Generative AI across the disciplines: an activity theory perspective on undergraduate students' AI use and disclosure practices

Synthesis: Generative AI across the disciplines — a multi-institutional quantitative study of 560 undergraduates across five academic domains examining how academic disciplines function as activity systems that shape students' GenAI use and disclosure practices. Jiang, Farag, Lucia, Vetter & Silvestro interpret disciplinary differences in GenAI engagement through the lens of activity theory.

Jiang et al. address the under-studied question of disciplinary differences in students' GenAI use and disclosure. Drawing on activity theory, they treat each academic discipline as an activity system in which students, GenAI applications, course policies, and role expectations interact to shape engagement. Survey data were analyzed with descriptive statistics, chi-square tests of association, and ordinal regression.

Method

• Design: Multi-institutional quantitative survey.
• Sample: 560 undergraduates across five academic domains.
• Analysis: Descriptive statistics, chi-square tests of association, ordinal regression; interpreted through disciplinary contradictions in activity theory.

Key Findings

• Significant associations between disciplinary affiliation and students' GenAI usage frequency and disclosure behaviors.
• Disciplines differ systematically in how students use and disclose GenAI, reflecting each field's norms, policies, and role expectations.
• Patterns are interpreted through disciplinary contradictions — tensions within each activity system (e.g., between course policies and actual practice, or between expectations for students and how GenAI is embedded).

What this means for practice

• Instructors. State GenAI disclosure expectations explicitly in each course, because disclosure varied significantly by discipline and by course-level policy rather than following a general institutional norm.
• Administrators. Write discipline-aware GenAI and Academic Integrity guidance instead of one uniform rule; identical policy language lands differently in fields whose course policies, instructor roles, and disciplinary norms diverge.
• Faculty developers. Help instructors align stated course policy with actual classroom practice, since the contradiction between written rules and tolerated use is itself what shapes students' use and disclosure.
• Researchers. Treat disciplinary activity systems — not individual ethics — as the unit of analysis when studying GenAI disclosure, which is a socially situated practice.

Limitations

• All four institutions sit in the northeastern United States and enroll mainly white, middle-class students, so generalization across geographic locations and institutional types is limited.
• The survey was advertised as one about GenAI, which likely recruited students already familiar with the tools; the most anxious or least experienced students probably did not respond, so reported non-use is understated.
• Responses were self-reported in a single wave with a gift card incentive, which may have encouraged quick completion, and some students may have underreported use despite anonymity assurances.
• Only 560 undergraduates at four institutions were surveyed, graduate students were excluded, and limited demographic data were collected.

Citation

Jiang, J., Farag, I., Lucia, B., Vetter, M. A., & Silvestro, J. (2026). Generative AI across the disciplines: an activity theory perspective on undergraduate students' AI use and disclosure practices. SSRN Working Paper.