Decolonizing Artificial Intelligence: AI and Education from the Perspectives of Indigenous Peoples

• In Event: Encountering the pluriverse through Indigenous pedagogies

Proposal

Abstract
Artificial Intelligence is transforming global education, yet it often reinforces Eurocentric worldviews that marginalize Indigenous knowledge, culture, and values. This paper examines AI in education through Indigenous epistemologies and ethical frameworks. Drawing from a study involving Indigenous participants in six countries, Indigenous people asked AI chatbots specific questions, and their responses were critically reviewed by community members. Analysis revealed key issues, including misinformation, misrepresentation, bias, and violations of data sovereignty. These findings highlight the urgent need for Indigenous-led AI development to ensure ethical, culturally responsive, and inclusive AI systems. The paper concludes with actionable recommendations for decolonizing AI, supporting Indigenous rights, knowledge sovereignty, and epistemic diversity.
Introduction
AI is increasingly reshaping global education systems, influencing how students learn, teachers instruct, and institutions assess outcomes. However, beneath the surface of these innovations lies a serious concern: most AI systems are trained and deployed within frameworks that reflect Western and/or dominant group epistemologies, often disregarding or marginalizing Indigenous knowledge systems, ways of knowing, languages, and pedagogies.
Methods:
This study investigates the intersection of AI, decolonization, and education from the perspectives of Indigenous Peoples. To explore AI from the perspectives of Indigenous communities, Indigenous peoples asked AI chatbots specific questions, and their responses were critically reviewed by community members. The questions include knowledge about cultural and historical events, knowledge of land, water, medicinal plants, and Indigenous ways of knowing. Since the questions are community specific and deeply embedded knowledge, elders and scholars from these communities are most knowledgeable to review AI chatbot responses. Then the Indigenous community members critically examines the AI chatbots responses, create a brief summary based on their findings. Analysis of their findings from multiple Indigenous communities provide the foundation of our understanding of the AI chatbot response as they related to the issues of reliability, ethics, bias, and privacy.
Findings:
Case study 1: Marma People of Chittagong Hill Tracts (CHT), Bangladesh, Kankanaey People in Benguet Province, the Philippines, and Chumash people in the United States. They used three AI chatbots (OpenAI ChatGPT, Microsoft Copilot, and Google Gemini). The queries include ceremony, oral stories, medicial plants, Manbuang system of knowledge, Kankanaway system of learning, etc.. More often than not, the AI chatbot responses include partially incorrect information or misinformation. More importantly, in several cases, AI chatbots provided incorrect information with high confidence in their responses. Chatbots provided biased information from the dominant majority point of view. AI chatbot responses also showed a lack of data privacy. These findings highlight the importance of ethical and responsible use of AI, data sovereignty, and culturally responsive curricula design in ensuring that AI serves as a tool of empowerment rather than misinformation and assimilation.

Case Study 2: Indigenous Language Models (ILMs) in Brazil. Linguists partnered with Indigenous communities to create language-specific AI tools, including spell-checkers and predictive text systems for endangered Indigenous languages (Cardoso et al., 2024). These models were trained on oral stories, transcribed interviews, and community-approved materials.
Highlights:
• Collaborative model development, not extractive.
• AI tools were tailored to local linguistic structures unfamiliar to general-purpose models.
• Focused on language preservation and intergenerational learning through digital platforms.
Even in low-resource environments, AI can amplify Indigenous literacy and education—but only when communities have a seat at the design table.

Case Study 3: Ethical AI for Early Childhood Education South Africa where researchers have critiqued the use of AI in early childhood developments as perpetuating colonial learning models. In response, local initiatives began embedding Indigenous languages, folktales, and child-centered values into AI-driven educational platforms (Kalinga, 2021). Highlights:
• AI must accommodate cultural learning styles, such as storytelling, reciprocity, and rhythm.
• Local educators, parents, and elders must co-create content and learning pathways.
• Generic AI learning models are insufficient for culturally responsive pedagogy.
The tension between globalized AI tools and localized pedagogical values, and the potential for community-authored systems to bridge that gap.

Based on the above case studies and supporting literature, the following themes emerge:
1. Indigenous-Led AI Projects Are Essential for Culturally Relevant Education
Across all cases, success hinged on community control and authorship—from data governance to algorithm design. Where AI was used to support Indigenous language revitalization, the tools were only effective when embedded in cultural protocols and led by Indigenous stakeholders.
2. Mainstream AI Models Reinforce Colonial Structures if Left Unchecked
Most general-purpose AI tools lack the nuance to represent Indigenous knowledge accurately. They tend to universalize Western pedagogies, overlook oral and land-based learning modes, and often prioritize dominant languages—thus undermining educational sovereignty.
3. Decolonizing AI Requires Epistemological Pluralism
Indigenous frameworks offer essential methodologies for rethinking AI in education. These frameworks allow coexistence between Indigenous and Western systems without subsuming one under the other, enabling more inclusive and dynamic learning technologies.
4. Data Sovereignty Is a Precondition for Ethical AI Use in Indigenous Education
Consent, transparency, and control over data and models are non-negotiable. Initiatives that respected Indigenous data sovereignty—like Te Hiku—produced more sustainable, trustworthy, and impactful outcomes compared to externally imposed models.
5. AI Can Empower Intergenerational Learning and Language Preservation
In communities where language and cultural transmission are threatened, AI can serve as a digital amplifier of Indigenous knowledge, especially when co-designed for intergenerational use.

Decolonizing AI is not a one-time intervention but an ongoing process that demands collaboration, humility, and a re-centering of Indigenous worldviews in technological development. It is a call to reimagine educational futures where AI supports, and not supplants, Indigenous ways of knowing and being.
Drawing from Smith’s (2021) foundational work, decolonizing AI demands a recognition of Indigenous Peoples not just as users or informants, but as theorists, researchers, and technologists in their own right. AI in education must be situated within Indigenous research paradigms that prioritize collective benefit, healing, accountability, and land-based knowledge systems. These perspectives reshapes not only what AI can do, but also what it should do, and for whom.
In summary, decolonizing AI is not about better diversity metrics—it is about restoring relationships, challenging power, and affirming Indigenous futurities.

Author

• Maung Ting Nyeu, University of California, Santa Barbara