Inclusive by design: Stanford initiative releases new paper that explores how AI can support learners with disabilities

reprint from Stanford.edu

A new white paper from the Stanford Accelerator for Learning highlights how artificial intelligence (AI) can be a powerful tool to support students with learning differences — but only if it is developed with their needs and voices at the center.

The report, “AI + Learning Differences: Designing a Future with No Boundaries,” is the result of a two-day Working Symposium and Hackathon hosted by the Accelerator that brought together more than 100 students, educators, researchers, entrepreneurs, policymakers, education industry leaders, and philanthropists. Participants, including many individuals with learning differences, spent the first day of the event engaging in collaborative dialogue on AI and learning differences and the second day designing and testing innovative new tools.

The conversations and hacking sessions at the symposium were grounded in the disability rights principle, “nothing about us without us,” emphasizing the importance of including people with lived experience in the design of emerging technologies. As artificial intelligence becomes increasingly integrated into education and everyday life, participants reflected on the assumptions being built into AI systems and who is being prioritized—or overlooked—in their development.

Discussions explored how AI could better support learners with differences, including helping identify where students may need additional support, enhancing assistive technologies, and improving the development and implementation of Individualized Education Plans (IEPs). Participants also explored how AI might impact students’ social and emotional well-being, and whether it can be designed to recognize and value the full spectrum of human intelligence, even when it doesn’t conform to traditional academic norms.

The paper, released on July 21, synthesizes the outcomes of the symposium and hackathon, presenting a vision of inclusion in an AI-infused world. It lays out 12 recommendations for developers, educators, researchers, and policymakers to ensure that AI systems equitably serve learners of all abilities.

“Empathy and access must not be afterthoughts,” said Elizabeth Kozleski, co-author and faculty co-director of the Accelerator’s Learning Differences Initiative. “This paper outlines how AI can be developed and deployed in ways that recognize the needs of all learners.”

Inclusive design benefits everyone

Designing AI with the needs of people with disabilities in mind doesn’t only support a specific population—it improves the experience for all learners. For example, real-time captioning supports both hard of hearing students, English language learners, and learners with auditory processing challenges. AI-driven tools that help with executive functioning can also assist office personnel managing logistics. Starting with accessibility leads to smarter, more adaptable systems that serve a wide range of users.

“When we design for the full range of human experience, we build better systems for everyone,” said Chris Lemons, co-author and incoming faculty director of the Learning Differences Initiative. “Tools created with learners with disabilities in mind often reveal gaps in the system that affect many others—like students who are multilingual, under-resourced, or just learn differently. Inclusive design is not a constraint—it’s a catalyst for better technology and stronger educational equity.”

Key recommendations include:

• Co-design with individuals who have learning differences
• Professional development for educators on integrating AI equitably
• Cross-sector collaboration to ensure inclusive policy and funding models
• Prioritization of privacy, transparency, and student agency

The white paper is accompanied by a Hackathon Toolkit to support schools, companies, and community organizations in creating inclusive AI tools with students, families, and educators.

“In many schools, disability is thought of as a problem and teaching solutions are based on ‘fixing kids,’” reflected 7th-grade student Mae T., who served as a hackathon judge.

“I have seen that happen at my school…at the meeting, I began to really understand that instead, the problem is our ideas about disability and teaching solutions should be based on fixing our ideas about learning differences…because everyone is different.”

DOWNLOAD THE WHITEPAPER

The symposium and hackathon were made possible through the generosity of the Alana Foundation and in collaboration with CAST, the Stanford Institute for Human-Centered Artificial Intelligence (HAI), and the Children’s Health Council (CHC).