Learning in AI-Augmented Environments Through Dual Pathways of AI Literacy

Learning in AI-Augmented Environments Through Dual Pathways of AI Literacy | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Learning in AI-Augmented Environments Through Dual Pathways of AI Literacy Jing Betty Feng, Gregorio Diaz, Zachary Long This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9173238/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Generative artificial intelligence (AI) tools are becoming embedded in higher education, raising questions about how learning unfolds in AI-augmented environments. This convergent mixed-methods study examines how three dimensions of AI literacy - technical, cognitive, and social-ethical - relate to learning processes and outcomes. Qualitative analysis of ten interviews with college students shows that learners engage with AI pragmatically but cautiously, using it to support efficiency, comprehension, and academic work while actively regulating use to preserve learning quality and integrity. Quantitative survey results show that technical AI literacy is associated with greater use of AI-assisted learning behaviors, which are in turn associated with higher perceived learning gains. Cognitive AI literacy is directly associated with perceived learning gains and stronger social-ethical awareness, whereas social-ethical AI literacy shows no direct association with learning behaviors or outcomes. Together, the findings indicate that learning in AI-augmented environments is shaped by two complementary pathways: 1) a technical–behavioral pathway that supports effective learning engagement with AI tools, and 2) a cognitive–ethical pathway that supports evaluative judgment and responsible use. These results highlight the importance of designing learning environments and assessment practices that support both procedural fluency and reflective judgment in AI-integrated higher education. Educational Philosophy and Theory AI literacy academic learning gains mixed-methods higher education AI-assisted learning behaviors AI ethics Figures Figure 1 Figure 2 Introduction The rapid diffusion of generative artificial intelligence (GAI), including large language models (LLMs) such as ChatGPT, Grammarly, and AI-augmented search engines, is reshaping contemporary academic learning environments. Rather than functioning as isolated instructional tools, these systems increasingly operate as persistent features of students’ everyday academic contexts, influencing how information is accessed, interpreted, and produced across learning tasks (D. Kim et al., 2025; Shardlow et al., 2022). Throughout this article, we use the term ‘AI’ to refer specifically to these GAI tools. As AI becomes embedded within the instructional ecology of higher education, the central question has shifted from whether students should use AI to how AI-augmented learning environments shape students’ learning processes and orientations. Research offers mixed evidence regarding the educational consequences of AI-rich environments. While some highlight the benefits such as improving access to information, facilitating deeper engagement, and promoting creativity (Eden et al., 2024), others caution that overreliance may diminish critical thinking and potential ethical risks (Grassini, 2023). In particular, a 2025 study at MIT’s Media Lab raised concerning results that the usage of ChatGPT could harm learning and reduce critical thinking abilities due to passive behaviors (such as copy/paste) (Kosmyna et al., 2025). These findings indicate that learning outcomes in AI-augmented environments depend not only on the presence of AI, but on how learners engage with it. Scholars have emphasized the importance of designing learning environments that support productive and responsible engagement with AI (Ifenthaler et al., 2024; Walter, 2024). AI literacy, the ability to understand, evaluate, and ethically engage with AI-generated content (Chang & Wong, 2025; Eisenbardt et al., 2025; Long & Magerko, 2020), has been emphasized as an emergent needs to ensure students learning outcomes and academic performance (Acosta-Enriquez et al., 2025). Extant literature has examined factors such as students’ perceptions, attitudes, trust, and perceived usefulness of AI (Bewersdorff et al., 2025; Hornberger et al., 2023; Long & Magerko, 2020). However, because AI literacy encompasses technical, cognitive, and social-ethical dimensions (Ng, 2012), it remains unclear which dimensions are empirically associated with students’ learning behaviors. In particular, AI-assisted learning behaviors (AALB), how students strategically incorporate AI into academic work, remain underexamined in relation to specific dimension of AI literacy (Elshall & Badir, 2025). This study aims to answer the research question: How does AI literacy shape learning processes and outcomes within AI-augmented learning environments? We employ a convergent mixed-methods (QUAL + QUAN) design (Fetters et al., 2013), integrating semi-structured interviews with a quantitative survey of college students. This approach enables examination of both learning outcomes and the meaning-making processes through which students interpret and navigate AI-supported environments. Qualitative findings identify key dimensions of student engagement, including academic enablement, AI literacy as competence, conditional trust and risk management, human–AI boundaries, and pedagogical guardrails . The quantitative results converge with and extend these insights by revealing two complementary pathways through which learning unfolds in AI-augmented environments. A technical–behavioral pathway, supported by Technical AI literacy, promotes AI-assisted learning behaviors and indirectly enhances perceived learning gains. A cognitive–ethical pathway, supported by Cognitive AI literacy, directly contributes to reflective judgment and perceived learning gains, alongside Social–ethical AI literacy. Notably, Social–ethical AI literacy shows no direct association with AI-assisted learning behaviors or performance outcomes. These findings suggest that AI-augmented learning environments do not produce uniform effects but instead support distinct pathways of engagement shaped by how learners interpret and regulate AI use within environmental conditions. This study offers a pedagogical framework for understanding how learning environments can be designed to support both procedural fluency in AI use and reflective, ethical engagement with AI-generated knowledge. Literature Review AI Literacy: Definition and Dimensions As AI becomes a persistent feature of academic learning environments rather than an optional tool, AI literacy increasingly functions as a capacity for navigating, interpreting, and regulating AI-augmented learning conditions (D. Kim et al., 2025). AI literacy is defined as “the ability to understand, use, monitor, and critically reflect on AI applications without necessarily being able to develop AI models themselves” (Laupichler et al., 2022). More broadly, AI literacy encompasses a multidimensional set of competencies that enables individuals to critically appraise AI technologies, interact productively with AI systems, and apply them effectively across academic, professional, and everyday contexts (Long & Magerko, 2020). Building on Ng’s (2012) framework for digital literacy, AI literacy comprises three interrelated dimensions: technical, cognitive, and social-emotional. Technical AI literacy refers to the ability of locating and using AI tools effectively in a given discipline and instructional contexts (Ng, 2012). Technical literacy shapes learners’ ability to engage with AI as an available environmental resource, including understanding their functionalities, distinguishing between academic and creative tasks, and selecting appropriate forms of AI support (Ng, 2012). This foundational capacity is critical for enabling students to participate meaningfully in AI-supported learning activities. Cognitive AI literacy refers to the awareness and ability to engage with AI as a collaborative and critically evaluating AI outputs for credibility, authority, and appropriateness (Chiu et al., 2024; Ng et al., 2021). Students must demonstrate and apply critical thinking and problem-solving skills when interacting with AI (Bewersdorff et al., 2025; Ng et al., 2021), including assessing reliability, recognizing bias, and determining when AI outputs are pedagogically appropriate. Cognitive AI literacy supports learners’ ability to interpret and evaluate information generated by AI, allowing learners to balance reliance on AI-generated information with reflective judgment and critical evaluation (Chen & Zare, 2025). Social-Ethical AI Literacy refers to the understanding of ethical implications and societal impacts of AI (Eisenbardt et al., 2025; Kajiwara & Kawabata, 2024). Socio-ethical literacy shapes how learners establish boundaries between human and algorithmic agency, including concerns related to transparency, data privacy, academic integrity, and the credibility of AI-generated knowledge (Acosta-Enriquez et al., 2025; Chang & Wong, 2025). While AI have shown to enhance students’ confidence, motivation, and engagement (Ji et al., 2025; Luca Liehner et al., 2023), they also raise concerns about responsible use and epistemic trust. Recent debates surrounding wearable and relational AI systems further underscore how AI is reshaping not only information practices but also learners’ relationships with knowledge and authority (Montali, 2025). Consequently, fostering AI literacy within academic environments requires attention to ethical responsibility, critical awareness, and institutional guidance (Crompton & Burke, 2023; Eden et al., 2024). Despite the growing scholarly attention to AI in education, notable gaps persist in understanding how each dimension of AI literacy translates into meaningful learning outcomes Much of the existing literature emphasizes psychological or pedagogical correlates such as self-efficacy, creativity, and satisfaction, while the environmentally situated cognitive and socio-ethical processes through which learners engage with AI remain underexplored (Chiu, 2025; Chun et al., 2025; Zawacki-Richter et al., 2019). Addressing these limitations requires moving beyond general accounts of AI literacy to examine how its distinct dimensions shape learners’ orientations and behaviors within AI-augmented learning environments, particularly in relation to critical thinking and perceived learning outcomes (Chiu, 2025; J. Kim et al., 2025; Walter, 2024). Integration of AI in Academic Learning The integration of AI into higher education has transformed students’ learning experiences, influencing academic performance, cognitive processes, and decision-making strategies (Grassini, 2023; Medina-Gual & Parejo, 2025). AI tools are increasingly used for writing, summarization, and content creation (Eden et al., 2024; Long & Magerko, 2020). As AI becomes environmentally embedded rather than episodically used, its influence extends beyond task completion to shaping how learners approach academic work more broadly. While AI applications can enhance efficiency and productivity, improper or unreflective engagement, such as uncritical copy-pasting or overreliance on generated outputs, may undermine deep learning and critical thinking (Kosmyna et al., 2025). This distinction between AI use and AI literacy is crucial in learning environments research: learners may operate in AI-augmented settings without possessing the evaluative or integrative capacities necessary to engage productively with AI-generated knowledge (Crompton & Burke, 2023; Ng et al., 2021). Empirical studies suggest that well-designed AI can enhance learners’ confidence, autonomy, and motivation, reinforcing cognitive engagement and problem-solving (Bewersdorff et al., 2025; Morales-García et al., 2024). AI-supported learning has also been associated with increased self-efficacy, creativity, and satisfaction among students and educators (Ji et al., 2025; Kong et al., 2022). These findings position AI not simply as a digital assistant, but as a feature of the learning environment that can function as a cognitive partner, depending on how learners interpret and engage with its affordances. Academic performance, however, captures only a partial dimension of learning. Grade point average (GPA) primarily reflects assessment outcomes rather than depth of understanding, critical thinking, or transfer of knowledge (Grassini, 2023). In contrast, perceived learning gains reflect students’ self-assessment of their intellectual growth, skill development, and ability to apply concepts meaningfully (Seymour et al., 2000). From a learning environments perspective, these indicators may diverge, as environmental conditions that shape student behavior can support performance-oriented outcomes without necessarily fostering deeper learning. Accordingly, perceived learning gains provide a complementary lens for examining how AI literacy shapes learning quality and longer-term educational development. Based on prior research on AI literacy, we propose the following hypotheses: H1. Each dimension of AI literacy – technical, cognitive, and ethical - is positively associated with perceived learning gains. AI-Assisted Learning Behaviors as Mediators While prior research demonstrates that digital and AI literacy are positively associated with academic performance (Hornberger et al., 2023), empirical evidence explaining how these competencies translate into learning gains remains limited (Walter, 2024). Within the AI-augmented learning environment, this gap suggests that literacy alone does not determine learning outcomes; rather, its effects depend on how learners enact AI-supported behaviors within specific contexts (J. Kim et al., 2025). Learning behaviors have long been recognized as critical determinants of academic success (Seymour et al., 2000). Within the context of AI, AALB may refer to students’ active and strategic engagement with AI tools to support their learning, include using AI for concept clarification, generating examples, revising drafts, conducting self-testing, and exploring ideas iteratively. These practices extend beyond passive tool use, reflecting the degree to which students leverage AI to enhance comprehension, critical thinking, and metacognitive awareness. Research indicates that AI-supported environments can foster confidence and academic engagement by helping students complete assignments, clarify complex concepts, and sustain motivation (Grassini, 2023). However, these benefits depend heavily on how students engage with AI: Learners with limited AI literacy may adopt surface-level or dependent practices, whereas learners with higher literacy demonstrate more reflective and self-regulated engagement (Luca Liehner et al., 2023). Exposure to AI technologies may promote trust and optimism, but productive learning outcomes require ethical reflection and adaptive use grounded in AI literacy (D. Kim et al., 2025; J. Kim et al., 2025). Accordingly, this study proposes that AALB represents a mechanism through which AI literacy enhances learning outcomes through a mediation relationship between AI literacy and learning gains H2. AI-assisted learning behaviors mediate the relationship between each dimension of AI literacy and learning gains. Methodology Research Design To examine how AI literacy shapes learning processes and outcomes within AI-augmented learning environments, this study employed a convergent mixed-methods design (QUAL + QUANT), in which qualitative and quantitative data were collected in parallel, analyzed separately, and integrated during interpretation(Fetters et al., 2013). This design allows for complementary examination of learning outcomes and the meaning-making processes through which students engage with AI-supported environments. Both qualitative and quantitative components were given equal analytic weight, enabling a balanced interpretation of learner experiences and outcome patterns (Younas et al., 2023). Rather than treating AI use as a discrete intervention, the mixed-methods approach was selected to capture how students interpret, navigate, and respond to the future of AI environment. Quantitative analyses identify relationships among AI literacy, AI-assisted learning behaviors, and learning outcomes, while qualitative analyses provide contextual insight into how these relationships are enacted in practice. Integrating both strands supports a more comprehensive understanding of how AI literacy operates within contemporary academic learning environments. Integrating both strands supports a more comprehensive understanding of how AI literacy operates within contemporary academic learning environments (Fetters et al., 2013; Younas et al., 2023). Qualitative Data Collection and Participants During the summer of 2025, we conducted 10 semi-structured interviews with students at a regional college in the Northeastern United States to explore their engagement with AI. The purpose of the interviews was to explore how students experience and interpret AI as part of their academic learning environment. Two trained student research assistants conducted the interviews using a common interview protocol designed to elicit students’ awareness of AI, their rationales for using AI tools in academic and personal contexts, their understanding of AI literacy, and the perceived influence of AI on their learning and future academic or professional