Developing generative AI literacies through self-regulated learning: A human-centered approach

Generative artificial intelligence (AI) creates both opportunities for enhanced learning and risks of skill erosion and dependency. This exploratory, mixed-methods study investigated how to design AI learning experiences using a human-centered approach that promotes student agency. We developed and investigated an integrated framework combining comprehensive generative AI literacies—functional, critical and ethical, and creative—with self-regulated learning (SRL) processes operationalized as a human-centered Plan, Iterate, Evaluate cycle. Thirty-eight undergraduate students enrolled in an “Artificial Intelligence and Writing” course completed scaffolded experiential challenges followed by self-directed creative projects. Quantitative analysis revealed significant growth in AI literacy self-efficacy across all dimensions, with students progressing from moderate initial confidence (M = 4.68, SD = 2.11) to high confidence levels (M = 8.39, SD = 1.04) on a 10-point scale (t = -9.86, p < .001). Qualitative analysis of project artifacts and student process reflections identified a taxonomy of human in the loop practices integrating AI literacies and self-regulation across the Plan, Iterate, Evaluate cycle. Planning practices involved activating domain knowledge to identify AI applications and establishing evaluative criteria. Iteration practices included developing multi-step workflows, refining prompts through dialogue, and monitoring output quality. Evaluation practices combined assessment of project outcomes with reflection on collaboration processes to inform future use. These practices illustrated adaptive human-AI collaboration strategies that augment rather than replace students’ disciplinary expertise and creative vision. These findings suggest scaffolded experiential learning integrating AI literacies and metacognitive processes can promote effective AI collaboration and empower students to actively direct their own learning. • Integrated AI literacies with self-regulated learning in Plan-Iterate-Evaluate cycle • Developed AI literacy self-efficacy measure with excellent reliability (α=.94-.96) • AI literacy self-efficacy increased significantly from M=4.68 to M=8.39 (p<.001) • Identified taxonomy of human in the loop practices for AI-enhanced learning • Experiential and project-based learning interventions promoted human agency with AI