AI Training is Increasing but not equally: AI Training Inequality in Higher Education 

• In Event: The Emerging Role of AI in Education and Workforce Preparation

Abstract

Artificial Intelligence (AI) is rapidly reshaping employers’ demand for skills. In response, higher education institutions have increasingly invested in AI education to prepare the future workforce with relevant competencies. However, the existing literature has limited evidence on how students’ access to and completion of AI-related coursework have evolved over time, which academic fields are driving the growth in AI education, and whether access to AI education is equitably distributed across different student populations.  

In this paper, we ask three questions. First, to what extent has AI course offerings and students’ AI exposure increased over time? Second, are there disparities in AI exposure by gender, race/ethnicity, field of study, and institutions, and are these gaps growing or shrinking? Third, what are the underlying drivers of aggregate AI exposure growth? Are STEM majoring students taking more AI courses, or are students sorting into STEM majors? Furthermore, how do these dynamics differ across student subpopulations?  

We use more than twenty years of student enrollment records from the Ohio Higher Education Information (HEI) system, linked with UMETRICS data and course descriptions from public university catalogs. We identify AI-related courses using two complementary approaches. First, we apply a large language model to course descriptions to generate course-level AI indicators. Second, we flag courses taught by researchers actively engaged in AI research, identified through UMETRICS grant records.  To examine how students’ AI exposure evolves over time, we measure AI exposure as the credit-hour-weighted share of AI coursework completed by each student. We further track these trends across cohorts by student demographics and academic fields. Finally, we use a decomposition framework that attributes growth in AI exposure to four channels: expansion of AI course offerings, sorting across institutions, sorting across majors, and course selection within majors.  

Our preliminary results show that AI exposure has increased substantially across cohorts, with the share of students with no AI courses shrinking and the share with high AI exposure rising sharply since the early 2010s. However, this growth is unequal. AI exposure has been increasing faster among men than women, among international and Asian students relative to other racial/ethnic groups, and among STEM majors relative to non-STEM students. Results from the decomposition analysis suggest that course offerings and major sorting account for most of the aggregate increase. Both mechanisms concentrate AI exposure gains in male-dominated, STEM-heavy fields, so the same forces driving aggregate growth are also widening the gender gap.   

These findings suggest that while higher education institutions have been successful in expanding AI-related coursework, this growth is concentrated in specific fields, especially STEM majors, and among already-advantaged demographic groups and international students. Unequal access to AI training across student groups may further widen disparities in labor market outcomes, including earnings, employment, and career advancement. Our results suggest that when expanding AI education, higher education institutions should broaden and equalize access to AI training to promote more inclusive skill development, as AI becomes increasingly central to the workforce.

Author

• Sujin Lee, The Ohio State University
  Presenting Author