trajectories. Participants represented a range of academic majors and grade levels, providing variation in disciplinary context and learning experience. Interviews lasted between 20 and 50 minutes, were audio recorded with participants’ consent, and subsequently transcribed verbatim for analysis. Table 1 presents demographic and academic characteristics of the interview participants. -------------------------------- Insert Table 1 About Here -------------------------------- Qualitative Data Analysis The qualitative phase employed Braun and Clarke’s (2006, 2021) inductive thematic analysis to examine patterns in students’ experiences with AI in academic learning contexts. During the initial open-coding phase, the research team conducted inductive first-cycle coding to identify recurring patterns in participants’ accounts of AI use in academic contexts. Coding focused on capturing concrete, descriptive features of students’ engagement with AI, including primary tools used, academic applications, efficiency-oriented practices, learning simplification strategies, concerns about detection, and perceived risks. An initial codebook was developed from the first interview transcript and iteratively refined as additional transcripts were analyzed. As coding progressed, new codes were added when novel themes emerged, and existing codes were refined, merged, or split to ensure conceptual clarity and alignment with the data. Throughout this process, the research team reviewed all coding decisions, verified illustrative quotations against transcripts, and resolved discrepancies through discussion. This iterative procedure resulted in a finalized set of first-level descriptive codes. In the second analytic phase, first-level codes were grouped into axial themes that reflected broader patterns of meaning across participants’ accounts. These themes captured students’ interpretations of AI as an academic resource, their navigation of accessibility and ease of use, their developing understanding of AI literacy, and their reflections on career preparation and future competencies. Finally, axial themes were synthesized into five aggregated dimensions that constitute the Student AI Engagement framework: Academic Enablement, AI Literacy as Competence, Conditional Trust and Risk Management, Human–AI Boundaries, and Pedagogical Guardrails . This structured analytic process enabled interpretation of how students orient themselves within AI-augmented learning environments, capturing both enabling and constraining aspects of AI use. The resulting qualitative insights provide contextual grounding for the quantitative findings and support integrated interpretation of learner orientations and learning outcomes. Figure 1 presents an overview of the qualitative coding structure. -------------------------------- Insert Figure 1 About Here -------------------------------- Quantitative Data Collection Participants The quantitative survey was conducted in parallel with the interviews and included undergraduate students from multiple disciplines. Participants were recruited through campus announcements and course mailing lists over a period of three weeks. Eligibility requires full-time enrollment and informed consent. We received a total of 146 responses. After removing incomplete responses and those who failed the attention check, the final sample comprised 81 valid participants (man = 53, woman = 44, non-binary = 1, a gender not listed = 1, prefer not to say = 5), with an average age of 23. Measures and Operational Definitions Data were collected using a structured survey instrument based on validated scales and adapted items from prior research. All constructs were measured using a 5-point Likert scale (1 = Strongly Disagree, 5 = Strongly Agree) unless otherwise noted. Adapted from Ng (2012) digital literacy measurement, AI literacy was operationalized as a multidimensional construct comprising technical, cognitive, and social–ethical dimensions, reflecting learners’ capacity to navigate and regulate engagement within AI-augmented learning environments . Technical AI literacy (4 items; Cronbach’s α = .75) captures learners’ ability to locate and use AI tools effectively for information access and academic support. Sample items include “I can identify AI tools that are relevant for locating information in my field of study or work” and “I use AI tools to find information more efficiently than traditional search engines.” Cognitive AI literacy (5 items; Cronbach’s α = .74) reflects learners’ ability to critically evaluate AI-generated outputs for credibility, bias, and appropriateness, and to integrate AI support into their own thinking. Sample items include “I assess the credibility of AI-generated outputs by verifying them against trustworthy sources” and “I apply AI outputs in ways that support and extend my own thinking rather than replace it.” Social-Ethical AI literacy (5 items; Cronbach’s α = .59) addresses learners’ awareness of ethical and responsible AI use, including transparency, academic integrity, and ethical implications of AI-supported work. Sample items include “I disclose when I have used AI tools in my assignments or creative work” and “I understand the ethical implications of using AI in communication or decision-making.” The lower internal consistency reflects it diverse aspects of social-ethical AI awareness, however, is acceptable for a preliminary exploratory study to cover multiple dimensions rather than achieve high internal consistency. AI-assisted learning behaviors (AALB) (6 items; Cronbach’s α = .91) represent learners’ active and strategic use of AI to support academic learning within AI-augmented environments. Items assess behaviors such as concept clarification, example generation, revision, and self-testing (e.g., “I use AI tools to explain concepts I do not fully understand,” “I revise my academic work based on feedback from AI tools,” and “I use AI tools to practice or quiz myself.” ). Perceived learning gains (6 items, Cronbach’s α =.93), adapted from Seymour et al. (2000), capture learners’ self-reported intellectual growth, understanding, and motivation associated with AI-supported learning. (e.g., Gained a deeper understanding of concepts by exploring explanations or examples generated by AI tools,” and “Increased my motivation to learn by making study tasks more engaging with AI tools.” Additional variables included demographic information (age, gender, year of study, major) and descriptive indicators of AI experience, such as frequency of AI use for coursework and types of tools used (e.g., ChatGPT, Grammarly, NotebookLM). Quantitative Data Analysis Descriptive statistics were used to summarize participant characteristics and scale scores for AI literacy dimensions, AI-assisted learning behaviors, and perceived learning gains. Internal consistency of multi-item scales was assessed using Cronbach’s alpha. Zero-order correlations were examined to assess associations among study variables. To evaluate the proposed relationships, multiple regression analyses were conducted, including mediation models examining whether AI-assisted learning behaviors explained associations between AI literacy dimensions and perceived learning gains. Indirect effects were tested using non-parametric bootstrapping with 5,000 resamples, allowing for robust estimation of mediation effects without assuming normality of the sampling distribution. Findings Qualitative Findings The qualitative findings reveal how students interpret and respond to AI learning environments in their academic work. Analysis of interview data yielded five aggregated dimensions of students’ orientations toward AI within their learning environments. These dimensions illustrate how students leverage AI to support academic work while actively regulating its use to maintain learning quality, academic integrity, and personal agency. Table 2 summarizes the definition of each dimension and selected illustrative quotations. --------------------------------- Insert Table 2 About Here ---------------------------------- Academic Enablement Students consistently described AI as an enabling feature of their academic learning environment, particularly in relation to efficiency, task support, and cognitive load management. AI support was commonly used for note-taking, summarizing readings, clarifying concepts, debugging code, and organizing assignments. One participant explained, “I use this AI that records the professor and translates the lecture into notes. I use those notes to answer my questions, so I understand the lecture better” (Int.1). Others emphasized time savings and simplification: “They speed up my academic work, and they’re good at breaking things down into simpler terms” (Int.2). AI use was also normalized through peer practices, with students describing AI as a routine academic resource across disciplines. As one student noted, “Especially in writing-heavy classes. Even bio majors use it more now” (Int.9). These accounts highlight how AI functions as an enabling environmental resource that supports students’ day-to-day academic engagement. AI Literacy as Competence Students conceptualized AI literacy as a multidimensional competence extending beyond basic tool use. Rather than equating literacy with familiarity alone, participants described AI literacy as a combination of technical skill, evaluative judgment, and future-oriented preparedness. One student summarized this view: “AI literacy means knowing how AI models work, how to use them responsibly, and how to evaluate their output” (Int.10). Students frequently emphasized prompting, iterative refinement, and verification as central competencies. As one participant stated, “The biggest skill is knowing how to write a prompt. That’s the most important thing” (Int.5). Many also linked AI literacy to anticipated professional expectations, framing it as preparation for evolving workplace environments: “AI is already showing up in search engines, browsers, and even places like fast food drive-throughs. It’s becoming normal, and being familiar with it will be a big advantage” (Int.2). Together, these accounts position AI literacy as an adaptive competence that supports effective engagement to align with AI learning environments, rather than as simple technical proficiency. Conditional Trust & Risk Management Despite frequent AI use, students described exercising selective and conditional trust. Participants articulated ongoing evaluations of accuracy, appropriateness, and risk, particularly in relation to plagiarism, misinformation, privacy, and institutional consequences. One student explained, “For general or common information, I usually trust it. But for things like history, people’s beliefs, or anything for a research paper, I double-check it before trusting it” (Int.3). Ethical considerations also shaped usage boundaries. As one participant noted, “I don’t use it for cheating or anything like that, it doesn’t help me in that situation” (Int.6). Concerns about privacy (Int.4), misinformation (Int.6), and plagiarism (Int.7) were frequently raised. These accounts suggest that trust in AI is actively regulated, with students continuously assessing when AI support is appropriate and when independent verification is necessary. Human–AI Boundaries Students articulated clear boundaries between tasks they considered appropriate for AI support and those they viewed as requiring human judgment, creativity, or experiential knowledge. While AI was seen as effective for generating structured content and explanations, participants emphasized its limitations in interpretive, creative, or values-based domains. As one student observed, “AI can generate designs, but those lack real human concepts and meaning” (Int.4). Concerns about overreliance were also common. One participant explained, “I try not to rely on it as much because I know I need to remember that material in the future” (Int.2). Students further noted that perceived boundaries varied by discipline, with AI seen as more disruptive in fields such as computer science or English than in hands-on domains like nursing (Int.1; Int.2). Pedagogical Guardrails Students generally supported the inclusion of AI within higher education but emphasized the importance of clear pedagogical boundaries governing its use. Participants consistently framed AI as a support for learning rather than a replacement for instruction. As one student stated, “Yes, but only as a support tool, not to replace teachers” (Int.8). Students also highlighted the need for institutional guardrails when integrating AI in education, including clear policies, ethical guidance, and equitable access. As one participant noted, “Schools would have to be very careful. If we can’t make it safe and credible, then I don’t think it should have a place in education” (Int.7). These perspectives reflect a preference for AI learning environments that balance innovation with instructional integrity, transparency, and fairness, while preserving the central role of human instruction. Overall, the qualitative findings portray students as pragmatic and reflective participants in AI-augmented learning environments. While students actively leverage AI for academic enablement, they simultaneously engage in self-regulation to preserve learning quality, ethical standards, and human expertise. These findings illuminate the conditions under which AI supports or constrains learning and provide contextual grounding for the quantitative results. Quantitative Finding Associations Among AI Literacy, Learning Behaviors, and Learning Gains Table 3 presents descriptive statistics and zero-order correlations among the study variables. Technical and Cognitive AI literacy were moderately correlated (r = .30, p < .01), supporting their treatment as related but empirically distinct dimensions. Technical AI literacy was positively associated with AALB and perceived learning gains, while AALB showed a strong positive association with learning gains. In contrast, Cognitive AI literacy was modestly correlated with learning gains but showed no meaningful association with AALB. Social–ethical AI literacy exhibited weak, non-significant correlations with the other constructs. ----------------------------------- Insert Table 3 About Here ------------------------------------ We first examined whether AI literacy dimensions predicted perceived learning gains while controlling for age and school (Table 4). In the baseline regression model (Model 1), Technical AI literacy showed a strong positive association with learning gains (b = 0.81, SE = 0.13, t = 6.25, p < .001). Cognitive AI literacy was positively but not significantly related to learning gains (b = 0.26, SE = 0.16, t = 1.66, p = .10), and Social–ethical AI literacy was not associated with learning gains (b = −0.03, SE = 0.17, t = −0.19, p = .85). Students in the health school reported slightly higher learning gains than those in the reference school (b = 0.86, SE = 0.43, t = 2.00, p = .050), while age and other school indicators were not significant. The model explained 46.8% of the variance in learning gains (R² = .47; adj. R² = .41). These results indicate that, at the bivariate outcome level, technical aspects of AI literacy are most strongly associated with perceived learning gains. ----------------------------------- Insert Table 4 About Here ------------------------------------ AALB as a Mediating Process We next examined whether AALB explained how AI literacy relates to learning gains (Table 4). In Model A, AALB was regressed on the three AI literacy dimensions and controls. Technical AI literacy was a strong predictor of AALB (b = 0.88, SE = 0.12, t = 7.18, p < .001), whereas Cognitive (b = −0.10, SE = 0.15, p = .51) and Social–ethical AI literacy (b = −0.03, SE = 0.16, p = .84) were not significant predictors. This model explained 46% of the variance in AALB (R² = .46; adj. R² = .40). In Model B, learning gains were regressed on AALB, the three literacy dimensions, and controls. AALB emerged as a strong predictor of learning gains (b = 0.72, SE = 0.10, t = 7.52, p < .001). After accounting for AALB, the direct effect of Technical AI literacy was substantially reduced and no longer significant (b = 0.18, SE = 0.13, p = .17), consistent with an indirect-only mediation pattern. In contrast, Cognitive AI literacy showed a significant direct association with learning gains (b = 0.33, SE = 0.12, t = 2.86, p = .006), indicating an effect that operates independently of AI-assisted learning behaviors. Social–ethical AI literacy remained non-significant. Model B explained 71.3% of the variance in learning gains (R² = .71; adj. R² = .68). To formally test indirect effects, we conducted a non-parametric bootstrap with 5,000 resamples. Bootstrapped mediation analyses confirmed a significant indirect effect of Technical AI literacy on learning gains through AALB (b = 0.64, 95% CI [0.40, 0.91]). Indirect effects for Cognitive (b = −0.07, 95% CI [−0.32, 0.16]) and Social–ethical AI literacy (b = −0.03, 95% CI [−0.28, 0.22]) were not significant. As figure 2 illustrates, these results indicate two distinct pathways linking AI literacy to learning gains. Technical AI literacy primarily supports learning by increasing students’ engagement in AALB, whereas Cognitive AI literacy contributes directly to learning gains, independent of behavior frequency. In unstandardized terms, approximately 78% of the total effect of Technical AI literacy on learning gains operates indirectly through AALB, while the effect