Exposure is not development
Most people who use AI regularly do not become AI literate through use alone. Exposure produces familiarity — the ability to get something useful out of an AI interaction. Development produces capability — the ability to judge when and how AI engagement serves genuine goals. The difference requires deliberate practice and reflection, not just accumulated hours.
Developing AI literacy
One-sentence definition: The progressive deepening of engagement with AI systems — from substitution through adaptation to transformation — through which the six dimensions of AI literacy develop in concert via authentic use and deliberate reflection.
Development is not the same as skill accumulation. A person can become highly familiar with AI tools — navigating interfaces, generating reasonable outputs, avoiding obvious errors — without developing the contextual judgement, critical evaluation, or ethical awareness that constitute genuine literacy. The distinction matters because familiarity is comfortable and self-reinforcing: outputs improve, which feels like progress, while the deeper dimensions remain undeveloped.
The three-stage trajectory
The developmental stages reflect a recognisable pattern in how people integrate new capabilities into professional practice.
Substitution is the natural entry point because cognitive load is lowest: you are using AI for tasks you already know how to do, which means you can evaluate outputs against existing knowledge. This stage primarily develops functional application (Dimension 3) and surface-level critical evaluation (Dimension 2). It is valuable as a starting point but tends to be self-limiting — once a workflow is efficient, there is little internal pressure to go further.
Adaptation requires allowing AI capabilities to change how you approach tasks, not just which tool you use for them. Rather than substituting AI for Google or a blank page, you restructure the activity around what AI makes possible — building understanding through staged dialogue, testing arguments through interlocution, decomposing problems before attempting synthesis. This stage develops creation and communication (Dimension 4) and deepens access and recognition (Dimension 1).
Transformation cannot be designed or rushed — it emerges from sustained practice with accumulated context and growing professional judgement. AI becomes integrated infrastructure rather than a discrete tool: context builds across sessions, engagement patterns become deliberate and domain-specific, and the central question shifts from “how do I use this?” to “when does this serve my goals?” This is where contextual judgement (Dimension 6) and ethical awareness (Dimension 5) develop most fully.
What development requires
Three conditions distinguish development from mere exposure.
Authentic use. Literacy develops through real tasks with genuine stakes. The disciplinary or professional content is primary; AI literacy develops as a byproduct of genuine engagement, not through exercises designed to practise AI use for its own sake.
Deliberate reflection. Moving between stages requires metacognitive work — evaluating the quality of engagement, not just the quality of outputs. Without reflection, people stabilise at substitution because it is efficient and comfortable. Reflection surfaces the dimensions that are not developing and creates the conditions for progression.
Foundational orientation. Before development can begin, learners need to understand what AI systems are, how they generate responses, and where they fail. This is not the same as studying AI as a topic — it is the minimum conceptual grounding that makes critical evaluation possible. Without it, the first stage collapses into uncritical use.
See AI literacy development framework for practical guidance on embedding these conditions into existing modules and courses.
Sources
- 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, 1–16. https://doi.org/10.1145/3313831.3376727
- Mollick, E., & Mollick, L. (2023). Assigning AI: Seven approaches for students, with prompts. SSRN Electronic Journal.
- UNESCO. (2024). Guidance for generative AI in education and research.
Notes
People often get stuck at substitution not because they lack capability but because substitution is self-reinforcing: outputs are better than nothing, efficiency improves, and there is no obvious signal that something is missing. Progression to adaptation typically requires external disruption — a task that substitution handles poorly, or deliberate reflection that surfaces the gap between familiarity and genuine competence.