Each dot is one college major (285 in all). Horizontal axis: percentile rank of the major's AI exposure among all 285 majors (0 = least exposed, 100 = most exposed). Vertical axis: median earnings one year after graduation, in US dollars. Exposure is the AIOE index (Felten, Raj and Seamans, 2021) averaged over the occupations each major leads to. It measures how much an occupation's tasks overlap with AI capabilities, not whether jobs will be automated.
Panels are my own grouping of the CIP fields into eight larger groups, ordered by the correlation in the panel. Engineering & computing: engineering, engineering technology, architecture, computer science. Natural sciences & agriculture: biological and physical sciences, mathematics and statistics, agriculture, natural resources, science technology. Social sciences & psychology also includes public administration, communication and legal studies. Humanities & arts: English, languages, philosophy, religion, history, area studies, liberal arts, visual and performing arts. Other: interdisciplinary majors and small fields (family and consumer sciences, homeland security, trades and others). Within-group correlations have wide intervals (95% intervals use the Fisher transformation) and should not be read as field-level effects. Majors are unweighted by size.
Data: College Scorecard, AIOE, NCES CIP-SOC crosswalk. TidyTuesday 2026, week 42. Ekaterina Enchikova.