of Cognitive AI literacy is primarily direct. -------------------------------- Insert Figure 2 About Here -------------------------------- Discussion This mixed-methods study examined how students engage with AI within AI-augmented learning environments, focusing on how AI literacy relates to learning gains through situated learning behaviours. By integrating qualitative and quantitative findings (Younas et al., 2023 ), the study clarifies that AI literacy does not operate as a uniform advantage, but through distinct pathways that shape how learners engage with and regulate AI within specific learning conditions. The qualitative dimensions describe how students leverage AI to support learning while actively regulating its use to preserve learning quality, integrity, and agency. The quantitative findings reinforce this interpretation by showing that learning gains are associated with behavioural engagement and evaluative judgment, rather than AI use alone (Chiu, 2025 ; Ng, 2012 ). Interpretation of Findings in AI-Augmented Learning Environments Quantitative results show that Technical AI literacy is strongly associated with AALB, which in turn account for a substantial proportion of perceived learning gains. When learning behaviours are considered, the direct association between Technical AI literacy and learning gains diminishes, indicating that learning gains emerge primarily through how students act within AI-supported environments, rather than from technical skill alone. This finding aligns with prior research emphasizing the role of strategic use and self-regulated learning behaviours in technology-enhanced learning (Ng, 2012 ; Walter, 2024 ). Qualitative findings contextualize this pattern. Students described routinized prompting, iterative refinement, and verification practices that enabled AI to support comprehension and task completion. These practices illustrate how technical literacy functions as an enabling condition, allowing learners to take advantage of AI features embedded in the learning environment without defaulting to passive use (Chiu, 2025 ; Ng et al., 2021 ). In contrast, Cognitive AI literacy exhibits a different pattern. Quantitatively, it predicts learning gains directly but is not associated with increased AALB. Qualitative accounts suggest that cognitive literacy shapes evaluative judgment, particularly in tasks involving interpretation, originality, or high academic stakes. Students described cross-checking AI outputs, comparing them with course materials, and deliberately limiting AI use when it threatened learning quality or integrity. This evaluative role is consistent with prior work emphasizing critical judgment and metacognitive regulation in AI-supported learning environments (Bewersdorff et al., 2025 ; Eden et al., 2024 ). Social–ethical AI literacy did not directly predict learning gains or AALB, but qualitative evidence indicates that it plays a regulatory role. Students frequently referenced concerns about plagiarism, privacy, misinformation, and institutional penalties, echoing broader debates in the literature on ethical governance and trust in educational AI (Crompton & Burke, 2023 ; Kajiwara & Kawabata, 2024 ). Rather than generating learning behaviours, social–ethical literacy appears to shape the normative boundaries within which AI engagement is judged acceptable Across both strands, results converge on a dual-pathway pattern of learning in AI-augmented environments (Fig. 2 ). One pathway is technical–behavioural, in which technical AI literacy supports frequent and strategic AI-assisted learning behaviours. The other is cognitive–ethical, in which cognitive understanding and ethical awareness shape how learners assess, limit, and integrate AI support. These pathways do not represent fixed learner types; rather, they reflect situated responses to learning environments in which AI is readily available and unevenly structured (Ifenthaler et al., 2024 ; Zawacki-Richter et al., 2019 ). Implications for the Design of AI-Augmented Learning Environments Consistent with prior work highlighting fragmented AI adoption in higher education (Eden et al., 2024 ; Kajiwara & Kawabata, 2024 ), the findings underscore the importance of designing learning environments that structure how AI is encountered and used, rather than leaving engagement entirely to individual discretion (Casanova et al., 2020 ). First, learning environments should be designed to support both pathways of AI-mediated learning. The technical–behavioural pathway requires environments that enable productive AI-assisted learning behaviours, such as opportunities for iterative exploration, feedback, and revision. The cognitive–ethical pathway requires environments that foreground judgment, interpretation, and restraint. Instructional designs that explicitly differentiate behavioural engagement (how AI is used) from evaluative reasoning (how AI outputs are judged) better align with how AI literacy operates in practice (Walter, 2024 ). The instructional environments can support productive AI engagement by making learning processes visible. Pairing AI-supported tasks with reflective artifacts, such as process logs, source-checking exercises, or comparisons between AI-assisted and non-AI work, helps align AI use with learning objectives rather than efficiency alone (Chiu, 2025 ). Second, traditional output-based assessment becomes increasingly insufficient for evaluating student learning and critical thinking, particularly in less hands-on disciplines such as business. Prior research in learning environments shows that when AI tools externalize cognitive work, observable products may no longer reliably reflect learners’ underlying reasoning or conceptual understanding (Elshall & Badir, 2025 ; Minn, 2022 ). Our findings suggest that assessment must shift toward making learners’ judgment and reasoning visible, rather than relying solely on written outputs. Dialogic and process-oriented and student-cantered assessment practices, such as brief oral explanations, structured discussion, guided questioning, peer critique, and instructor–student conferences, have long been recognized as effective for supporting deep learning (Baeten et al., 2016 ). They become especially critical in AI-augmented learning environments, where written products alone provide limited insight into students’ thinking. By foregrounding reasoning in interaction, these approaches allow instructors to evaluate how students interpret, justify, and regulate AI use in real time, supporting metacognitive regulation and meaningful assessment without imposing excessive documentation demands. Limitation and Conclusion As in prior AI-in-education research, this study relies on self-reported measures, which may introduce response bias. The cross-sectional design also limits causal inference. Future research should employ longitudinal or experimental designs to examine how learner orientations evolve as AI becomes further embedded in learning environments, and how changes in instructional design influence engagement, judgment, and learning outcomes over time. As AI becomes a persistent feature of academic learning environments, understanding how students adapt to its presence is increasingly critical. This study demonstrates that learning gains in AI-augmented settings are shaped not by AI use alone, but by two complementary learning pathways through which AI literacy operates. The first is a technical–behavioural pathway, in which technical AI literacy enables students to engage in AI-assisted learning behaviours that support comprehension, practice, and revision, thereby indirectly contributing to learning gains. The second is a cognitive–ethical pathway, in which cognitive and social–ethical AI literacy support learning through evaluative judgment, guiding how students assess AI outputs, set boundaries on use, and integrate AI support responsibly. These pathways highlight that effective learning in AI-rich environments depends on both productive engagement with AI tools and reflective judgment about their use. For educators, the central challenge is therefore not whether students use AI, but how learning environments and assessment practices can be designed to surface reasoning, support judgment, and sustain learning quality as AI becomes an integral part of academic practice. Declarations A statement of ethics approval - The study was approved by Farmingdale State College IRB/ethics committee. Conflict of interest The author confirms that there are no conflicts of interest. References Acosta-Enriquez BG, Valle G, Arbulu Ballesteros MLA, Arbulu Castillo M, Perez Vargas JCA, Torres CG, Silva Leon IS, P. M., Saavedra Tirado K (2025) What is the influence of psychosocial factors on artificial intelligence appropriation in college students? BMC Psychol 13(1):7. https://www.ncbi.nlm.nih.gov/pubmed/39755638 Baeten M, Dochy F, Struyven K, Parmentier E, Vanderbruggen A (2016) Student-centred learning environments: an investigation into student teachers’ instructional preferences and approaches to learning. Learn Environ Res 19(1):43–62 Bewersdorff A, Hornberger M, Nerdel C, Schiff DS (2025) AI advocates and cautious critics: How AI attitudes, AI interest, use of AI, and AI literacy build university students' AI self-efficacy. Computers and Education: Artificial Intelligence , 8 Braun V, Clarke V (2006) Using thematic analysis in psychology. Qualitative Res Psychol 3(2):77–101 Braun V, Clarke V (2021) Thematic analysis: A practical guide Casanova D, Huet I, Garcia F, Pessoa T (2020) Role of technology in the design of learning environments. Learn Environ Res 23(3):413–427 Chang X, Wong GK (2025) A systematic review of how educators integrate ethics into artificial intelligence curriculum. J Res Technol Educ, 1–18 Chen Y, Zare S (2025) University students’ AI thinking: A Latent Profile Analysis and predictive effects of psychological factors. J Res Technol Educ, 1–14 Chiu M-L (2025) Exploring user awareness and perceived usefulness of generative AI in higher education: The moderating role of trust. Educ Inform Technol, 1–35 Chiu TK, Chen Y, Yau KW, Chai C-s, Meng H, King I, Wong S, Yam Y (2024) Developing and validating measures for AI literacy tests: From self-reported to objective measures. Computers Education: Artif Intell 7:100282 Chun Z, Ning Y, Chen J, Wijaya TT (2025) Exploring the interplay among artificial intelligence literacy, creativity, self-efficacy, and academic achievement in college students: Findings from PLS-SEM and FsQCA. Educ Inform Technol, 1–34 Crompton H, Burke D (2023) Artificial intelligence in higher education: the state of the field. Int J Educational Technol High Educ 20(1):22 Eden CA, Chisom ON, Adeniyi IS (2024) Integrating AI in education: Opportunities, challenges, and ethical considerations. Magna Scientia Adv Res Reviews 10(2):006–013 Eisenbardt M, Kurtz G, Gafni R, Leiba M, Dettmer S, Mullins R, Siegert I, Eisenbardt T, Gal E (2025) Students’ Readiness and Ethical Considerations to Adopt Generative AI in their learning: An International Comparative Study. Technol Knowl Learn, 1–28 Elshall AS, Badir A (2025) Balancing AI-assisted learning and traditional assessment: the FACT assessment in environmental data science education. Frontiers in Education Fetters MD, Curry LA, Creswell JW (2013) Achieving integration in mixed methods designs—principles and practices. Health Serv Res 48(6pt2):2134–2156 Grassini S (2023) Development and validation of the AI attitude scale (AIAS-4): a brief measure of general attitude toward artificial intelligence. Front Psychol 14:1191628 Hornberger M, Bewersdorff A, Nerdel C (2023) What do university students know about Artificial Intelligence? Development and validation of an AI literacy test. Computers Education: Artif Intell, 5 Ifenthaler D, Majumdar R, Gorissen P, Judge M, Mishra S, Raffaghelli J, Shimada A (2024) Artificial intelligence in education: Implications for policymakers, researchers, and practitioners. Technol Knowl Learn 29(4):1693–1710 Ji Y, Zhong M, Lyu S, Li T, Niu S, Zhan Z (2025) How does AI literacy affect individual innovative behavior: the mediating role of psychological need satisfaction, creative self-efficacy, and self-regulated learning. Educ Inform Technol 30(11):16133–16162 Kajiwara Y, Kawabata K (2024) AI literacy for ethical use of chatbot: Will students accept AI ethics? Computers Education: Artif Intell 6:100251 Kim D, Wang CK, Borowiec K, Rein N, Cho J-HA, Liu JJ (2025) Unlocking the Future: A Comprehensive Review of ChatGPT in Education. Technol Knowl Learn, 1–53 Kim J, Yu S, Lee S-S, Detrick R (2025) Students’ prompt patterns and its effects in AI-assisted academic writing: Focusing on students’ level of AI literacy. J Res Technol Educ, 1–18 Kong S-C, Cheung WM-Y, Zhang G (2022) Evaluating artificial intelligence literacy courses for fostering conceptual learning, literacy and empowerment in university students: Refocusing to conceptual building. Computers Hum Behav Rep 7:100223 Kosmyna N, Hauptmann E, Yuan YT, Situ J, Liao X-H, Beresnitzky AV, Braunstein I, Maes P (2025) Your brain on ChatGPT: Accumulation of cognitive debt when using an AI assistant for essay writing task Laupichler MC, Aster A, Schirch J, Raupach T (2022) Artificial intelligence literacy in higher and adult education: A scoping literature review. Computers Education: Artif Intell 3:100101 Long D, Magerko B (2020) What is AI Literacy? Competencies and Design Considerations Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems Liehner L, Hick G, Biermann A, Brauner H, P., Ziefle M (2023) Perceptions, attitudes and trust toward artificial intelligence — An assessment of the public opinion. Artificial Intelligence and Social Computing Medina-Gual L, Parejo J-L (2025) University Students’ Engagement with Artificial Intelligence: A Cluster Analysis of Learner Profiles in AI Literacy. Technol Knowl Learn, 1–19 Minn S (2022) AI-assisted knowledge assessment techniques for adaptive learning environments. Computers Education: Artif Intell 3:100050 Montali S (2025) Oct 7th, 2025). A Debate About A.I. Plays Out on the Subway Walls. New York Times . https://www.nytimes.com/2025/10/07/style/friend-ai-subway-ads-new-york.html Morales-García WC, Sairitupa-Sanchez LZ, Morales-García SB, Morales-García M (2024) Adaptation and psychometric properties of a brief version of the general self-efficacy scale for use with artificial intelligence (GSE-6AI) among university students. Frontiers in Education , 9 Ng DTK, Leung JKL, Chu SKW, Qiao MS (2021) Conceptualizing AI literacy: An exploratory review. Computers Education: Artif Intell 2:100041 Ng W (2012) Can we teach digital natives digital literacy? Comput Educ 59(3):1065–1078 Seymour E, Wiese D, Hunter A, Daffinrud SM (2000) Creating a better mousetrap: On-line student assessment of their learning gains. National Meeting of the American Chemical Society Shardlow M, Sellar S, Rousell D (2022) Collaborative augmentation and simplification of text (CoAST): Pedagogical applications of natural language processing in digital learning environments. Learn Environ Res 25(2):399–421 Walter Y (2024) Embracing the future of Artificial Intelligence in the classroom: the relevance of AI literacy, prompt engineering, and critical thinking in modern education. Int J Educational Technol High Educ 21(1):15 Younas A, Fàbregues S, Creswell JW (2023) Generating metainferences in mixed methods research: A worked example in convergent mixed methods designs. Methodological Innovations 16(3):276–291 Zawacki-Richter O, Marín VI, Bond M, Gouverneur F (2019) Systematic review of research on artificial intelligence applications in higher education – where are the educators? Int J Educational Technol High Educ, 16 (1) Tables Table 1 Interviewees Profiles ID Gender Major Grade level 1 M Sports Management Freshman 2 M Nursing Freshman 3 M Business Analytics Sophomore 4 M Architecture Sophomore 5 F Computer Science Junior 6 M Science, Technology Society Sophomore 7 M Mechanical Engineering Technology Sophomore 8 F Architecture Sophomore 9 F Biology Freshman 10 M Computer Science Sophomore Table 2 Student AI Engagement Framework: Dimensions, Definitions, and Illustrative Quotes Dimension Core Idea Illustrative Quotes Academic Enablement AI increases efficiency, comprehension, and confidence in coursework. “It helps you take notes better, study better—like reviewing the notes AI helped me generate. It's a really good study tool.” (Int. 1) “I can ask AI to summarize it. It gives me the main points quickly and saves a lot of time.” (Int.2) “It’s helped explain things when I asked in-depth questions. It’s given me context or extra knowledge I didn’t have before. It’s helped me learn more material, which in turn helps me complete more assignments.” (Int.7) “AI makes computer science more exciting. It adds tools, not threats.” (Int.10) AI Literacy as Competence Students emphasize prompting, verification, and critical thinking as core skills for effective use. “I think it means understanding how AI communicates and works—how it gives answers and what it’s capable of. It’s not just about using it but knowing how to interpret what it gives you.” (Int.5) “The biggest skill is knowing how to write a prompt.” (Int.5) “You have to be flexible and good at adapting. AI is always changing as new information comes out. You need to be open-minded and always look at what’s new.” (Int.7) “Using AI is surface-level. Literacy is about control and critical understanding.” (Int.10) Conditional Trust & Risk Management