Enhancing AI Literacy Course Satisfaction Through Empowerment in AI Problem-Solving and Ethical Awareness: Development and Validation of an AI Project-Based Learning Scale

Synthesis: Zhu and Kong develop and validate a context-grounded AI project-based learning scale (AI-PBLS) for measuring students' perceived Project-Based Learning experiences when using AI to solve real-world problems, then test how those experiences relate to satisfaction with AI Literacy courses. Using data from 1,027 Hong Kong secondary and university students enrolled in an AI literacy program — 446 with complete data for structural equation modeling (SEM) — they show that empowerment in using AI for Problem Solving and AI ethical awareness mediate the relationship between perceived PBL and AI literacy course satisfaction. The study positions AI literacy as multidimensional, spanning cognitive, metacognitive, affective, and social dimensions, and grounds the mediation model in self-determination theory and social cognitive theory.

Key Findings

• The AI-PBLS was validated as a reliable 7-item, single-factor scale after exploratory (EFA, N = 513) and confirmatory (CFA, N = 514) factor analyses, with high internal consistency (Cronbach's α = 0.963) and a context-grounded design where the "AI" element was defined by the instructional setting rather than item wording — a contribution to Educational Measurement of AI learning.
• Perceived PBL was positively associated with AI literacy course satisfaction (direct β = 0.202), supporting PBL as a suitable Pedagogies and Teaching Strategies for AI literacy education.
• Empowerment in using AI for problem-solving was a stronger mediator than AI ethical awareness — the indirect effect through empowerment (β = 0.393) far exceeded that through ethical awareness (β = 0.029) — suggesting educators should prioritize fostering student empowerment.
• Empowerment strongly predicted ethical awareness (β = 0.668), and the final SEM model explained 58.4%, 66.1%, and 67.1% of variance in empowerment, ethical awareness, and course satisfaction respectively.
• All measures correlated strongly and positively with course satisfaction (r = 0.67–0.77, p < 0.001), underscoring the links among PBL, engagement, empowerment, and ethics.

What this means for practice

• Instructional designers. Design AI literacy courses around authentic PBL projects — collaboration, hands-on artifact creation, real-world problem contexts — over isolated AI tasks, allocating roughly 4–6 weeks per project cycle so students can define problems, revise ideas, and reflect on outcomes.
• Instructional designers. Prioritize student empowerment over pure technical skill by giving learners meaningful choices, encouraging comparison of alternative AI tools or outputs, and asking them to justify how and why AI is used at each project stage.
• Instructional designers. Build empowerment in AI problem-solving as a measured course objective rather than an assumed by-product: it carried an indirect effect on course satisfaction of β = 0.393 against β = 0.029 for ethical awareness, and it predicted ethical awareness itself (β = 0.668).
• Educators. Embed ethical awareness throughout the project process rather than as a separate topic — for example, using prompts such as "What bias may exist in this AI output?" or "Who might be disadvantaged?" in proposals, peer discussions, Reflection journals, and final presentations.
• Administrators. Promote guidelines for responsible and equitable AI use and invest in professional development for teachers implementing PBL-based AI instruction.

Limitations

• The design is cross-sectional, so the associations between perceived PBL, empowerment, ethical awareness, and course satisfaction cannot be read as causal; the authors propose longitudinal data as the next step.
• All measures are self-reported, which the authors identify as a possible source of method bias and which captures perceptions rather than the dynamic development of empowerment or ethical awareness; Harman's single-factor test showed four factors with eigenvalues above 1, so common-method variance did not explain the pattern, but no behavioral or achievement data were collected.
• The AI-PBLS was not subjected to formal expert validation or cognitive interviews, so its content validity is described as preliminary; the "AI" element is defined by the instructional context rather than item wording, which means the scale assumes every respondent has completed AI-based problem-solving activities in the same course.
• Of 1,027 participating Hong Kong secondary and university students, only 446 provided complete data for the SEM, and the sample sits within a Confucian heritage culture emphasizing teacher authority, achievement, and collectivism, so the mediation estimates may not transfer to other settings.

Citation

Zhu, J., & Kong, S. C. (2026). Enhancing AI literacy course satisfaction through empowerment in AI problem-solving and ethical awareness: Development and validation of an AI project-based learning scale. Computers and Education: Artificial Intelligence, 11, 100624.