Community-Based AI Learning: Redistributing Artificial Intelligence's Epistemic Authority in Education

Synthesis: A perspective paper introducing community-based AI learning as a framework that repositions epistemic authority, grounding AI engagement in learners' lived and community-based epistemologies rather than treating AI systems as authoritative knowledge sources. Drawing on community-driven learning and constructionist traditions, the framework rests on three commitments—epistemic fine-tuning, redistribution of authority, and situated discernment—that together localize critical AI literacy by calibrating trust, foregrounding community knowledge, and supporting collective judgment about when to design with, interrogate, or reject AI. The authors argue that equitable AI education requires negotiating authority through place, history, and social context.

Key Findings

• Generative AI systems are frequently treated as authoritative knowers in classrooms: their confident, fluent outputs and widespread institutional adoption grant them "epistemic authority," yet this authority is not neutral but historically structured through power relations that marginalize community-based, Indigenous, and Global South epistemologies in favor of dominant Eurocentric formations.
• Dominant AI-education paradigms design AI as instructional agents (tutors, co-learners), implicitly positioning AI systems as default reference points for what counts as correct, relevant, or complete knowledge—particularly where learners lack prior expertise.
• Existing equity-oriented work (computational empowerment, critical AI literacy, youth AI auditing and design) redistributes technical agency but often stops short of centering learners' communities as primary epistemic resources for interpreting AI systems.
• Community-based AI learning is defined as an approach to AI education in which learners engage with AI through issues, practices, and forms of knowledge grounded in their communities, treating local experience and lived realities as central resources for learning with and about AI.
• The framework rests on three commitments: epistemic fine-tuning (learner-centered calibration of how AI outputs are interpreted and trusted, comparing them against lived experience, local histories, and community expertise); redistribution of authority (refusing AI as ultimate authority, positioning learners and communities as legitimate knowers, and cultivating technoskepticism); and situated discernment (collective examination of how AI infrastructures intersect with social worlds, with criticality shaped by place, geography, and local histories).
• In practice, learners design something real for their community using AI as a design resource while community knowledge and community practitioners serve as the evaluative standard; the approach can function modularly alongside technical instruction.

What this means for practice

• Instructors. Have students evaluate AI outputs against locally grounded criteria of relevance, harm, and usefulness, treating confident answers as partial and decontextualized claims rather than as reference points for what counts as correct.
• Instructors. Structure a unit in which students design something real for their community — a restaurant concept, a neighborhood tool, a local cultural archive — using AI as a design resource while community knowledge serves as the evaluative standard.
• Instructors. Sequence a community-site visit before AI engagement and bring community practitioners in as co-evaluative judges, so local expertise rather than the model sets the criteria for a good outcome.
• Administrators. Shift critical AI literacy from generalized awareness of sociopolitical issues toward situated engagement with AI's consequences for a specific community, and leave room for limitation, refusal, or strategic non-use where critical engagement requires it.
• Researchers. Treat place, infrastructure, and lived experience as constitutive dimensions of AI learning rather than as context variables, and report whether the three commitments transfer across settings through the practice of epistemic discernment.

Limitations

• This is a perspective paper with no empirical study behind it: the three commitments (epistemic fine-tuning, redistribution of authority, situated discernment) and the classroom unit described are illustrative, so nothing here shows that the approach changes learner or community outcomes.
• The framework's warrant is axiological rather than measured — it argues from epistemic justice and from the claim that community knowledge is the kind of knowing AI cannot produce, and neither premise is tested.
• It generalizes from community-based science education and constructionist traditions; no analysis is offered of how the approach performs in the settings where epistemic marginalization is most acute or how it interacts with mandated technical curricula.
• Modular adoption "alongside technical instruction" is asserted rather than costed: the paper gives no account of the teacher time, community partnerships, or coordination a unit requires, or of how community co-evaluation would scale beyond a single unit.

Citation

Ojeda-Ramirez, S., Gyles, S., & Peppler, K. (2026). Community-Based AI Learning: Redistributing Artificial Intelligence's Epistemic Authority in Education. Proceedings of the 2026 Conference for Research on Equitable and Sustained Participation in Engineering, Computing, and Technology (RESPECT 2026). ACM.