Guided Integration or Prohibition? A Quasi-Experimental Study of Generative AI in Project-Based Learning

Abstract

Purpose: The rapid expansion of generative artificial intelligence (AI) tools in higher education has prompted institutional responses ranging from prohibition to structured integration. However, empirical evidence comparing these approaches within authentic classroom contexts remains limited. This quasi-experimental study examines whether guided AI integration enhances student learning outcomes in a project-based undergraduate course. Methodology: Two intact sections of a senior-level global trade course (N = 60) completed an identical semester-long group project. One section operated under a structured AI policy (Experimental Group), while the other followed a strict human-only policy (Control Group). Findings: Independent-samples t-tests revealed statistically significant differences in overall project performance favoring the Experimental Group across written and presentation components. Students in the Experimental Group demonstrated stronger analytical organization, clearer conceptual application, and more effective synthesis of course content. Self-reported measures further indicated higher perceived learning, greater professional growth, and reduced stress. Indicators consistent with AI-assisted writing were observed in some Control Group submissions; however, these observations were not formally quantified and are interpreted cautiously. Unique Contribution to Theory, Policy and Practice: Findings suggest that structured and transparent AI integration, when aligned with instructional design principles, may support stronger performance outcomes than prohibition-based approaches within this course context. The study contributes empirical evidence to ongoing discussions regarding AI governance in higher education and offers implications for instructional policy design.