Usefulness is balanced with concerns about accuracy, plagiarism, privacy, and institutional penalties. “Sometimes AI gives wrong answers… so I double-check before submitting assignments.” (Int.1) “No. I always avoid [enter private information]—it’s learning from users, so there’s no guarantee your data stays private.” (Int.4) “Responsibility and balance. You have to know how to use AI responsibly and also develop your own skills.” (Int.5) “I believe misinformation is the biggest problem.” (Int.6) “My main concern is plagiarism because AI just pulls information from the Internet or its own databases.” (Int.7) Human–AI Boundaries Students view creativity, judgment, ethics, and hands-on learning as distinctly human strengths. “AI doesn’t understand human emotions or have emotional intelligence. I wouldn’t trust it to define me.” (Int.1) “I try not to rely on it as much because I know I need to remember that material in the future.” (Int.2) “I’ve always wanted to make sure I actually learn and retain the material for future applications…If I handed in something AI-generated, I wouldn’t know what I wrote and wouldn’t retain it. (Int.7) “Experimental design, ethical reasoning, and hands-on lab work need human input.” (Int.9) “Critical thinking, scientific literacy, and knowing when to rely on AI versus human expertise..” (Int.9) Pedagogical Guardrails Students support AI in education, but only with clear guidelines and as a supplement—not a replacement for instructors. “School should teach how to use AI responsibly.” (Int.3) “Schools would have to be very careful. If we can’t make it safe and credible, then I don’t think it should have a place in education.” (Int.7) “Yes, but only as a support tool—not to replace teachers.” (Int.8) “It should support learning, not replace human instruction.” (Int.9) Table 3 Correction Table of the main variables ( N=81) Mean S.D 1 2 3 4 5 1 Technical AI 3.88 0.86 1 2 Cognitive AI 3.98 0.77 .299** 1 3 Social-Ethical AI 4.23 0.68 0.047 .421** 1 4 AALB 3.36 1.05 .682** 0.198 -0.003 1 5 Learning Gain 3.42 1.16 .668** .339** 0.088 .813** 1 * p < .05, ** p < .01, *** p < .001. Table 4 Regression results for AI literacy, AI-Assisted Learning Behavior (AALB) and Learning Gains. Dependent variables: Model 1 & B = Learning Gain; Model A = Learning Behavior. Predictor Model 1: Learning Gain (no mediator) Model A: Learning Behavior Model B: Learning Gain (with mediator) Technical AI 0.81*** (0.13) 0.88*** (0.12) 0.18 (0.13) Cognitive AI 0.26 (0.16) −0.10 (0.15) 0.33** (0.12) Social-ethical AI −0.03 (0.17) −0.03 (0.16) −0.01 (0.13) AALB — — 0.72*** (0.10) Age (years) 0.01 (0.02) −0.00 (0.02) 0.02 (0.01) School: Business 0.19 (0.26) 0.21 (0.25) 0.04 (0.20) School: Engineering 0.32 (0.33) −0.10 (0.31) 0.40 (0.25) School: Health 0.86* (0.43) 0.38 (0.40) 0.58† (0.32) R² 0.47 0.46 0.71 Adj. R² 0.41 0.40 0.68 N 75 75 75 Notes: Entries are unstandardized regression coefficients with standard errors in parentheses. All models control for age and school (reference category not shown). Indirect effect for TechAI via AALB from bootstrap: 0.64, 95% CI [0.40, 0.91]. † p < .10, * p < .05, ** p < .01, *** p < .001. Additional Declarations The authors declare no competing interests. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9173238","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":609136462,"identity":"055f928d-0258-44bf-8018-f918c20ac760","order_by":0,"name":"Jing Betty Feng","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA+klEQVRIiWNgGAWjYBAC9gYgIVEhByQZgUw2BhmQqAQ+LTwHQCrOGMO18BCnhbHNGMolSgt77+EXlvMM5M2nHW57+KXMjoefgfngbR58WnjOpVlIbjMwnHM7sd1Y5lwyj2QDW7I1Pi32EjlmBpLb/jDOkE5sk5ZsY+YxOMBjJo3XFvk3QC1zDOyhWup57A/wf8OvRYLH+IFkg0EiSIvkx7bDPAYMPGz4tfDkmDFIHDNIBtvCcO44j8RhNmPLOfi0sJ8x/ixRY2A7Qzr9meSPsmo5/vbmhzfe4NECBGzSsGhgBruHGb9ysJKPH6Asxh+EVY+CUTAKRsEIBABs7UKOoYP9BQAAAABJRU5ErkJggg==","orcid":"https://orcid.org/0000-0001-8596-9766","institution":"Farmingdale State College","correspondingAuthor":true,"prefix":"","firstName":"Jing","middleName":"Betty","lastName":"Feng","suffix":""},{"id":609136463,"identity":"0f1efcd4-4629-49d7-beb9-74cd2ed83546","order_by":1,"name":"Gregorio Diaz","email":"","orcid":"","institution":"Farmingdale State College","correspondingAuthor":false,"prefix":"","firstName":"Gregorio","middleName":"","lastName":"Diaz","suffix":""},{"id":609136464,"identity":"83193c9b-e737-4eaa-be7a-a9444af0ac23","order_by":2,"name":"Zachary Long","email":"","orcid":"","institution":"Farmingdale State College","correspondingAuthor":false,"prefix":"","firstName":"Zachary","middleName":"","lastName":"Long","suffix":""}],"badges":[],"createdAt":"2026-03-19 22:47:37","currentVersionCode":1,"declarations":{"humanSubjects":true,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":true,"humanSubjectConsent":true,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-9173238/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9173238/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":105312341,"identity":"cc6c2801-daf4-4e57-9e2a-60359f5f0847","added_by":"auto","created_at":"2026-03-24 15:30:05","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":152515,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eCoding Scheme of Student Engagement with AI\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"Figure1AIEngageCoding.png","url":"https://assets-eu.researchsquare.com/files/rs-9173238/v1/bde0585bfe73e7f6e0422373.png"},{"id":105312342,"identity":"d6daccdb-28f0-466b-acc8-a4356d6d6c76","added_by":"auto","created_at":"2026-03-24 15:30:05","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":47589,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eThe Dual-Path Model of AI-Literacy on Learning Gains\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"Figure2DualPathofAILiteracy.png","url":"https://assets-eu.researchsquare.com/files/rs-9173238/v1/ae382be53c3918d9bfc2afd8.png"},{"id":105564348,"identity":"ca9eee8c-cfcc-46d4-a745-6a20d308d1a9","added_by":"auto","created_at":"2026-03-27 12:49:20","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1684238,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9173238/v1/99729f96-58b0-497e-a947-dfca5b4dab8f.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003e\u003cstrong\u003eLearning in AI-Augmented Environments Through Dual Pathways of AI Literacy\u003c/strong\u003e\u003c/p\u003e","fulltext":[{"header":"Introduction","content":"\u003cp\u003eThe rapid diffusion of generative artificial intelligence (GAI), including large language models (LLMs) such as ChatGPT, Grammarly, and AI-augmented search engines, is reshaping contemporary academic learning environments. Rather than functioning as isolated instructional tools, these systems increasingly operate as persistent features of students’ everyday academic contexts, influencing how information is accessed, interpreted, and produced across learning tasks (D. Kim et al., 2025; Shardlow et al., 2022). Throughout this article, we use the term ‘AI’ to refer specifically to these GAI tools. As AI becomes embedded within the instructional ecology of higher education, the central question has shifted from whether students should use AI to how AI-augmented learning environments shape students’ learning processes and orientations.\u003c/p\u003e\n\u003cp\u003eResearch offers mixed evidence regarding the educational consequences of AI-rich environments. While some highlight the benefits such as improving access to information, facilitating deeper engagement, and promoting creativity (Eden et al., 2024), others caution that overreliance may diminish critical thinking and potential ethical risks (Grassini, 2023). In particular, a 2025 study at MIT’s Media Lab raised concerning results that the usage of ChatGPT could harm learning and reduce critical thinking abilities due to passive behaviors (such as copy/paste)\u0026nbsp;(Kosmyna et al., 2025). These findings indicate that learning outcomes in AI-augmented environments depend not only on the presence of AI, but on how learners engage with it.\u003c/p\u003e\n\u003cp\u003eScholars have emphasized the importance of designing learning environments that support productive and responsible engagement with AI (Ifenthaler et al., 2024; Walter, 2024). AI literacy, the ability to understand, evaluate, and ethically engage with AI-generated content (Chang \u0026amp; Wong, 2025; Eisenbardt et al., 2025; Long \u0026amp; Magerko, 2020), has been emphasized as an emergent needs to ensure students learning outcomes and academic performance (Acosta-Enriquez et al., 2025). Extant literature has examined factors such as students’ perceptions, attitudes, trust, and perceived usefulness of AI (Bewersdorff et al., 2025; Hornberger et al., 2023; Long \u0026amp; Magerko, 2020). However, because AI literacy encompasses technical, cognitive, and social-ethical dimensions (Ng, 2012), it remains unclear which dimensions are empirically associated with students’ learning behaviors. In particular, AI-assisted learning behaviors (AALB), how students strategically incorporate AI into academic work, remain underexamined in relation to specific dimension of AI literacy (Elshall \u0026amp; Badir, 2025).\u003c/p\u003e\n\u003cp\u003eThis study aims to answer the research question: \u003cem\u003eHow does AI literacy shape learning processes and outcomes within AI-augmented learning environments?\u0026nbsp;\u003c/em\u003eWe employ a convergent mixed-methods (QUAL + QUAN) design (Fetters et al., 2013), integrating semi-structured interviews with a quantitative survey of college students. This approach enables examination of both learning outcomes and the meaning-making processes through which students interpret and navigate AI-supported environments.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eQualitative findings identify key dimensions of student engagement, including \u003cem\u003eacademic enablement, AI literacy as competence, conditional trust and risk management, human–AI boundaries,\u0026nbsp;\u003c/em\u003eand\u003cem\u003e\u0026nbsp;pedagogical guardrails\u003c/em\u003e. The quantitative results converge with and extend these insights by revealing two complementary pathways through which learning unfolds in AI-augmented environments. A technical–behavioral pathway, supported by Technical AI literacy, promotes AI-assisted learning behaviors and indirectly enhances perceived learning gains. A cognitive–ethical pathway, supported by Cognitive AI literacy, directly contributes to reflective judgment and perceived learning gains, alongside Social–ethical AI literacy. Notably, Social–ethical AI literacy shows no direct association with AI-assisted learning behaviors or performance outcomes. These findings suggest that AI-augmented learning environments do not produce uniform effects but instead support distinct pathways of engagement shaped by how learners interpret and regulate AI use within environmental conditions. This study offers a pedagogical framework for understanding how learning environments can be designed to support both procedural fluency in AI use and reflective, ethical engagement with AI-generated knowledge.\u003c/p\u003e"},{"header":"Literature Review","content":"\u003ch3\u003e\u003cstrong\u003eAI\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;Literacy: Definition and Dimensions\u003c/strong\u003e\u003c/h3\u003e\n\u003cp\u003eAs AI becomes a persistent feature of academic learning environments rather than an optional tool, AI literacy increasingly functions as a capacity for navigating, interpreting, and regulating AI-augmented learning conditions (D. Kim et al., 2025). AI literacy is defined as “the ability to understand, use, monitor, and critically reflect on AI applications without necessarily being able to develop AI models themselves” (Laupichler et al., 2022). More broadly, AI literacy encompasses a multidimensional set of competencies that enables individuals to critically appraise AI technologies, interact productively with AI systems, and apply them effectively across academic, professional, and everyday contexts (Long \u0026amp; Magerko, 2020).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eBuilding on Ng’s (2012) framework for digital literacy, AI literacy comprises three interrelated dimensions: technical, cognitive, and social-emotional.\u0026nbsp;\u003cem\u003eTechnical AI literacy\u003c/em\u003e refers to the ability of locating\u0026nbsp;and using AI tools effectively in a given discipline and instructional contexts (Ng, 2012).\u0026nbsp;Technical literacy shapes learners’ ability to engage with AI as an available environmental resource, including understanding their functionalities, distinguishing between academic and creative tasks, and selecting appropriate forms of AI support\u0026nbsp;(Ng, 2012). This foundational capacity is critical for enabling students to participate meaningfully in AI-supported learning activities.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eCognitive AI literacy\u003c/em\u003e refers to the awareness and ability to engage with AI as a collaborative and critically evaluating AI outputs for credibility, authority, and appropriateness (Chiu et al., 2024; Ng et al., 2021). Students must demonstrate and apply critical thinking and problem-solving skills when interacting with AI (Bewersdorff et al., 2025; Ng et al., 2021), including assessing reliability, recognizing bias, and determining when AI outputs are pedagogically appropriate. Cognitive AI literacy supports learners’ ability to interpret and evaluate information generated by AI, allowing learners to balance reliance on AI-generated information with reflective judgment and critical evaluation (Chen \u0026amp; Zare, 2025).\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eSocial-Ethical AI Literacy\u003c/em\u003e refers to the understanding of ethical implications and societal impacts of AI (Eisenbardt et al., 2025; Kajiwara \u0026amp; Kawabata, 2024). Socio-ethical literacy shapes how learners establish boundaries between human and algorithmic agency, including concerns related to transparency, data privacy, academic integrity, and the credibility of AI-generated knowledge (Acosta-Enriquez et al., 2025; Chang \u0026amp; Wong, 2025). \u0026nbsp;While AI have shown to enhance students’ confidence, motivation, and engagement (Ji et al., 2025; Luca Liehner et al., 2023), they also raise concerns about responsible use and epistemic trust. Recent debates surrounding wearable and relational AI systems further underscore how AI is reshaping not only information practices but also learners’ relationships with knowledge and authority (Montali, 2025). Consequently, fostering AI literacy within academic environments requires attention to ethical responsibility, critical awareness, and institutional guidance (Crompton \u0026amp; Burke, 2023; Eden et al., 2024).\u003c/p\u003e\n\u003cp\u003eDespite the growing scholarly attention to AI in education, notable gaps persist in understanding \u003cem\u003ehow\u003c/em\u003e each dimension of AI literacy translates into meaningful learning outcomes Much of the existing literature emphasizes psychological or pedagogical correlates such as self-efficacy, creativity, and satisfaction, while the environmentally situated cognitive and socio-ethical processes through which learners engage with AI remain underexplored (Chiu, 2025; Chun et al., 2025; Zawacki-Richter et al., 2019). Addressing these limitations requires moving beyond general accounts of AI literacy to examine how its distinct dimensions shape learners’ orientations and behaviors within AI-augmented learning environments, particularly in relation to critical thinking and perceived learning outcomes (Chiu, 2025; J. Kim et al., 2025; Walter, 2024).\u003c/p\u003e\n\u003ch3\u003eIntegration of AI in Academic Learning\u003c/h3\u003e\n\u003cp\u003eThe integration of AI into higher education has transformed students’ learning experiences, influencing academic performance, cognitive processes, and decision-making strategies (Grassini, 2023; Medina-Gual \u0026amp; Parejo, 2025). AI tools are increasingly used for writing, summarization, and content creation (Eden et al., 2024; Long \u0026amp; Magerko, 2020). As AI becomes environmentally embedded rather than episodically used, its influence extends beyond task completion to shaping how learners approach academic work more broadly.\u003c/p\u003e\n\u003cp\u003eWhile AI applications can enhance efficiency and productivity, improper or unreflective engagement, such as uncritical copy-pasting or overreliance on generated outputs, may undermine deep learning and critical thinking (Kosmyna et al., 2025). This distinction between \u003cem\u003eAI use\u003c/em\u003e and \u003cem\u003eAI literacy\u003c/em\u003e is crucial in learning environments research: learners may operate in AI-augmented settings without possessing the evaluative or integrative capacities necessary to engage productively with AI-generated knowledge (Crompton \u0026amp; Burke, 2023; Ng et al., 2021).\u003c/p\u003e\n\u003cp\u003eEmpirical studies suggest that well-designed AI can enhance learners’ confidence, autonomy, and motivation, reinforcing cognitive engagement and problem-solving (Bewersdorff et al., 2025; Morales-García et al., 2024). AI-supported learning has also been associated with increased self-efficacy, creativity, and satisfaction among students and educators (Ji et al., 2025; Kong et al., 2022). These findings position AI not simply as a digital assistant, but as a feature of the learning environment that can function as a cognitive partner, depending on how learners interpret and engage with its affordances.