Federal Reserve Board research maps generative AI exposure across college majors, demographics, and institution types

The Federal Reserve Board published a FEDS Note assessing which college majors, demographic groups, and types of postsecondary institutions may be most affected by generative artificial intelligence (AI) through its predicted impact on the labor market. The analysis links occupational exposure scores for two generative AI applications, language modeling and image generation, to the National Survey of College Graduates (2013–2021) to construct relative exposure measures for fields of study and 1994 Carnegie institutional classifications. The note finds that Mathematics and Computer Science-related fields have relatively high exposure to both language modeling and image generation, while Engineering and Technology-related fields are most exposed to image generation; Accounting and Political Science and Government also feature among the most exposed majors for language modeling. It reports significant correlations suggesting that majors with higher shares of Hispanic graduates are more exposed to language modeling (correlation coefficient 0.21) and majors with higher shares of Asian graduates are more exposed to image generation (0.47), while majors with higher shares of female, white, and black graduates are less exposed to image generation (-0.67, -0.29, and -0.25). By institution type, Liberal Arts I institutions show the highest relative exposure to language modeling and Research University I institutions the highest relative exposure to image generation at both bachelor’s and graduate levels, implying that any generative AI-driven changes in labor demand could translate into more pronounced shifts in student engagement and staffing at those institutions.