Artificial intelligence (AI) is increasingly embedded in educational practice, yet

AI literacy is often defined through predefined competencies that say less about

how it is understood, enacted, and shaped in specific contexts. This research

investigates how AI literacy is conceptualised in research and policy, understood

by teachers, enacted through interaction with generative AI (GenAI), and used

to inform the design of educational GenAI systems that support reflective and

critical engagement.

The methodology combines Emergent Research Design structured with De-

sign Research Methodology. This allows the research focus to develop in response

to empirical findings, methodological insights, and rapid changes in the educa-

tional AI landscape, while maintaining a coherent progression from problem

clarification and descriptive inquiry towards exploratory design-oriented knowl-

edge. Adopting a practice-oriented perspective, the thesis synthesises conceptual,

empirical, interactional, and design-oriented inquiry across eight studies.

Findings show that AI literacy is frequently framed in operational and

competency-oriented terms, while sociotechnical, critical, and agentive dimen-

sions remain less developed. Teachers’ understandings of AI are often shaped by

fragmentedandinformallearning,althoughtheirconcernsextendtotrust,respon-

sibility, authorship, bias, and pedagogical judgement. Technical understanding

does not automatically lead to broader sociotechnical reasoning, and recognising

AI-related risks is not equivalent to acting critically in practice. Teacher–GenAI in-

teraction reveals AI literacy as situated and relational: it becomes visible through

how teachers delegate tasks, interpret and adapt outputs, question suggestions,

exercise judgement, and negotiate responsibility.

The thesis further argues that literacy is increasingly bound up with machine

processes. Meaning-making, knowledge production, and pedagogical action

are no longer exclusively human activities but emerge through relations among

teachers, learners, AI systems, infrastructures, and institutional conditions. This

resonates with critical posthumanist accounts of literacy and shifts attention from

what individuals know about AI to how humans and non-humans jointly shape

what can be known and done.

Supporting critical AI literacy therefore requires attention to both education

anddesign. EducationalGenAIsystemsshouldbeevaluatednotonlyforaccuracy,

usability, and efficiency, but also for whether they support reflection, surface

uncertainties and question assumptions. The thesis contributes a situated account

of AI literacy and an interaction-informed rationale for educational GenAI system

design rather than a universal literacy framework or validated design solution.