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAcademic performance, however, captures only a partial dimension of learning. Grade point average (GPA) primarily reflects assessment outcomes rather than depth of understanding, critical thinking, or transfer of knowledge (Grassini, 2023). In contrast, \u003cem\u003eperceived learning gains\u003c/em\u003e reflect students’ self-assessment of their intellectual growth, skill development, and ability to apply concepts meaningfully (Seymour et al., 2000). From a learning environments perspective, these indicators may diverge, as environmental conditions that shape student behavior can support performance-oriented outcomes without necessarily fostering deeper learning. Accordingly, perceived learning gains provide a complementary lens for examining how AI literacy shapes learning quality and longer-term educational development. Based on prior research on AI literacy, we propose the following hypotheses:\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eH1.\u003c/em\u003e\u003c/strong\u003e\u003cem\u003e\u0026nbsp;Each dimension of AI literacy – technical, cognitive, and ethical - is positively associated with perceived learning gains.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAI-Assisted Learning Behaviors as Mediators\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWhile prior research demonstrates that digital and AI literacy are positively associated with academic performance (Hornberger et al., 2023), empirical evidence explaining \u003cem\u003ehow\u003c/em\u003e these competencies translate into learning gains remains limited (Walter, 2024). Within the AI-augmented learning environment, this gap suggests that literacy alone does not determine learning outcomes; rather, its effects depend on how learners enact AI-supported behaviors within specific contexts (J. Kim et al., 2025).\u003c/p\u003e\n\u003cp\u003eLearning behaviors have long been recognized as critical determinants of academic success (Seymour et al., 2000). Within the context of AI, AALB may refer to students’ active and strategic engagement with AI tools to support their learning, include using AI for concept clarification, generating examples, revising drafts, conducting self-testing, and exploring ideas iteratively. These practices extend beyond passive tool use, reflecting the degree to which students leverage AI to enhance comprehension, critical thinking, and metacognitive awareness.\u003c/p\u003e\n\u003cp\u003eResearch indicates that AI-supported environments can foster confidence and academic engagement by helping students complete assignments, clarify complex concepts, and sustain motivation (Grassini, 2023). However, these benefits depend heavily on \u003cem\u003ehow\u003c/em\u003e students engage with AI: Learners with limited AI literacy may adopt surface-level or dependent practices, whereas learners with higher literacy demonstrate more reflective and self-regulated engagement (Luca Liehner et al., 2023). Exposure to AI technologies may promote trust and optimism, but productive learning outcomes require ethical reflection and adaptive use grounded in AI literacy (D. Kim et al., 2025; J. Kim et al., 2025).\u003c/p\u003e\n\u003cp\u003eAccordingly, this study proposes that AALB represents a mechanism through which AI literacy enhances learning outcomes through a mediation relationship between AI literacy and learning gains\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eH2.\u003c/em\u003e\u003c/strong\u003e\u003cem\u003e\u0026nbsp;AI-assisted learning behaviors mediate the relationship between each dimension of AI literacy and learning gains.\u003c/em\u003e\u003c/p\u003e"},{"header":"Methodology","content":"\u003ch3\u003e\u003cstrong\u003eResearch Design\u003c/strong\u003e\u003c/h3\u003e\n\u003cp\u003eTo examine how AI literacy shapes learning processes and outcomes within AI-augmented learning environments, this study employed a \u003cstrong\u003econvergent mixed-methods design\u003c/strong\u003e (QUAL + QUANT), in which qualitative and quantitative data were collected in parallel, analyzed separately, and integrated during interpretation(Fetters et al., 2013). This design allows for complementary examination of learning outcomes and the meaning-making processes through which students engage with AI-supported environments. Both qualitative and quantitative components were given equal analytic weight, enabling a balanced interpretation of learner experiences and outcome patterns (Younas et al., 2023).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eRather than treating AI use as a discrete intervention, the mixed-methods approach was selected to capture how students interpret, navigate, and respond to the future of AI environment. Quantitative analyses identify relationships among AI literacy, AI-assisted learning behaviors, and learning outcomes, while qualitative analyses provide contextual insight into how these relationships are enacted in practice. Integrating both strands supports a more comprehensive understanding of how AI literacy operates within contemporary academic learning environments. Integrating both strands supports a more comprehensive understanding of how AI literacy operates within contemporary academic learning environments (Fetters et al., 2013; Younas et al., 2023).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eQualitative Data Collection and Participants\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eDuring the summer of 2025, we conducted 10 semi-structured interviews with students at a regional college in the Northeastern United States to explore their engagement with AI. The purpose of the interviews was to explore how students experience and interpret AI as part of their academic learning environment. Two trained student research assistants conducted the interviews using a common interview protocol designed to elicit students’ awareness of AI, their rationales for using AI tools in academic and personal contexts, their understanding of AI literacy, and the perceived influence of AI on their learning and future academic or professional trajectories. Participants represented a range of academic majors and grade levels, providing variation in disciplinary context and learning experience. Interviews lasted between 20 and 50 minutes, were audio recorded with participants’ consent, and subsequently transcribed verbatim for analysis. Table 1 presents demographic and academic characteristics of the interview participants.\u003c/p\u003e\n\u003cp\u003e--------------------------------\u003c/p\u003e\n\u003cp\u003eInsert Table 1 About Here\u003c/p\u003e\n\u003cp\u003e--------------------------------\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eQualitative Data Analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe qualitative phase employed Braun and Clarke’s (2006, 2021) inductive thematic analysis to examine patterns in students’ experiences with AI in academic learning contexts. During the initial open-coding phase, the research team conducted inductive first-cycle coding to identify recurring patterns in participants’ accounts of AI use in academic contexts. Coding focused on capturing concrete, descriptive features of students’ engagement with AI, including primary tools used, academic applications, efficiency-oriented practices, learning simplification strategies, concerns about detection, and perceived risks.\u003c/p\u003e\n\u003cp\u003eAn initial codebook was developed from the first interview transcript and iteratively refined as additional transcripts were analyzed. As coding progressed, new codes were added when novel themes emerged, and existing codes were refined, merged, or split to ensure conceptual clarity and alignment with the data. Throughout this process, the research team reviewed all coding decisions, verified illustrative quotations against transcripts, and resolved discrepancies through discussion. This iterative procedure resulted in a finalized set of first-level descriptive codes.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIn the second analytic phase, first-level codes were grouped into axial themes that reflected broader patterns of meaning across participants’ accounts. These themes captured students’ interpretations of AI as an academic resource, their navigation of accessibility and ease of use, their developing understanding of AI literacy, and their reflections on career preparation and future competencies. Finally, axial themes were synthesized into five aggregated dimensions that constitute the Student AI Engagement framework: \u003cem\u003eAcademic Enablement, AI Literacy as Competence, Conditional Trust and Risk Management, Human–AI Boundaries,\u003c/em\u003e and \u003cem\u003ePedagogical Guardrails\u003c/em\u003e.\u003c/p\u003e\n\u003cp\u003eThis structured analytic process enabled interpretation of how students orient themselves within AI-augmented learning environments, capturing both enabling and constraining aspects of AI use. The resulting qualitative insights provide contextual grounding for the quantitative findings and support integrated interpretation of learner orientations and learning outcomes. Figure 1 presents an overview of the qualitative coding structure.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e--------------------------------\u003c/p\u003e\n\u003cp\u003eInsert Figure 1 About Here\u003c/p\u003e\n\u003cp\u003e--------------------------------\u003c/p\u003e\n\u003ch3\u003e\u003cstrong\u003eQuantitative Data Collection\u003c/strong\u003e\u003c/h3\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eParticipants\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe quantitative survey was conducted in parallel with the interviews and included undergraduate students from multiple disciplines. Participants were recruited through campus announcements and course mailing lists over a period of three weeks. Eligibility requires full-time enrollment and informed consent. We received a total of 146 responses. After removing incomplete responses and those who failed the attention check, the final sample comprised 81 valid participants (man = 53, woman = 44, non-binary = 1, a gender not listed = 1, prefer not to say = 5), with an average age of 23.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eMeasures and Operational Definitions\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eData were collected using a structured survey instrument based on validated scales and adapted items from prior research. All constructs were measured using a 5-point Likert scale (1 = Strongly Disagree, 5 = Strongly Agree) unless otherwise noted.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAdapted from\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003eNg (2012)\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;digital literacy measurement, AI literacy was operationalized as a multidimensional construct comprising technical, cognitive, and social–ethical dimensions, reflecting learners’ capacity to navigate and regulate engagement within AI-augmented learning environments\u003c/strong\u003e.\u0026nbsp;\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003e\u003cstrong\u003eTechnical AI literacy\u003c/strong\u003e (4 items; Cronbach’s α = .75) captures learners’ ability to locate and use AI tools effectively for information access and academic support. Sample items include \u003cem\u003e“I can identify AI tools that are relevant for locating information in my field of study or work”\u003c/em\u003e and \u003cem\u003e“I use AI tools to find information more efficiently than traditional search engines.”\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cstrong\u003eCognitive AI literacy (5 items; Cronbach’s α = .74) reflects learners’ ability to critically evaluate AI-generated outputs for credibility, bias, and appropriateness, and to integrate AI support into their own thinking. Sample items include\u003cem\u003e\u0026nbsp;“I assess the credibility of AI-generated outputs by verifying them against trustworthy sources”\u0026nbsp;\u003c/em\u003eand \u003cem\u003e“I apply AI outputs in ways that support and extend my own thinking rather than replace it.”\u003c/em\u003e\u003c/strong\u003e\u003c/li\u003e\n \u003cli\u003e\u003cstrong\u003eSocial-Ethical AI literacy (5 items;\u0026nbsp;\u003c/strong\u003eCronbach’s α =\u003cstrong\u003e\u0026nbsp;.59)\u003c/strong\u003e addresses learners’ awareness of ethical and responsible AI use, including transparency, academic integrity, and ethical implications of AI-supported work. Sample items include \u003cem\u003e“I disclose when\u0026nbsp;\u003c/em\u003e\u003cem\u003eI have used AI tools in my\u003c/em\u003e\u003cem\u003e\u0026nbsp;assignments or creative work”\u003c/em\u003e and \u003cem\u003e“I understand\u0026nbsp;\u003c/em\u003e\u003cem\u003ethe ethical implications\u003c/em\u003e\u003cem\u003e\u0026nbsp;of using AI in communication or\u0026nbsp;\u003c/em\u003e\u003cem\u003edecision-making.”\u0026nbsp;\u003c/em\u003e The lower internal consistency reflects it diverse aspects of social-ethical AI awareness, however, is acceptable for a preliminary exploratory study to cover multiple dimensions rather than achieve high internal consistency.\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003e\u003cstrong\u003eAI-assisted learning behaviors\u003c/strong\u003e (AALB) (6 items; Cronbach’s α = .91) represent learners’ active and strategic use of AI to support academic learning within AI-augmented environments. Items assess behaviors such as concept clarification, example generation, revision, and self-testing (e.g., \u003cem\u003e“I use AI tools to explain concepts I do not fully understand,” “I revise my academic work based on feedback from AI tools,”\u003c/em\u003e and \u003cem\u003e“I use AI tools to practice or quiz myself.”\u003c/em\u003e).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePerceived learning gains\u003c/strong\u003e (6 items, Cronbach’s α =.93), adapted from Seymour et al. (2000), capture learners’ self-reported intellectual growth, understanding, and motivation associated with AI-supported learning. (e.g., \u003cem\u003eGained a deeper understanding of concepts by exploring explanations or examples generated by AI tools,”\u003c/em\u003e and \u003cem\u003e“Increased my motivation to learn by making study tasks more engaging with AI tools.”\u003c/em\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAdditional variables included demographic information (age, gender, year of study, major) and descriptive indicators of AI experience, such as frequency of AI use for coursework and types of tools used (e.g., ChatGPT, Grammarly, NotebookLM).\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eQuantitative Data Analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eDescriptive statistics were used to summarize participant characteristics and scale scores for AI literacy dimensions, AI-assisted learning behaviors, and perceived learning gains. Internal consistency of multi-item scales was assessed using Cronbach’s alpha. Zero-order correlations were examined to assess associations among study variables.\u003c/p\u003e\n\u003cp\u003eTo evaluate the proposed relationships, multiple regression analyses were conducted, including mediation models examining whether AI-assisted learning behaviors explained associations between AI literacy dimensions and perceived learning gains. Indirect effects were tested using non-parametric bootstrapping with 5,000 resamples, allowing for robust estimation of mediation effects without assuming normality of the sampling distribution.\u003c/p\u003e"},{"header":"Findings","content":"\u003cp\u003e\u003cstrong\u003eQualitative Findings\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe qualitative findings reveal how students interpret and respond to AI learning environments in their academic work. Analysis of interview data yielded five aggregated dimensions of students’ orientations toward AI within their learning environments. These dimensions illustrate how students leverage AI to support academic work while actively regulating its use to maintain learning quality, academic integrity, and personal agency. Table 2 summarizes the definition of each dimension and selected illustrative quotations.\u003c/p\u003e\n\u003cp\u003e---------------------------------\u003c/p\u003e\n\u003cp\u003eInsert Table 2 About Here\u003c/p\u003e\n\u003cp\u003e----------------------------------\u003c/p\u003e\n\u003ch2\u003e\u003cem\u003eAcademic Enablement\u003c/em\u003e\u003c/h2\u003e\n\u003cp\u003eStudents consistently described AI as an enabling feature of their academic learning environment, particularly in relation to efficiency, task support, and cognitive load management. AI support was commonly used for note-taking, summarizing readings, clarifying concepts, debugging code, and organizing assignments. One participant explained, \u003cem\u003e“I use this AI that records the professor and translates the lecture into notes. I use those notes to answer my questions, so I understand the lecture better”\u003c/em\u003e (Int.1). Others emphasized time savings and simplification: \u003cem\u003e“They speed up my academic work, and they’re good at breaking things down into simpler terms”\u003c/em\u003e (Int.2).