- The paper presents a framework that redistributes epistemic authority from AI systems to community-based knowledge through processes like epistemic fine tuning.
- It employs constructionist and culturally sustaining methodologies to validate local expertise over universal AI outputs.
- Findings indicate that participatory, context-sensitive practices can enhance critical AI literacy and promote educational equity.
Introduction
"Community-Based AI Learning: Redistributing Artificial Intelligence's Epistemic Authority in Education" (2604.21986) critiques dominant paradigms in AI-mediated education, arguing they inappropriately situate AI systems as epistemic authorities at the expense of learners' lived and community-based epistemologies. Rooted in constructionist and community-driven learning traditions, this paper theorizes a framework where community knowledge operates as the primary site of epistemic legitimacy, challenging universalist knowledge regimes embedded in AI and foregrounding processes of epistemic fine tuning, redistribution of authority, and situated discernment. This orientation fundamentally repositions AI from a site of authoritative knowledge delivery to one of negotiated, community-anchored interpretation, with far-reaching implications for equity and the development of critical AI literacy.
Critique of Epistemic Authority in Mainstream AI Education
Contemporary AI education predominantly frames AI/ML systems as reliable instructional supports, such as tutors, scaffolds, or analytical tools. These systems supply confident, institutionally-endorsed outputs that are frequently received as complete and provisional sources of truth, particularly in contexts where learners lack prior expertise or epistemic confidence. The paper underscores that this uncritical deference institutionalizes epistemic hierarchies, authorizing AI as a "view from nowhere" (Haraway) while marginalizing local, Indigenous, and Global South knowledge systems.
Empirical research further demonstrates that learners, especially those whose epistemic beliefs privilege institutional authority, tend to integrate AI-generated content (e.g., from ChatGPT) into academic work indiscriminately, including inaccurate outputs [(2604.21986); 48]. The underlying issue is not simply factuality, but how epistemic authority is configured when AI systems become actors in pedagogical processes—an issue rarely interrogated in extant models of AI literacy, which tend to treat sociopolitical analysis as a matter of abstract bias or generic fairness.
The proposed framework draws from three core traditions:
- Community-Based Science Education: Knowledge is negotiated through relationships anchored in place, cultural history, and shared lived experience. Expertise is heterogeneous and includes non-institutional, community-rooted ways of knowing.
- Critical and Culturally Sustaining Pedagogy: The framework not only leverages community knowledge for enrichment but also insists on its primacy as the evaluative standard for all sources, including AI.
- Constructionism: Learning is materialized and socially negotiated through the construction of public artifacts that externalize both understanding and epistemic stances towards technology.
Community-based AI learning is thus conceptualized as an epistemic process where learners utilize their lived and community-grounded experience to critically interrogate, contextualize, and, when necessary, contest AI-generated outputs. Knowledge construction extends beyond cognitive understanding into a relational and sociopolitically situated activity, repositioning epistemic agency at the intersection of community, culture, and technology.
Key Processes: Epistemic Fine Tuning, Redistribution of Authority, and Situated Discernment
The operationalization of the framework centers on three intertwined commitments:
Epistemic Fine Tuning
Distinct from technical model fine tuning, epistemic fine tuning tasks learners with continually recalibrating their stance toward AI outputs. Trust in AI is rendered conditional, grounded in lived experience and local expertise rather than the presumptive authority of model outputs. In practice, this means learners treat AI claims as decontextualized and partial, actively scrutinizing them against local realities and community histories.
This process directly addresses asymmetries in epistemic authority generated by the dominance of Eurocentric/westernized datasets within LLM training, reversing the tendency to relegate community ways of knowing to secondary status [(2604.21986); 10, 42, 52].
Redistribution of Authority
The framework demands a reordering of power in knowledge-making: AI becomes one interpretive resource among many, and learners' communities occupy the role of primary epistemic agents. Authority is reframed as relational and context-sensitive—students, practitioners, and elders with localized expertise are placed at the center of evaluative judgment. This approach simultaneously requires and enables critical AI literacy to become locally-embedded rather than abstract or universally-defined.
Situated Discernment
Instead of prescriptive or uniform engagement with AI, the framework supports collective decision-making about when and how AI should be incorporated into learning or design processes—and under what conditions it should be resisted or excluded. This is particularly vital where algorithmic infrastructures materially shape, surveil, or disadvantage marginalized communities. Situated discernment is thus both an educational and civic process, guiding learners and communities to deliberate about technological adoption and its real-world ramifications.
Practical Implications and Implementation
For computing educators, curriculum designers, and policymakers, the adoption of community-based AI learning implies substantive shifts in pedagogical design and epistemic practices:
- Instructional activities should be anchored in community-driven phenomena, with learners and practitioners externalizing and critically examining their relationships to AI through project-based, participatory construction.
- Evaluative authority must move from algorithmic outputs to community sites, with local practitioners and family members empowered as co-judges of relevance and accuracy.
- Critical AI literacy instruction should move past abstract bias mitigation and interrogate concrete consequences of AI infrastructure on local social, cultural, and economic conditions.
- Methodologically, research efforts must prioritize participatory designs, analytic approaches, and partnerships that center place, geography, and lived experience as constitutive of AI learning.
This approach is explicitly modular—it coexists with technical instruction but adds layers of ethical calibration, reflexivity, and technoskepticism, equipping learners to negotiate and contest the authority and limits of AI systems in concrete community contexts.
Theoretical and Practical Implications for AI Development
Practically, this framework foregrounds the necessity of co-design and participatory engagement in the development and deployment of AI in educational spaces. It cautions against the uncritical importation of algorithmic solutions and highlights the risks of epistemic injustice that accompany the universalization of dominant cultural frames.
Theoretically, the framework challenges dominant epistemologies of AI—those built on presumed neutrality and transferability—by demonstrating their limitations in pluralist, historically-situated social worlds. It suggests that future AI systems and pedagogical models must build infrastructural capacity for epistemic humility, adaptability, and responsiveness to diverse communities. This may imply the development of more localized or context-aware AI systems, the integration of participatory feedback mechanisms, and ongoing scrutiny of whose knowledge is encoded, amplified, or erased by algorithmic infrastructures.
Conclusion
The paper posits that equitable and critically-grounded AI education is unattainable without a redistribution of epistemic authority from AI systems to the communities experiencing their impacts. Community-based AI learning provides a robust framework for recalibrating trust, authorizing community knowledge, and enabling differential and context-sensitive engagement with AI. As AI further integrates into educational infrastructure, this approach offers a path toward computational empowerment that is both epistemically just and democratically responsive to the lived realities of learners and their communities (2604.21986).
