Joint committee holds informational briefing on artificial intelligence in education

The Joint Committee on Higher Education and the House Committee on Education convened an informational briefing on artificial intelligence in education, featuring classroom demonstrations by an educator, policy guidance from the Hawaii Department of Education and technical and governance perspectives from University of Hawaii at Manoa officials.

The session matter-of-factly addressed near-term classroom uses, privacy and bias risks, workforce implications and equity. Gabriel Yanagihara, an educator and consultant who has run classroom pilots across the islands, told lawmakers he “completely replaced myself with AI as an experiment” to learn quickly what works and what fails. Heidi Armstrong, deputy superintendent for academics at the Department of Education, described department guidance and professional development intended to let teachers “use [AI] as a tool to support learning, absolutely not replace any teachers.” Provost Bruno and University of Hawaii colleagues emphasized data governance, research integrity and economic-development opportunities tied to AI adoption.

Yanagihara described hands-on student projects and teacher training he has led: AI-generated community “postcards from the future,” student-illustrated books (including a tsunami-safety book produced with Hawaii Emergency Management), and a summer video-game design class that used image-generation tools to help students iterate and revise. He said those efforts have reached teachers and students across Maui, Molokai and other islands and that he has trained “over 2,500 teachers across the state.” Yanagihara urged structured literacy lessons so students learn to use AI ethically and to revise their own work rather than rely on AI to produce final drafts.

Armstrong outlined the department’s approach: a public guidance website and training portfolio for staff and students that covers data protection, human oversight, fairness and citation practices. She described pilots the department is using to evaluate tools, including a Conmigo tutoring pilot that involved about 400 students, and a pilot called Magic School that bundles chat and lesson-planning tools for classroom pilots. Armstrong said the department emphasizes a staged rollout, templates for data-sharing agreements, and both in-person and asynchronous professional development modules (prompt engineering, AI for assessment, conversational AI, and related topics).

University of Hawaii officials framed adoption under five pillars—governance and ethics, outreach and engagement, research and innovation, workforce development, and economic development—and stressed existing data-governance and privacy frameworks that must be applied to AI. University speakers said they have worked with Google to deploy Gemini under the hawaii.edu domain with enterprise protections and are offering a Google AI essentials online course through Coursera for campus users. They also cited large research grants and cross-campus projects that used generative AI to accelerate course design and research work.

Committee members used the question period to press presenters on specific risks and operational issues: Rep. Chica asked about data governance and how the DOE detects and responds when vendor services move data offshore; Dr. Winston Sakurai of the DOE said staff monitor server locations and, in one case, discontinued a service when it sought to move data to a jurisdiction of concern. Rep. Capella asked about the environmental and power requirements of large AI models; university speakers said training and serving large models are energy-intensive and typically run in large data centers, underscoring the need for more energy-efficient hardware and distributed (edge) approaches.

Other questions covered admissions integrity (how universities will evaluate application essays in an era of AI assistance), social-emotional impacts of conversational agents on students, and strategies to ensure equitable access so rural and lower-resource schools can benefit. University speakers and DOE staff said tools for detecting AI-derived text exist but that the research-integrity landscape is evolving; they emphasized holistic admissions reviews and ongoing pilots rather than single technical fixes.

No formal votes or policy decisions were taken; presenters and committee members described ongoing pilots, guidance and planned next steps. University officials noted proposed legislation to formalize an AI institute to coordinate research, workforce development and outreach; the DOE said it will continue scaling pilots and professional development. The briefing closed with members urging continued collaboration among teachers, the department and the university.