\u003c/p\u003e\n\u003cp\u003eAI use was also normalized through peer practices, with students describing AI as a routine academic resource across disciplines. As one student noted, \u003cem\u003e“Especially in writing-heavy classes. Even bio majors use it more now”\u003c/em\u003e (Int.9). These accounts highlight how AI functions as an enabling environmental resource that supports students’ day-to-day academic engagement.\u003c/p\u003e\n\u003ch2\u003e\u003cem\u003eAI Literacy as Competence\u003c/em\u003e\u003c/h2\u003e\n\u003cp\u003eStudents conceptualized AI literacy as a multidimensional competence extending beyond basic tool use. Rather than equating literacy with familiarity alone, participants described AI literacy as a combination of technical skill, evaluative judgment, and future-oriented preparedness. One student summarized this view: \u003cem\u003e“AI literacy means knowing how AI models work, how to use them responsibly, and how to evaluate their output”\u003c/em\u003e (Int.10).\u003c/p\u003e\n\u003cp\u003eStudents frequently emphasized prompting, iterative refinement, and verification as central competencies. As one participant stated, \u003cem\u003e“The biggest skill is knowing how to write a prompt. That’s the most important thing”\u003c/em\u003e (Int.5). Many also linked AI literacy to anticipated professional expectations, framing it as preparation for evolving workplace environments:\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cem\u003e“AI is already showing up in search engines, browsers, and even places like fast food drive-throughs. It’s becoming normal, and being familiar with it will be a big advantage”\u003c/em\u003e (Int.2).\u003c/p\u003e\n\u003cp\u003eTogether, these accounts position AI literacy as an adaptive competence that supports effective engagement to align with AI learning environments, rather than as simple technical proficiency.\u003c/p\u003e\n\u003ch2\u003e\u003cem\u003eConditional Trust \u0026amp; Risk Management\u003c/em\u003e\u003c/h2\u003e\n\u003cp\u003eDespite frequent AI use, students described exercising selective and conditional trust. Participants articulated ongoing evaluations of accuracy, appropriateness, and risk, particularly in relation to plagiarism, misinformation, privacy, and institutional consequences. One student explained, \u003cem\u003e“For general or common information, I usually trust it. But for things like history, people’s beliefs, or anything for a research paper, I double-check it before trusting it”\u003c/em\u003e (Int.3).\u003c/p\u003e\n\u003cp\u003eEthical considerations also shaped usage boundaries. As one participant noted, \u003cem\u003e“I don’t use it for cheating or anything like that, it doesn’t help me in that situation”\u003c/em\u003e (Int.6). Concerns about privacy (Int.4), misinformation (Int.6), and plagiarism (Int.7) were frequently raised. These accounts suggest that trust in AI is actively regulated, with students continuously assessing when AI support is appropriate and when independent verification is necessary.\u003c/p\u003e\n\u003ch2\u003e\u003cem\u003eHuman–AI Boundaries\u003c/em\u003e\u003c/h2\u003e\n\u003cp\u003eStudents articulated clear boundaries between tasks they considered appropriate for AI support and those they viewed as requiring human judgment, creativity, or experiential knowledge. While AI was seen as effective for generating structured content and explanations, participants emphasized its limitations in interpretive, creative, or values-based domains. As one student observed, \u003cem\u003e“AI can generate designs, but those lack real human concepts and meaning”\u003c/em\u003e (Int.4).\u003c/p\u003e\n\u003cp\u003eConcerns about overreliance were also common. One participant explained, \u003cem\u003e“I try not to rely on it as much because I know I need to remember that material in the future”\u003c/em\u003e (Int.2). Students further noted that perceived boundaries varied by discipline, with AI seen as more disruptive in fields such as computer science or English than in hands-on domains like nursing (Int.1; Int.2).\u0026nbsp;\u003c/p\u003e\n\u003ch2\u003e\u003cem\u003ePedagogical Guardrails\u003c/em\u003e\u003c/h2\u003e\n\u003cp\u003eStudents generally supported the inclusion of AI within higher education but emphasized the importance of clear pedagogical boundaries governing its use. Participants consistently framed AI as a support for learning rather than a replacement for instruction. As one student stated, \u003cem\u003e“Yes, but only as a support tool, not to replace teachers”\u003c/em\u003e (Int.8).\u003c/p\u003e\n\u003cp\u003eStudents also highlighted the need for institutional guardrails when integrating AI in education, including clear policies, ethical guidance, and equitable access. As one participant noted, \u003cem\u003e“Schools would have to be very careful. If we can’t make it safe and credible, then I don’t think it should have a place in education”\u003c/em\u003e (Int.7). These perspectives reflect a preference for AI learning environments that balance innovation with instructional integrity, transparency, and fairness, while preserving the central role of human instruction.\u003c/p\u003e\n\u003cp\u003eOverall, the qualitative findings portray students as pragmatic and reflective participants in AI-augmented learning environments. While students actively leverage AI for academic enablement, they simultaneously engage in self-regulation to preserve learning quality, ethical standards, and human expertise. These findings illuminate the conditions under which AI supports or constrains learning and provide contextual grounding for the quantitative results.\u0026nbsp;\u003c/p\u003e\n\u003ch2\u003e\u003cstrong\u003eQuantitative Finding\u003c/strong\u003e\u003c/h2\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eAssociations Among AI Literacy, Learning Behaviors, and Learning Gains\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTable 3 presents descriptive statistics and zero-order correlations among the study variables. Technical and Cognitive AI literacy were moderately correlated (r = .30, p \u0026lt; .01), supporting their treatment as related but empirically distinct dimensions. Technical AI literacy was positively associated with AALB and perceived learning gains, while AALB showed a strong positive association with learning gains. In contrast, Cognitive AI literacy was modestly correlated with learning gains but showed no meaningful association with AALB. Social–ethical AI literacy exhibited weak, non-significant correlations with the other constructs.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e-----------------------------------\u003c/p\u003e\n\u003cp\u003eInsert Table 3 About Here\u003c/p\u003e\n\u003cp\u003e------------------------------------\u003c/p\u003e\n\u003cp\u003eWe first examined whether AI literacy dimensions predicted perceived learning gains while controlling for age and school (Table 4). In the baseline regression model (Model 1), Technical AI literacy showed a strong positive association with learning gains (b = 0.81, SE = 0.13, t = 6.25, p \u0026lt; .001). Cognitive AI literacy was positively but not significantly related to learning gains (b = 0.26, SE = 0.16, t = 1.66, p = .10), and Social–ethical AI literacy was not associated with learning gains (b = −0.03, SE = 0.17, t = −0.19, p = .85). Students in the health school reported slightly higher learning gains than those in the reference school (b = 0.86, SE = 0.43, t = 2.00, p = .050), while age and other school indicators were not significant. The model explained 46.8% of the variance in learning gains (R² = .47; adj. R² = .41).\u003c/p\u003e\n\u003cp\u003eThese results indicate that, at the bivariate outcome level, technical aspects of AI literacy are most strongly associated with perceived learning gains.\u003c/p\u003e\n\u003cp\u003e-----------------------------------\u003c/p\u003e\n\u003cp\u003eInsert Table 4 About Here\u003c/p\u003e\n\u003cp\u003e------------------------------------\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eAALB as a Mediating Process\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe next examined whether AALB explained how AI literacy relates to learning gains (Table 4). In Model A, AALB was regressed on the three AI literacy dimensions and controls. Technical AI literacy was a strong predictor of AALB (b = 0.88, SE = 0.12, t = 7.18, p \u0026lt; .001), whereas Cognitive (b = −0.10, SE = 0.15, p = .51) and Social–ethical AI literacy (b = −0.03, SE = 0.16, p = .84) were not significant predictors. This model explained 46% of the variance in AALB (R² = .46; adj. R² = .40).\u003c/p\u003e\n\u003cp\u003eIn Model B, learning gains were regressed on AALB, the three literacy dimensions, and controls. AALB emerged as a strong predictor of learning gains (b = 0.72, SE = 0.10, t = 7.52, p \u0026lt; .001). After accounting for AALB, the direct effect of Technical AI literacy was substantially reduced and no longer significant (b = 0.18, SE = 0.13, p = .17), consistent with an indirect-only mediation pattern. In contrast, Cognitive AI literacy showed a significant direct association with learning gains (b = 0.33, SE = 0.12, t = 2.86, p = .006), indicating an effect that operates independently of AI-assisted learning behaviors. Social–ethical AI literacy remained non-significant. Model B explained 71.3% of the variance in learning gains (R² = .71; adj. R² = .68).\u003c/p\u003e\n\u003cp\u003eTo formally test indirect effects, we conducted a non-parametric bootstrap with 5,000 resamples. Bootstrapped mediation analyses confirmed a significant indirect effect of Technical AI literacy on learning gains through AALB (b = 0.64, 95% CI [0.40, 0.91]). Indirect effects for Cognitive (b = −0.07, 95% CI [−0.32, 0.16]) and Social–ethical AI literacy (b = −0.03, 95% CI [−0.28, 0.22]) were not significant.\u003c/p\u003e\n\u003cp\u003eAs figure 2 illustrates, these results indicate two distinct pathways linking AI literacy to learning gains. Technical AI literacy primarily supports learning by increasing students’ engagement in AALB, whereas Cognitive AI literacy contributes directly to learning gains, independent of behavior frequency. In unstandardized terms, approximately 78% of the total effect of Technical AI literacy on learning gains operates indirectly through AALB, while the effect of Cognitive AI literacy is primarily direct.\u003c/p\u003e\n\u003cp\u003e--------------------------------\u003c/p\u003e\n\u003cp\u003eInsert Figure 2 About Here\u003c/p\u003e\n\u003cp\u003e--------------------------------\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eThis mixed-methods study examined how students engage with AI within AI-augmented learning environments, focusing on how AI literacy relates to learning gains through situated learning behaviours. By integrating qualitative and quantitative findings (Younas et al., \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2023\u003c/span\u003e), the study clarifies that AI literacy does not operate as a uniform advantage, but through distinct pathways that shape how learners engage with and regulate AI within specific learning conditions. The qualitative dimensions describe how students leverage AI to support learning while actively regulating its use to preserve learning quality, integrity, and agency. The quantitative findings reinforce this interpretation by showing that learning gains are associated with behavioural engagement and evaluative judgment, rather than AI use alone (Chiu, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Ng, \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2012\u003c/span\u003e).\u003c/p\u003e\n\u003ch3\u003eInterpretation of Findings in AI-Augmented Learning Environments\u003c/h3\u003e\n\u003cp\u003eQuantitative results show that Technical AI literacy is strongly associated with AALB, which in turn account for a substantial proportion of perceived learning gains. When learning behaviours are considered, the direct association between Technical AI literacy and learning gains diminishes, indicating that learning gains emerge primarily through how students act within AI-supported environments, rather than from technical skill alone. This finding aligns with prior research emphasizing the role of strategic use and self-regulated learning behaviours in technology-enhanced learning (Ng, \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2012\u003c/span\u003e; Walter, \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eQualitative findings contextualize this pattern. Students described routinized prompting, iterative refinement, and verification practices that enabled AI to support comprehension and task completion. These practices illustrate how technical literacy functions as an enabling condition, allowing learners to take advantage of AI features embedded in the learning environment without defaulting to passive use (Chiu, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Ng et al., \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2021\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eIn contrast, Cognitive AI literacy exhibits a different pattern. Quantitatively, it predicts learning gains directly but is not associated with increased AALB. Qualitative accounts suggest that cognitive literacy shapes evaluative judgment, particularly in tasks involving interpretation, originality, or high academic stakes. Students described cross-checking AI outputs, comparing them with course materials, and deliberately limiting AI use when it threatened learning quality or integrity. This evaluative role is consistent with prior work emphasizing critical judgment and metacognitive regulation in AI-supported learning environments (Bewersdorff et al., \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Eden et al., \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eSocial\u0026ndash;ethical AI literacy did not directly predict learning gains or AALB, but qualitative evidence indicates that it plays a regulatory role. Students frequently referenced concerns about plagiarism, privacy, misinformation, and institutional penalties, echoing broader debates in the literature on ethical governance and trust in educational AI (Crompton \u0026amp; Burke, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Kajiwara \u0026amp; Kawabata, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Rather than generating learning behaviours, social\u0026ndash;ethical literacy appears to shape the normative boundaries within which AI engagement is judged acceptable\u003c/p\u003e \u003cp\u003eAcross both strands, results converge on a dual-pathway pattern of learning in AI-augmented environments (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). One pathway is technical\u0026ndash;behavioural, in which technical AI literacy supports frequent and strategic AI-assisted learning behaviours. The other is cognitive\u0026ndash;ethical, in which cognitive understanding and ethical awareness shape how learners assess, limit, and integrate AI support. These pathways do not represent fixed learner types; rather, they reflect situated responses to learning environments in which AI is readily available and unevenly structured (Ifenthaler et al., \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Zawacki-Richter et al., \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2019\u003c/span\u003e).\u003c/p\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003eImplications for the Design of AI-Augmented Learning Environments\u003c/h2\u003e \u003cp\u003eConsistent with prior work highlighting fragmented AI adoption in higher education (Eden et al., \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Kajiwara \u0026amp; Kawabata, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2024\u003c/span\u003e), the findings underscore the importance of designing learning environments that structure how AI is encountered and used, rather than leaving engagement entirely to individual discretion (Casanova et al., \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2020\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eFirst, learning environments should be designed to support both pathways of AI-mediated learning. The technical\u0026ndash;behavioural pathway requires environments that enable productive AI-assisted learning behaviours, such as opportunities for iterative exploration, feedback, and revision. The cognitive\u0026ndash;ethical pathway requires environments that foreground judgment, interpretation, and restraint. Instructional designs that explicitly differentiate behavioural engagement (how AI is used) from evaluative reasoning (how AI outputs are judged) better align with how AI literacy operates in practice (Walter, \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). The instructional environments can support productive AI engagement by making learning processes visible. Pairing AI-supported tasks with reflective artifacts, such as process logs, source-checking exercises, or comparisons between AI-assisted and non-AI work, helps align AI use with learning objectives rather than efficiency alone (Chiu, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eSecond, traditional output-based assessment becomes increasingly insufficient for evaluating student learning and critical thinking, particularly in less hands-on disciplines such as business. Prior research in learning environments shows that when AI tools externalize cognitive work, observable products may no longer reliably reflect learners\u0026rsquo; underlying reasoning or conceptual understanding (Elshall \u0026amp; Badir, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Minn, \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Our findings suggest that assessment must shift toward making learners\u0026rsquo; judgment and reasoning visible, rather than relying solely on written outputs.