.
  • Canadian Census Data (2016): This data was used to enrich the application dataset by providing demographic information (income and education levels) based on the applicants' home postal codes.

Key Insights and Ideas

  • Beyond Keywords: The study highlights the power of semantic understanding in text analysis. Simply counting keywords misses nuanced motivations; for example, essays about a robotics project and a hospital internship might both stem from a "previous experience" motivation, even with different vocabulary.
  • The "Why" Behind Engineering: The research moves beyond simply identifying what students want to do and delves into why they choose engineering, connecting these motivations to broader demographic and cultural contexts.
  • Gender as a Dominant Factor: The finding that gender has the strongest and most robust impact on motivations is particularly striking, suggesting that gender-based differences in aspirations and perceptions are deeply ingrained and persistent across other demographic factors.
  • Cultural Nuances in Nationality: The differences observed between Canadian and international applicants point to the influence of cultural values (individualistic vs. collectivist) on career motivations.

Key Results

  • Top Motivators: The two most powerful motivators for engineering applicants are "Interest in technology" and the desire to "make a social impact."
  • Gender Differences:
    • Males are more likely to be motivated by "childhood dreams" and "previous experience" in STEM.
  • Females are more likely to emphasize "public issues," "problem solving," and "societal contributions."
    • Female applicants are also more likely to express a "love of science."
    • These gender differences were found to be largely consistent across nationality and income levels.
  • Nationality Differences:
    • Canadian applicants are more driven by "individualistic" factors like personal interest, professional development, and technical interests.
  • International applicants (particularly from Asia, the Middle East, and North Africa) show a greater emphasis on "collective good," public issues, and mentorship.
  • Socioeconomic Status (Income & Education):
    • Higher family income is associated with motivations like "problem solving," "love of science," and "professional development."
    • Lower income is linked to interests in "technology" and "childhood dreams," with a greater focus on societal contributions.
  • Family education also plays a role, with higher education levels correlating with "technical interests" and "problem solving," while lower education is linked to "mentorship" and "professional development," especially in higher-income families.
  • Gender's Robustness: Gender-based differences in motivations were found to be the most consistent and robust, persisting across nationality, family income, and education levels, with only minor exceptions.

Limitations and Future Work

  • Data Limitations: The study acknowledges the lack of relevant data for international students regarding family income and education, preventing a full analysis of these factors for this group. Ethnicity was also not collected, which could provide further insights. The absence of final admission decisions limits the ability to use them as a proxy for educational attainment.
  • Future Work: The authors suggest that future research could explore the impact of ethnicity and use admission decisions as a proxy for attainment. The findings provide a strong basis for developing targeted recruitment strategies, particularly for underrepresented groups, by emphasizing factors that resonate with them, such as practical significance, social impact, and mentorship opportunities.

Summary updated

Abstract

Women, visible minorities, and other socially disadvantaged groups continue to be underrepresented in STEM education. Understanding students' motivations for pursuing a STEM major, and the roles gender, nationality, parental education attainment, and socio-economic background play in shaping students' motivations can support the design of more effective recruitment efforts towards these groups. In this paper, we propose and develop a novel text mining approach incorporating the Latent Dirichlet Allocation and word embeddings to analyze applicants' motivational factors for choosing an engineering program. We apply the proposed method to a dataset of 43,645 applications to the engineering school of a large Canadian university. We then investigate the relationship between applicants' gender, nationality, and family income and educational attainment, and their stated motivations for applying to their engineering program of choice. We find that interest in technology and the desire to make social impact are the two most powerful motivators for applicants. Additionally, while we find significant motivational differences related to applicants' nationality and family socio-economic status, gender has the strongest and the most robust impact on students' motivations for studying engineering.
BibTeX
@misc{lin2021text,
  author = {Bo Lin and Bissan Ghaddar and Ada Hurst},
  title = {{Text Mining Undergraduate Engineering Programs' Applications: the Role of Gender, Nationality, and Socio-economic Status}},
  year = {2021},
  eprint = {2107.14034},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2107.14034},
}
Figure