The structural pathways from AI attitude to AI literacy among university students: considering the serial mediators of AI self-efficacy and AI motivation

As artificial intelligence rapidly transitions from a peripheral technological novelty into a foundational pillar of the global educational ecosystem, higher education institutions face an unprecedented mandate: equipping students with robust artificial intelligence literacy. While previous academic inquiries have heavily emphasized external macro-level interventions—such as institutional policies, instructional designs, and classroom environments—a comprehensive new empirical study shifts the focus inward. Researchers have uncovered that a student’s internal psychological makeup, specifically their attitudes, self-efficacy, and motivations regarding artificial intelligence, plays a decisive role in shaping their overall AI competency.

The formal investigation, conducted by scholars X. Wang and J. Nonprofit, explored the intricate structural relationships connecting student attitudes toward artificial intelligence with their ultimate digital literacy levels. Utilizing a sophisticated chain mediation model, the researchers surveyed 1,323 undergraduate students across two major universities in Shandong Province, China, representing both natural sciences and humanities disciplines. The findings offer a groundbreaking look at how subjective cognitive evaluations cascade into high-order technological capabilities, providing a vital roadmap for educators navigating the intelligent era.

Chronology and Methodology of the Study

The empirical investigation unfolded over several distinct phases, beginning with questionnaire distribution between December 2025 and January 2026, coinciding with the first semester of the academic year. The research team partnered with class counselors across academic departments, deploying electronic surveys via online data-collection platforms to target freshmen, sophomores, juniors, and seniors. Out of 1,453 initial respondents, rigorous data-cleaning protocols—including listwise deletion for missing values and the removal of invariant or inattentive responses—yielded a final, highly reliable sample size of 1,323 undergraduate students.

To measure the core psychological constructs, the research team employed established psychometric instruments subjected to rigorous translation and back-translation procedures. Student attitudes were gauged using a unidimensional scale evaluating perceived human utility, while AI self-efficacy was assessed through dimensions of problem-solving confidence and continuous learning capacity. Furthermore, AI utilization motivation was categorized into four distinct orientations: instrumental productivity, entertainment, social connection, and stress escape. Finally, overall AI literacy was measured across four vital pillars: basic awareness, practical application, critical evaluation, and ethical compliance. Data analysis via SPSS and bootstrapping macro models allowed the team to evaluate direct paths as well as single and serial mediating effects.

Statistical Findings and Mediation Pathways

The empirical results provide compelling quantitative evidence regarding the psychological mechanisms driving student competence in automated environments. Descriptive statistics revealed high overall mean scores for student attitudes (3.939 out of 5), self-efficacy (3.802 out of 5), and AI literacy (4.932 out of 7), indicating a generally receptive and confident student body. Pearson correlation analyses confirmed significant positive relationships across all focal variables, with variance inflation factor (VIF) metrics remaining safely below standard multicollinearity thresholds.

The structural equation modeling and bootstrapping procedures unveiled a powerful chain of psychological associations:

• Direct Association: AI attitude exhibited a strong, direct positive association with AI literacy ($beta = 0.243, p < 0.001$), demonstrating that holding a favorable view of technology directly correlates with enhanced competency.
• Single Mediation of AI Self-Efficacy: The indirect path operating through AI self-efficacy proved to be the most substantial mechanism identified in the study ($beta = 0.563$), indicating that positive attitudes heavily bolster student confidence in tackling complex algorithmic challenges, which in turn elevates literacy.
• Single Mediation of AI Motivation: AI motivation also served as a significant independent mediator ($beta = 0.030$), reflecting that favorable attitudes stimulate purposeful utilization drives that aid competency acquisition.
• Serial Mediation: The sequential pathway moving from AI attitude through AI self-efficacy and subsequently to AI motivation before reaching AI literacy was statistically significant ($beta = 0.031$), confirming that confidence acts as a vital bridge enabling operational attitudes to translate into goal-directed behavioral drives.

Theoretical Foundations and Expert Insights

To contextualize these findings, the research integrated Albert Bandura’s renowned self-efficacy theory with Katz et al.’s Uses and Gratifications Theory (UGT). According to the study’s theoretical framework, an individual’s evaluative attitude functions as an affective filter that reduces technophobia and anxiety. When students perceive artificial intelligence as an empowering ally rather than an academic threat, their cognitive bandwidth is liberated to build task-specific self-efficacy.

Experts note that while motivation acts as an initial catalyst for engagement, robust self-efficacy is required to sustain the cognitive persistence necessary to master steep technological learning curves. As automated systems become ubiquitous in modern workflows, students who lack internal confidence or harbor defensive psychological resistance are significantly more vulnerable to structural disadvantages in future labor markets. Consequently, the research highlights that external technological proliferation alone is insufficient; institutions must actively cultivate internal psychological readiness.

Implications for Higher Education and Policy

The findings carry profound practical implications for university administrators, curriculum developers, and academic policymakers worldwide. Rather than relying exclusively on top-down technical instruction or hardware upgrades, educational stakeholders are urged to adopt a holistic, psychological-informed approach to digital pedagogy.

1. Attitudinal Interventions: Introductory courses should incorporate structured reflection and ethical dialogues to dispel unwarranted fears regarding algorithmic displacement and foster affirmative baseline perceptions.
2. Scaffolding Self-Efficacy: Curriculum designs must feature graduated learning tasks—progressing from low-stakes auxiliary tools to complex problem-solving environments—allowing students to accumulate mastery experiences and build unshakeable technical confidence.
3. Channeling Motivation: Instructors should anchor AI tasks in authentic disciplinary demands and real-world problem-solving, satisfying students’ instrumental and collaborative needs to maintain long-term engagement.

While the authors note certain methodological limitations—such as the reliance on self-reported perception scales and the exploratory nature of unrecorded classroom-level clustering—the study establishes a critical foundation for future longitudinal and experimental research. Ultimately, the investigation redefines digital education, proving that cultivating future-ready competencies requires nurturing the cognitive, affective, and motivational agency of every student.