\u003c/p\u003e \u003cp\u003eDialogic and process-oriented and student-cantered assessment practices, such as brief oral explanations, structured discussion, guided questioning, peer critique, and instructor\u0026ndash;student conferences, have long been recognized as effective for supporting deep learning (Baeten et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). They become especially critical in AI-augmented learning environments, where written products alone provide limited insight into students\u0026rsquo; thinking. By foregrounding reasoning in interaction, these approaches allow instructors to evaluate how students interpret, justify, and regulate AI use in real time, supporting metacognitive regulation and meaningful assessment without imposing excessive documentation demands.\u003c/p\u003e \u003c/div\u003e"},{"header":"Limitation and Conclusion","content":"\u003cp\u003eAs in prior AI-in-education research, this study relies on self-reported measures, which may introduce response bias. The cross-sectional design also limits causal inference. Future research should employ longitudinal or experimental designs to examine how learner orientations evolve as AI becomes further embedded in learning environments, and how changes in instructional design influence engagement, judgment, and learning outcomes over time.\u003c/p\u003e \u003cp\u003eAs AI becomes a persistent feature of academic learning environments, understanding how students adapt to its presence is increasingly critical. This study demonstrates that learning gains in AI-augmented settings are shaped not by AI use alone, but by two complementary learning pathways through which AI literacy operates. The first is a technical\u0026ndash;behavioural pathway, in which technical AI literacy enables students to engage in AI-assisted learning behaviours that support comprehension, practice, and revision, thereby indirectly contributing to learning gains. The second is a cognitive\u0026ndash;ethical pathway, in which cognitive and social\u0026ndash;ethical AI literacy support learning through evaluative judgment, guiding how students assess AI outputs, set boundaries on use, and integrate AI support responsibly. These pathways highlight that effective learning in AI-rich environments depends on both productive engagement with AI tools and reflective judgment about their use. For educators, the central challenge is therefore not whether students use AI, but how learning environments and assessment practices can be designed to surface reasoning, support judgment, and sustain learning quality as AI becomes an integral part of academic practice.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003eA statement of ethics approval - The study was approved by Farmingdale State College IRB/ethics committee.\u003c/p\u003e \u003cp\u003e \u003cstrong\u003eConflict of interest\u003c/strong\u003e \u003cp\u003eThe author confirms that there are no conflicts of interest.\u003c/p\u003e \u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAcosta-Enriquez BG, Valle G, Arbulu Ballesteros MLA, Arbulu Castillo M, Perez Vargas JCA, Torres CG, Silva Leon IS, P. M., Saavedra Tirado K (2025) What is the influence of psychosocial factors on artificial intelligence appropriation in college students? BMC Psychol 13(1):7. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/39755638\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/39755638\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBaeten M, Dochy F, Struyven K, Parmentier E, Vanderbruggen A (2016) Student-centred learning environments: an investigation into student teachers\u0026rsquo; instructional preferences and approaches to learning. Learn Environ Res 19(1):43\u0026ndash;62\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBewersdorff A, Hornberger M, Nerdel C, Schiff DS (2025) AI advocates and cautious critics: How AI attitudes, AI interest, use of AI, and AI literacy build university students' AI self-efficacy. \u003cem\u003eComputers and Education: Artificial Intelligence\u003c/em\u003e, \u003cem\u003e8\u003c/em\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBraun V, Clarke V (2006) Using thematic analysis in psychology. Qualitative Res Psychol 3(2):77\u0026ndash;101\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBraun V, Clarke V (2021) Thematic analysis: A practical guide\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCasanova D, Huet I, Garcia F, Pessoa T (2020) Role of technology in the design of learning environments. Learn Environ Res 23(3):413\u0026ndash;427\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChang X, Wong GK (2025) A systematic review of how educators integrate ethics into artificial intelligence curriculum. J Res Technol Educ, 1\u0026ndash;18\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen Y, Zare S (2025) University students\u0026rsquo; AI thinking: A Latent Profile Analysis and predictive effects of psychological factors. J Res Technol Educ, 1\u0026ndash;14\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChiu M-L (2025) Exploring user awareness and perceived usefulness of generative AI in higher education: The moderating role of trust. Educ Inform Technol, 1\u0026ndash;35\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChiu TK, Chen Y, Yau KW, Chai C-s, Meng H, King I, Wong S, Yam Y (2024) Developing and validating measures for AI literacy tests: From self-reported to objective measures. Computers Education: Artif Intell 7:100282\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChun Z, Ning Y, Chen J, Wijaya TT (2025) Exploring the interplay among artificial intelligence literacy, creativity, self-efficacy, and academic achievement in college students: Findings from PLS-SEM and FsQCA. Educ Inform Technol, 1\u0026ndash;34\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCrompton H, Burke D (2023) Artificial intelligence in higher education: the state of the field. Int J Educational Technol High Educ 20(1):22\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEden CA, Chisom ON, Adeniyi IS (2024) Integrating AI in education: Opportunities, challenges, and ethical considerations. Magna Scientia Adv Res Reviews 10(2):006\u0026ndash;013\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEisenbardt M, Kurtz G, Gafni R, Leiba M, Dettmer S, Mullins R, Siegert I, Eisenbardt T, Gal E (2025) Students\u0026rsquo; Readiness and Ethical Considerations to Adopt Generative AI in their learning: An International Comparative Study. Technol Knowl Learn, 1\u0026ndash;28\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eElshall AS, Badir A (2025) Balancing AI-assisted learning and traditional assessment: the FACT assessment in environmental data science education. Frontiers in Education\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFetters MD, Curry LA, Creswell JW (2013) Achieving integration in mixed methods designs\u0026mdash;principles and practices. Health Serv Res 48(6pt2):2134\u0026ndash;2156\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGrassini S (2023) Development and validation of the AI attitude scale (AIAS-4): a brief measure of general attitude toward artificial intelligence. Front Psychol 14:1191628\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHornberger M, Bewersdorff A, Nerdel C (2023) What do university students know about Artificial Intelligence? Development and validation of an AI literacy test. Computers Education: Artif Intell, \u003cem\u003e5\u003c/em\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIfenthaler D, Majumdar R, Gorissen P, Judge M, Mishra S, Raffaghelli J, Shimada A (2024) Artificial intelligence in education: Implications for policymakers, researchers, and practitioners. Technol Knowl Learn 29(4):1693\u0026ndash;1710\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJi Y, Zhong M, Lyu S, Li T, Niu S, Zhan Z (2025) How does AI literacy affect individual innovative behavior: the mediating role of psychological need satisfaction, creative self-efficacy, and self-regulated learning. Educ Inform Technol 30(11):16133\u0026ndash;16162\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKajiwara Y, Kawabata K (2024) AI literacy for ethical use of chatbot: Will students accept AI ethics? Computers Education: Artif Intell 6:100251\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKim D, Wang CK, Borowiec K, Rein N, Cho J-HA, Liu JJ (2025) Unlocking the Future: A Comprehensive Review of ChatGPT in Education. Technol Knowl Learn, 1\u0026ndash;53\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKim J, Yu S, Lee S-S, Detrick R (2025) Students\u0026rsquo; prompt patterns and its effects in AI-assisted academic writing: Focusing on students\u0026rsquo; level of AI literacy. J Res Technol Educ, 1\u0026ndash;18\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKong S-C, Cheung WM-Y, Zhang G (2022) Evaluating artificial intelligence literacy courses for fostering conceptual learning, literacy and empowerment in university students: Refocusing to conceptual building. Computers Hum Behav Rep 7:100223\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKosmyna N, Hauptmann E, Yuan YT, Situ J, Liao X-H, Beresnitzky AV, Braunstein I, Maes P (2025) Your brain on ChatGPT: Accumulation of cognitive debt when using an AI assistant for essay writing task\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLaupichler MC, Aster A, Schirch J, Raupach T (2022) Artificial intelligence literacy in higher and adult education: A scoping literature review. Computers Education: Artif Intell 3:100101\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLong D, Magerko B (2020) \u003cem\u003eWhat is AI Literacy? Competencies and Design Considerations\u003c/em\u003e Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiehner L, Hick G, Biermann A, Brauner H, P., Ziefle M (2023) Perceptions, attitudes and trust toward artificial intelligence \u0026mdash; An assessment of the public opinion. Artificial Intelligence and Social Computing\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMedina-Gual L, Parejo J-L (2025) University Students\u0026rsquo; Engagement with Artificial Intelligence: A Cluster Analysis of Learner Profiles in AI Literacy. Technol Knowl Learn, 1\u0026ndash;19\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMinn S (2022) AI-assisted knowledge assessment techniques for adaptive learning environments. Computers Education: Artif Intell 3:100050\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMontali S (2025) Oct 7th, 2025). A Debate About A.I. Plays Out on the Subway Walls. \u003cem\u003eNew York Times\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.nytimes.com/2025/10/07/style/friend-ai-subway-ads-new-york.html\u003c/span\u003e\u003cspan address=\"https://www.nytimes.com/2025/10/07/style/friend-ai-subway-ads-new-york.html\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMorales-Garc\u0026iacute;a WC, Sairitupa-Sanchez LZ, Morales-Garc\u0026iacute;a SB, Morales-Garc\u0026iacute;a M (2024) Adaptation and psychometric properties of a brief version of the general self-efficacy scale for use with artificial intelligence (GSE-6AI) among university students. \u003cem\u003eFrontiers in Education\u003c/em\u003e, \u003cem\u003e9\u003c/em\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNg DTK, Leung JKL, Chu SKW, Qiao MS (2021) Conceptualizing AI literacy: An exploratory review. Computers Education: Artif Intell 2:100041\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNg W (2012) Can we teach digital natives digital literacy? Comput Educ 59(3):1065\u0026ndash;1078\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSeymour E, Wiese D, Hunter A, Daffinrud SM (2000) Creating a better mousetrap: On-line student assessment of their learning gains. National Meeting of the American Chemical Society\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShardlow M, Sellar S, Rousell D (2022) Collaborative augmentation and simplification of text (CoAST): Pedagogical applications of natural language processing in digital learning environments. Learn Environ Res 25(2):399\u0026ndash;421\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWalter Y (2024) Embracing the future of Artificial Intelligence in the classroom: the relevance of AI literacy, prompt engineering, and critical thinking in modern education. Int J Educational Technol High Educ 21(1):15\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYounas A, F\u0026agrave;bregues S, Creswell JW (2023) Generating metainferences in mixed methods research: A worked example in convergent mixed methods designs. Methodological Innovations 16(3):276\u0026ndash;291\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZawacki-Richter O, Mar\u0026iacute;n VI, Bond M, Gouverneur F (2019) Systematic review of research on artificial intelligence applications in higher education \u0026ndash; where are the educators? Int J Educational Technol High Educ, \u003cem\u003e16\u003c/em\u003e(1)\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"},{"header":"Tables","content":"\u003cp\u003e\u003cstrong\u003eTable 1 Interviewees Profiles\u003c/strong\u003e\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eID\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eGender\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 268px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eMajor\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 34px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eGrade level\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003eM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 268px;\"\u003e\n \u003cp\u003eSports Management\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 34px;\"\u003e\n \u003cp\u003eFreshman\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003eM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 268px;\"\u003e\n \u003cp\u003eNursing\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 34px;\"\u003e\n \u003cp\u003eFreshman\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003eM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 268px;\"\u003e\n \u003cp\u003eBusiness Analytics\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 34px;\"\u003e\n \u003cp\u003eSophomore\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003eM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 268px;\"\u003e\n \u003cp\u003eArchitecture\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 34px;\"\u003e\n \u003cp\u003eSophomore\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003eF\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 268px;\"\u003e\n \u003cp\u003eComputer Science\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 34px;\"\u003e\n \u003cp\u003eJunior\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003eM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 268px;\"\u003e\n \u003cp\u003eScience, Technology Society\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 34px;\"\u003e\n \u003cp\u003eSophomore\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003eM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 268px;\"\u003e\n \u003cp\u003eMechanical Engineering Technology\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 34px;\"\u003e\n \u003cp\u003eSophomore\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003eF\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 268px;\"\u003e\n \u003cp\u003eArchitecture\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 34px;\"\u003e\n \u003cp\u003eSophomore\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003eF\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 268px;\"\u003e\n \u003cp\u003eBiology\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 34px;\"\u003e\n \u003cp\u003eFreshman\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003eM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 268px;\"\u003e\n \u003cp\u003eComputer Science\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 34px;\"\u003e\n \u003cp\u003eSophomore\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cstrong\u003eTable 2 Student AI Engagement Framework: Dimensions, Definitions, and Illustrative Quotes\u003c/strong\u003e\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"642\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 120px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eDimension\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 144px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eCore Idea\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 378px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eIllustrative Quotes\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 120px;\"\u003e\n \u003cp\u003eAcademic Enablement\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 144px;\"\u003e\n \u003cp\u003eAI increases efficiency, comprehension, and confidence in coursework.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 378px;\"\u003e\n \u003cp\u003e\u0026ldquo;It helps you take notes better, study better\u0026mdash;like reviewing the notes AI helped me generate. It\u0026apos;s a really good study tool.\u0026rdquo; (Int. 1)\u003c/p\u003e\n \u003cp\u003e\u0026ldquo;I can ask AI to summarize it. It gives me the main points quickly and saves a lot of time.