AI Literacy: An Exercise in Power-Knowledge

Synthesis: Argues existing AI literacy frameworks, dominated by technical competency and responsible-use principles, enforce a consumer orientation toward AI rather than fostering genuine epistemic agency. Draws on Foucault's power-knowledge framework to propose a critical AI literacy that empowers learners to shape and challenge AI systems rather than merely use them.

Key Findings

• The paper argues that dominant AI literacy frameworks — dominated by technical competency and responsible-use principles — enforce a "consumer" orientation toward AI rather than fostering genuine epistemic agency.
• Drawing on Foucault's concept of power-knowledge, the authors contend that the absence of power as a construct in AI literacy discourse reflects a deeper conceptual failure: treating AI systems as neutral tools rather than as apparatuses that structure what can be known and by whom.
• Grounded in Freire's Pedagogies and Teaching Strategies of critical consciousness and scholarship on digital literacy, the paper reconceptualizes AI literacy as a critical practice that equips individuals not just to use AI systems but to critically evaluate them, resist their structuring assumptions, and participate in their governance.
• Unequal access to AI tools, the authors argue, recapitulates longstanding epistemic injustices, so a literacy framework oriented toward empowerment must account for these structural inequities.
• A three-part framework — contextual use, critical interrogation, and participatory governance — frames AI literacy as the cultivation of epistemic "agents" rather than the training of competent consumers of AI-generated information.

Theoretical Foundations

Generative AI has emerged not only as a new class of technologies but as an infrastructure for the creation and dissemination of knowledge, embedded in search engines, writing tools, research platforms, and industries including education. The authors argue that most institutional approaches reduce complex epistemic and ethical questions to technical proficiency: understanding how these models work, evaluating their outputs for accuracy, and using the tools "responsibly." Against this, the power-knowledge lens asks who can access and shape AI-mediated knowledge production, and how literacy frameworks either reproduce or challenge existing distributions of epistemic authority.

What this means for practice

• Instructors. Teach generative prompting before you teach answer retrieval: have learners use AI in iterative dialogue to develop and refine their own knowledge rather than to outsource it.
• Assign counter-prompting as an exercise. Students ask a system for the dominant view on a contested question, then a dissenting view, then ask which perspectives are absent from its training data, and write up what the system could not see.
• Treat critical interrogation of AI as curricular content in its own right rather than a distraction from efficient tool use; the framework positions learners as epistemic agents whose questioning of a system's structuring assumptions belongs in the syllabus.
• Audit who can actually access which AI tools in your setting before assuming a skills-based intervention will close gaps: inequity here is structural, not individual, because unequal access recapitulates epistemic injustice.
• Give learners a real route into AI Governance — co-designing course or institutional AI policy, for example — since participatory governance is the dimension the paper identifies as the most neglected in existing AI literacy frameworks.

Limitations

• This is a conceptual paper built on Foucault's power-knowledge and Freire's pedagogy of critical consciousness; it presents no data, no case study, and no empirical test of the Contextual Use–Critical Interrogation–Participatory Governance framework.
• The critique of existing AI literacy frameworks is argued rather than measured: the authors characterize that literature as competence- and compliance-oriented but do not systematically sample or code the frameworks they reject.
• The framework is deliberately anti-hierarchical — described as "not a ladder but a practice" — so it supplies no proficiency levels, sequencing, or completion criteria that a program could adopt directly.
• Claims about access and educational capital rest on cited secondary evidence rather than new measurement, and the paper does not test whether the framework changes learner behavior or epistemic agency.

Citation

Brady D. Lund, Zoë Abbie Teel (2026). AI Literacy: An Exercise in Power-Knowledge.