Bridging Psychometric and Content Development Practices with AI: A Community-Based Workflow for Augmenting Hawaiian Language Assessments

AI can amplify human expertise while preserving cultural authority: The project successfully used AI tools (NotebookLM and Claude 3.5 Sonnet) to accelerate analysis of 58 educational assessment items while maintaining human oversight at every stage and ensuring Hawaiian language and cultural integrity remained under community control, demonstrating that AI doesn't require replacing human judgment in culturally sensitive contexts.

Systematic oversight frameworks are essential for responsible AI in education: The KĀ'EO AI Policy Framework enabled the team to identify concrete problems—such as linguistic ambiguity and misaligned difficulty levels—while maintaining ethical boundaries, suggesting that clear governance structures can make AI tools safer and more effective for high-stakes educational decisions.

AI-assisted analysis can uncover systemic design issues at scale: The workflow identified six major categories of problems across Hawaiian Language Arts, Mathematics, and Science assessments that might have been missed through manual review alone, showing that responsible AI implementation can improve educational quality for underserved communities without compromising their autonomy.

This model is replicable for other Indigenous language programs: By documenting their community-based workflow, the KĀ'EO team created a template that other Indigenous-language educational programs could adapt, addressing a critical gap since KĀ'EO is the only Native language assessment used for federal accountability in the United States.