Proceedings of the Annual Meeting of the Cognitive Science Society

As access to information expands, learners must decide not only what to learn, but from whom or where. Research in self-regulated learning and epistemic cognition shows that individuals hold differentiated beliefs about source reliability and usefulness, yet it remains unclear whether these evaluations guide source selection in early adulthood. We interviewed 116 college students who imagined preparing an internship presentation on either a novel or partially known concept. Participants described how they would learn it and evaluated their chosen source on perceived knowledge, trustworthiness, expected learning, and enjoyment. Non-AI platforms were most frequently selected (73.28%), followed by AI (21.55%) and human-sources (5.17%). Ordinal models revealed that humans were rated as significantly more trustworthy, enjoyable, and effective for learning, despite being rarely chosen. Their explanations prioritized availability and speed. Together with developmental findings in children, results reveal a dissociation between epistemic evaluation and learning decisions, highlighting efficiency-driven constraints on self-regulated learning.