This study explores how cognitive traits, demographics, and learning strategies influence the adoption of an AI Virtual Teaching Assistant Jill Watson (JW) in a self-regulated learning (SRL) environment. Among 81 students, 46% actively engaged with JW, underscoring the importance of understanding adoption factors. Adopters exhibited significantly higher Need for Cognition (NFC) scores (p < 0.001), suggesting that students inclined toward deeper cognitive engagement were more likely to incorporate AI support in their SRL strategies. Age also showed a strong positive association with adoption (p < 0.001), with older learners (mean age = 35) being more likely to use JW than younger ones (mean age = 32). These findings indicate that older learners may perceive AI assistance as a valuable tool for navigating complex learning tasks, leveraging it for both theoretical understanding and task-specific problem-solving.

Interestingly, students who frequently engaged in help-seeking (HS) (p = 0.019) and peer learning (PL) (p = 0.002) were less likely to adopt JW, suggesting a potential preference for human-based learning interactions. This challenges the expectation that students with strong SRL behaviors would incorporate AI as an additional support mechanism. The reduced adoption among socially oriented learners highlights a key limitation of AI-based learning tools, which may lack the dynamic, interactive qualities that peer collaboration and instructor support provide. While prior research has suggested that self-efficacy (SE) plays a critical role in SRL and technology adoption, our analysis found no significant difference in SE between adopters and non-adopters (p = 0.377). This suggests that while SE is important for overall SRL behaviors, it may not directly influence the choice of AI-based support versus traditional learning strategies.

These findings underscore key barriers to AI adoption in education. Students with lower NFC may struggle with AI engagement due to difficulties in meta-cognitive monitoring, aversion to ambiguity, and challenges in identifying information needs. AI tools that passively respond to queries may not sufficiently support learners who require guided scaffolding to enhance their SRL processes. Additionally, socially driven learners may find AI assistants less appealing due to their lack of interpersonal dynamics. To foster broader adoption, AI tools should integrate proactive features such as structured prompts, meta-cognitive monitoring support, and strategic learning recommendations that cater to diverse SRL needs.

By leveraging a combination of user log data and survey measures—rather than relying solely on self-reported perceptions—this study provides novel insights into the cognitive and behavioral factors influencing AI adoption. These results offer a foundation for designing AI learning assistants that align with different learner profiles, promoting engagement across diverse educational settings. Future research should explore how AI tools can better replicate social learning interactions and provide adaptive support to accommodate varying cognitive traits and SRL strategies.

Keywords: AI Adoption, Education, e-learning, Virtual Teaching Assistants, Need for Cognition, Self Efficacy, Demographics, Help Seeking, Peer Learning, Self-directed Learning, Inquiry-based Learning.