Building AI Competencies

AI competency is more than knowing how to prompt. Learners and educators need a working understanding of AI systems, the ability to make ethical and human-centred choices, skill in evaluating output, and judgment about when AI supports or displaces learning.

UNESCO’s teacher framework describes five dimensions—human-centred mindset, ethics, foundations and applications, AI pedagogy, and professional learning—at three progression levels: acquire, deepen, and create.¹ UNESCO’s student framework similarly combines a human-centred mindset, ethics, AI techniques and applications, and AI system design.²

A four-part learning pathway

1. Understand what the system is

Learners should be able to distinguish a model from a product, describe in plain language how generative systems produce output, and explain why fluency does not establish accuracy or understanding.

Suggested preparation: What is generative AI? and Limitations and risks.

Learning activity: give learners descriptions of a model, chatbot, search engine, writing assistant, and retrieval-based course tool. Ask them to label which components generate, retrieve, calculate, store, or present information and to identify what must be checked in each workflow.

Evidence of learning: an annotated system map plus a short explanation of one limitation that interface fluency can conceal.

2. Make a responsible use decision

Learners should be able to connect purpose, consequence of error, data, verification, access, disclosure, and human responsibility.

Suggested preparation: When to use generative AI and Responsible AI use.

Learning activity: compare three scenarios—a low-stakes lesson brainstorm, feedback on identifiable student work, and an admissions recommendation. Ask learners to decide whether to proceed, modify, escalate, or avoid each use and to justify the safeguards.

Evidence of learning: a completed review record that names the purpose, data, consequence, verification, decision owner, access conditions, and stop criteria.

3. Evaluate an output

Learners should be able to separate claims from presentation, verify sources, identify omissions, assess uncertainty, and examine whose perspectives are represented.

Suggested preparation: Teaching students to critically evaluate AI outputs and Why AI detection does not prove authorship.

Learning activity: provide an educator-vetted AI artefact and an approved source pack. Learners map claims to sources, correct one error, identify one important omission, and explain what cannot be concluded from style or a detector score.

Evidence of learning: a claim-to-source table and a justified revision.

4. Apply and evaluate AI in a discipline

Learners should be able to choose an appropriate role for AI, preserve the human learning activity, test the workflow, and evaluate its effects.

Suggested preparation: The practical prompting pattern, Assessing learning in the age of AI, and one relevant teaching workflow.

Learning activity: redesign a real course task with AI-free, AI-limited, and AI-integrated variants. For each, map the outcome, student activity, evidence, risks, access conditions, and human review.

Evidence of learning: a selected design, assignment AI-use statement, test cases, observed failure patterns, and a decision to adopt, revise, or reject the workflow.

Progress from acquire to create

Progression should reflect judgment and responsibility, not the number of tools used or the complexity of a prompt.

Evaluate the learning, not confidence with a product

• Assess transferable concepts rather than memorised interface steps.
• Include tasks where the correct decision is not to use AI.
• Require verification with authoritative sources.
• Test application in an unfamiliar scenario.
• Provide non-AI or institutionally supported access routes.
• Avoid rewarding costly subscriptions, fluent English, or prior exposure.
• Revisit competencies when tools, policy, or disciplinary practice changes.

References & Footnotes