\u0026rdquo; (Int.2)\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026ldquo;It\u0026rsquo;s helped explain things when I asked in-depth questions. It\u0026rsquo;s given me context or extra knowledge\u0026nbsp;I didn\u0026rsquo;t have before. It\u0026rsquo;s helped me learn more material, which in turn helps me complete more assignments.\u0026rdquo; (Int.7)\u003c/p\u003e\n \u003cp\u003e\u0026ldquo;AI makes computer science more exciting. It adds tools, not threats.\u0026rdquo; (Int.10)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 120px;\"\u003e\n \u003cp\u003eAI Literacy as Competence\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 144px;\"\u003e\n \u003cp\u003eStudents emphasize prompting, verification, and critical thinking as core skills for effective use.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 378px;\"\u003e\n \u003cp\u003e\u0026ldquo;I think it means understanding how AI communicates and works\u0026mdash;how it gives answers and what it\u0026rsquo;s capable of. It\u0026rsquo;s not just about using it but knowing how to interpret what it gives you.\u0026rdquo; (Int.5) \u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026ldquo;The biggest skill is knowing how to write a prompt.\u0026rdquo; (Int.5)\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026ldquo;You have to be flexible and good at adapting. AI is always changing as new information comes out. You need to be open-minded and always look at what\u0026rsquo;s new.\u0026rdquo; (Int.7)\u003c/p\u003e\n \u003cp\u003e\u0026ldquo;Using AI is surface-level. Literacy is about control and critical understanding.\u0026rdquo; (Int.10)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 120px;\"\u003e\n \u003cp\u003eConditional Trust \u0026amp; Risk Management\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 144px;\"\u003e\n \u003cp\u003eUsefulness is balanced with concerns about accuracy, plagiarism, privacy, and institutional penalties.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 378px;\"\u003e\n \u003cp\u003e\u0026ldquo;Sometimes AI gives wrong answers\u0026hellip; so I double-check before submitting assignments.\u0026rdquo; (Int.1)\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u0026ldquo;No. I always avoid [enter private information]\u0026mdash;it\u0026rsquo;s learning from users, so there\u0026rsquo;s no guarantee your data stays private.\u0026rdquo; (Int.4)\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026ldquo;Responsibility and balance. You have to know how to use AI responsibly and also develop your own skills.\u0026rdquo; (Int.5)\u003c/p\u003e\n \u003cp\u003e\u0026ldquo;I believe misinformation is the biggest problem.\u0026rdquo; (Int.6)\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026ldquo;My main concern is plagiarism because AI just pulls information from the Internet or its own databases.\u0026rdquo; (Int.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 120px;\"\u003e\n \u003cp\u003eHuman\u0026ndash;AI Boundaries\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 144px;\"\u003e\n \u003cp\u003eStudents view creativity, judgment, ethics, and hands-on learning as distinctly human strengths.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 378px;\"\u003e\n \u003cp\u003e\u0026ldquo;AI doesn\u0026rsquo;t understand human emotions or have emotional intelligence. I wouldn\u0026rsquo;t trust it to define me.\u0026rdquo; (Int.1)\u003c/p\u003e\n \u003cp\u003e\u0026ldquo;I try not to rely on it as much because I know I need to remember that material in the future.\u0026rdquo; (Int.2)\u003c/p\u003e\n \u003cp\u003e\u0026ldquo;I\u0026rsquo;ve always wanted to make sure I actually learn and retain the material for future applications\u0026hellip;If I handed in something AI-generated, I wouldn\u0026rsquo;t know what I wrote and wouldn\u0026rsquo;t retain it. \u0026nbsp;(Int.7)\u003c/p\u003e\n \u003cp\u003e\u0026ldquo;Experimental design, ethical reasoning, and hands-on lab work need human input.\u0026rdquo; (Int.9)\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026ldquo;Critical thinking, scientific literacy, and knowing when to rely on AI versus human expertise..\u0026rdquo; (Int.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 120px;\"\u003e\n \u003cp\u003ePedagogical Guardrails\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 144px;\"\u003e\n \u003cp\u003eStudents support AI in education, but only with clear guidelines and as a supplement\u0026mdash;not a replacement for instructors.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 378px;\"\u003e\n \u003cp\u003e\u0026ldquo;School should teach how to use AI responsibly.\u0026rdquo; (Int.3)\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026ldquo;Schools would have to be very careful. If we can\u0026rsquo;t make it safe and credible, then I don\u0026rsquo;t think it should have a place in education.\u0026rdquo; (Int.7)\u003c/p\u003e\n \u003cp\u003e\u0026ldquo;Yes, but only as a support tool\u0026mdash;not to replace teachers.\u0026rdquo; (Int.8) \u0026nbsp; \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026ldquo;It should support learning, not replace human instruction.\u0026rdquo; (Int.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cstrong\u003eTable 3 Correction Table of the main variables (\u003c/strong\u003e\u003cstrong\u003eN=81)\u003c/strong\u003e\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 132px;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 61px;\"\u003e\n \u003cp\u003eMean\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 59px;\"\u003e\n \u003cp class=\"MsoNormal\"\u003e\u003cspan lang=\"EN-US\"\u003eS.D\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 69px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 69px;\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 72px;\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 69px;\"\u003e\n \u003cp\u003e4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 64px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 132px;\"\u003e\n \u003cp\u003eTechnical AI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 61px;\"\u003e\n \u003cp\u003e3.88\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 59px;\"\u003e\n \u003cp\u003e0.86\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 69px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 69px;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 72px;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 69px;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 64px;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 132px;\"\u003e\n \u003cp\u003eCognitive AI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 61px;\"\u003e\n \u003cp\u003e3.98\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 59px;\"\u003e\n \u003cp\u003e0.77\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 69px;\"\u003e\n \u003cp\u003e.299**\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 69px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 72px;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 69px;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 64px;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 132px;\"\u003e\n \u003cp\u003eSocial-Ethical AI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 61px;\"\u003e\n \u003cp\u003e4.23\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 59px;\"\u003e\n \u003cp\u003e0.68\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 69px;\"\u003e\n \u003cp\u003e0.047\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 69px;\"\u003e\n \u003cp\u003e.421**\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 72px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 69px;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 64px;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 132px;\"\u003e\n \u003cp\u003eAALB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 61px;\"\u003e\n \u003cp\u003e3.36\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 59px;\"\u003e\n \u003cp\u003e1.05\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 69px;\"\u003e\n \u003cp\u003e.682**\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 69px;\"\u003e\n \u003cp\u003e0.198\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 72px;\"\u003e\n \u003cp\u003e-0.003\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 69px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 64px;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 132px;\"\u003e\n \u003cp\u003eLearning Gain\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 61px;\"\u003e\n \u003cp\u003e3.42\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 59px;\"\u003e\n \u003cp\u003e1.16\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 69px;\"\u003e\n \u003cp\u003e.668**\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 69px;\"\u003e\n \u003cp\u003e.339**\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 72px;\"\u003e\n \u003cp\u003e0.088\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 69px;\"\u003e\n \u003cp\u003e.813**\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 64px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e* p \u0026lt; .05, ** p \u0026lt; .01, *** p \u0026lt; .001.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 4 Regression results for\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;AI literacy, AI-Assisted Learning Behavior (AALB) and Learning Gains.\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eDependent variables: Model 1 \u0026amp; B = Learning Gain; Model A = Learning Behavior.\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 180px;\"\u003e\n \u003cp\u003e\u003cstrong\u003ePredictor\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 157px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eModel 1: Learning Gain (no mediator)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 119px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eModel A: Learning Behavior\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 167px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eModel B: Learning Gain (with mediator)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 180px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTechnical AI\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 157px;\"\u003e\n \u003cp\u003e0.81*** (0.13)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 119px;\"\u003e\n \u003cp\u003e0.88*** (0.12)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 167px;\"\u003e\n \u003cp\u003e0.18 (0.13)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 180px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eCognitive AI\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 157px;\"\u003e\n \u003cp\u003e0.26 (0.16)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 119px;\"\u003e\n \u003cp\u003e\u0026minus;0.10 (0.15)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 167px;\"\u003e\n \u003cp\u003e0.33** (0.12)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 180px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSocial-ethical AI\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 157px;\"\u003e\n \u003cp\u003e\u0026minus;0.03 (0.17)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 119px;\"\u003e\n \u003cp\u003e\u0026minus;0.03 (0.16)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 167px;\"\u003e\n \u003cp\u003e\u0026minus;0.01 (0.13)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 180px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAALB\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 157px;\"\u003e\n \u003cp\u003e\u0026mdash;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 119px;\"\u003e\n \u003cp\u003e\u0026mdash;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 167px;\"\u003e\n \u003cp\u003e0.72*** (0.10)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 180px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAge (years)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 157px;\"\u003e\n \u003cp\u003e0.01 (0.02)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 119px;\"\u003e\n \u003cp\u003e\u0026minus;0.00 (0.02)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 167px;\"\u003e\n \u003cp\u003e0.02 (0.01)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 180px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSchool: Business\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 157px;\"\u003e\n \u003cp\u003e0.19 (0.26)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 119px;\"\u003e\n \u003cp\u003e0.21 (0.25)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 167px;\"\u003e\n \u003cp\u003e0.04 (0.20)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 180px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSchool: Engineering\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 157px;\"\u003e\n \u003cp\u003e0.32 (0.33)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 119px;\"\u003e\n \u003cp\u003e\u0026minus;0.10 (0.31)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 167px;\"\u003e\n \u003cp\u003e0.40 (0.25)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 180px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSchool: Health\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 157px;\"\u003e\n \u003cp\u003e0.86* (0.43)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 119px;\"\u003e\n \u003cp\u003e0.38 (0.40)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 167px;\"\u003e\n \u003cp\u003e0.58\u0026dagger; (0.32)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 180px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eR\u0026sup2;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 157px;\"\u003e\n \u003cp\u003e0.47\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 119px;\"\u003e\n \u003cp\u003e0.46\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 167px;\"\u003e\n \u003cp\u003e0.71\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 180px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAdj. R\u0026sup2;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 157px;\"\u003e\n \u003cp\u003e0.41\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 119px;\"\u003e\n \u003cp\u003e0.40\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 167px;\"\u003e\n \u003cp\u003e0.68\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 180px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eN\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 157px;\"\u003e\n \u003cp\u003e75\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 119px;\"\u003e\n \u003cp\u003e75\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 167px;\"\u003e\n \u003cp\u003e75\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eNotes: Entries are unstandardized regression coefficients with standard errors in parentheses. All models control for age and school (reference category not shown). Indirect effect for TechAI via AALB from bootstrap: 0.64, 95% CI [0.40, 0.91]. \u0026dagger; p \u0026lt; .10, * p \u0026lt; .05, ** p \u0026lt; .01, *** p \u0026lt; .001.\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"Farmingdale State College","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"AI literacy, academic learning gains, mixed-methods, higher education, AI-assisted learning behaviors, AI ethics","lastPublishedDoi":"10.21203/rs.3.rs-9173238/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9173238/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eGenerative artificial intelligence (AI) tools are becoming embedded in higher education, raising questions about how learning unfolds in AI-augmented environments. This convergent mixed-methods study examines how three dimensions of AI literacy - technical, cognitive, and social-ethical - relate to learning processes and outcomes. Qualitative analysis of ten interviews with college students shows that learners engage with AI pragmatically but cautiously, using it to support efficiency, comprehension, and academic work while actively regulating use to preserve learning quality and integrity. Quantitative survey results show that technical AI literacy is associated with greater use of AI-assisted learning behaviors, which are in turn associated with higher perceived learning gains. Cognitive AI literacy is directly associated with perceived learning gains and stronger social-ethical awareness, whereas social-ethical AI literacy shows no direct association with learning behaviors or outcomes. Together, the findings indicate that learning in AI-augmented environments is shaped by two complementary pathways: 1) a technical\u0026ndash;behavioral pathway that supports effective learning engagement with AI tools, and 2) a cognitive\u0026ndash;ethical pathway that supports evaluative judgment and responsible use. These results highlight the importance of designing learning environments and assessment practices that support both procedural fluency and reflective judgment in AI-integrated higher education.\u003c/p\u003e","manuscriptTitle":"Learning in AI-Augmented Environments Through Dual Pathways of AI Literacy","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-03-24 15:30:00","doi":"10.21203/rs.3.rs-9173238/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"d5e636fa-270c-4efa-8109-a3e14a7aee4f","owner":[],"postedDate":"March 24th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":64817805,"name":"Educational Philosophy and Theory"}],"tags":[],"updatedAt":"2026-03-24T15:30:00+00:00","versionOfRecord":[],"versionCreatedAt":"2026-03-24 15:30:00","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9173238","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9173238","identity":"rs-9173238","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}