Technical Summary: Beyond Tool Adoption – A Practical Five-Stage Developmental Continuum for AI Literacy in Higher Education

Problem Statement

While Artificial Intelligence (AI) literacy is increasingly recognized as a foundational competency for university graduates, current student engagement with AI tools exhibits a problematic polarization. Students tend to cluster at two extremes: avoidance (driven by fear, mistrust, ethical concerns, or lack of access) and uncritical reliance (producing fluent outputs that mask a lack of disciplinary understanding, described as "performance without learning").

Existing AI literacy frameworks (e.g., UNESCO, OECD, AI Literacy Heptagon) primarily offer dimensional definitions of competency (what students should know and do) but lack developmental guidance on how learners progress from their current state toward responsible engagement. Specifically, these frameworks often fail to address pre-literacy states such as principled avoidance or fluent but uncritical use, leaving educators without a diagnostic tool to identify where students begin or how to design instruction that moves them forward. Furthermore, the distinction between mere tool adoption and genuine literacy is often blurred, leading to assessments that may reward surface-level performance over deep understanding.

Methodology

This paper proposes a conceptual framework grounded in design-based implementation and observational practice rather than controlled experimental validation.

1. Conceptual Framework Development: The authors synthesized multi-year instructional experiences across AI and generative AI courses and intensive workshops at North Carolina State University (NC State). They analyzed patterns of student engagement to identify distinct developmental orientations, resulting in a five-stage continuum.
2. Case Study Implementation: The framework was applied to a series of initiatives at NC State between Fall 2024 and Spring 2026. These included:
   • A 1-credit graduate course ("Generative AI for Science").
   • Multiple 2-day intensive hands-on workshops for diverse populations (undergraduates, staff, international students, high schoolers).
   • Ongoing 1-credit cross-campus courses.
   • Total participation: >330 individuals.
3. Data Collection: Evidence was gathered through post-workshop feedback surveys, instructor observations of engagement during training, and the examination of student-produced artifacts (e.g., prompt logs, revised outputs, workflow designs).
4. Limitations of Methodology: The authors explicitly note that the study did not use validated pre/post instruments or a control group. Consequently, findings are framed as observational and practice-based. Stage transitions are identified through behavioral indicators consistent with the proposed stages rather than psychometrically validated measurements.

Key Contributions

The paper introduces the AI Literacy Continuum, a five-stage developmental model designed to bridge the gap between abstract competency definitions and classroom practice.

The Five Stages

1. Stage 0: Not Yet Engaged: Learners avoid AI tools due to fear, mistrust, ethical objections, or lack of exposure. This includes both uninformed avoidance and principled refusal.
2. Stage 1: Uncritical Use: Learners use AI but accept outputs as authoritative without verification. This "copy-paste" orientation leads to "performance without learning," where fluent AI prose masks gaps in disciplinary knowledge.
3. Stage 2: Informed Use (Graduation Baseline): Learners understand basic mechanisms, recognize limitations (e.g., hallucinations), and apply human judgment. They verify facts, document use, and understand ethical implications.
4. Stage 3: Critical Evaluation: Learners systematically assess AI outputs for bias, uncertainty, and disciplinary appropriateness. This stage requires deep domain expertise to evaluate the reasoning and assumptions behind AI-generated content.
5. Stage 4: Improvement: Learners actively contribute to improving AI systems through design, customization, prompt engineering, fine-tuning, or research.

Alignment and Diagnostic Utility

• Complementarity: The continuum complements dimensional frameworks (like UNESCO's) by adding a temporal axis. It explicitly addresses the "pre-literacy" gap (Stages 0–1) that UNESCO's "Understand, Apply, Create" model assumes is already crossed.
• Assessment Gap Analysis: The paper maps existing instruments (e.g., SNAIL, MAILS, GLAT) to the continuum, revealing that current tools poorly assess Stages 0, 1, and 4, highlighting a need for new performance-based diagnostic tools.
• Institutional Baseline: The authors propose that Stages 2–3 should serve as the graduation baseline for higher education, ensuring all graduates possess informed and critical engagement skills.

Results

Based on observational evidence from NC State initiatives, the paper reports the following findings:

• Feasibility of Stage Transitions: Brief, well-designed interventions (1–2 days) were sufficient to move participants from Stage 0 (avoidance) or Stage 1 (uncritical use) toward Stage 2 (informed use). Participants demonstrated behaviors such as recognizing hallucinations, verifying outputs against sources, and articulating when human judgment is required.
• Role of Sustained Engagement: While short workshops facilitated initial transitions, evidence of Stage 3 (Critical Evaluation) and Stage 4 (Improvement) was more strongly associated with sustained, discipline-embedded experiences (e.g., semester-long courses, capstone projects).
• Universality of the Trajectory: The developmental trajectory appeared robust across diverse populations, including graduate students, undergraduates, high school students, and international visitors (e.g., Korean students), suggesting the continuum is applicable across age, discipline, and cultural contexts.
• Curricular Pathways: The case study demonstrated that a mix of general education modules (for Stages 0–2) and discipline-specific research experiences (for Stages 3–4) provides a scalable pathway for institutional implementation.

Significance and Claims

The paper argues that AI literacy should not be understood as tool adoption alone, but as a developmental capacity to understand, evaluate, and responsibly apply AI systems in specific contexts.

• Conceptual Clarity: The primary significance lies in distinguishing "AI use" from "AI literacy." By identifying "Uncritical Use" as a distinct, problematic stage, the framework prevents the conflation of tool familiarity with competence.
• Practical Diagnostic Tool: The continuum provides educators with a shared language to diagnose student starting points and design instruction that targets specific developmental needs (e.g., demystification for Stage 0 vs. verification habits for Stage 1).
• Policy-to-Practice Bridge: It offers a mechanism for translating high-level policy goals (e.g., UNESCO frameworks) into actionable curricular and assessment strategies.
• Modest Claims: The authors explicitly state that the framework is a pedagogical heuristic based on observational evidence, not a rigid psychological model. They call for future research to include formal empirical validation, controlled studies with pre/post assessments, and cross-cultural adaptation to confirm the model's robustness.

In conclusion, the paper posits that higher education must move beyond policies of prohibition or permissive adoption toward structured developmental approaches. The AI Literacy Continuum offers a practical, achievable framework for ensuring graduates can engage with AI as informed, critical, and responsible agents.