It is undeniable that artificial intelligence (AI) is, sooner rather than later, going to impact the work of teaching and learning in special education. Given formal adoption of AI technologies by schools and districts and informal uses of ChatGPT and similar platforms by educators and students, the field of special education research needs to take seriously how advancements in AI can complement and potentially improve our work. But there are also ways that these advancements can go astray. With these technologies advancing so quickly, and with AI models being trained on populations that may not include individuals with disabilities, there is a real risk that AI will fail to improve—or worse, harm—learning experiences for students with disabilities. Therefore, there is a pressing need to ensure that voices from within the special education community are included in the development of these new technologies.
At NCSER, we are committed to investing in research on AI technologies in a way that privileges the expertise of the special education community, including researchers, educators, and students with disabilities and their families. Below, we highlight two NCSER-funded projects that demonstrate this commitment.
Using AI to support speech-language pathologists
In 2023, NCSER partnered with the National Science foundation to fund AI4ExceptionalEd, a new AI Institute that focuses on transforming education for children with speech and language disorders. Currently, there is a drastic shortage of speech-language pathologists (SLPs) to identify and instruct students with speech and language needs. AI4ExceptionalEd brings together researchers from multiple disciplines including special education, communication disorders, learning sciences, linguistics, computer science, and AI from nine different universities across the United States to tackle pressing educational issues around the identification of students and the creation of specially designed, individualized instruction for students with speech and language disorders.
By bringing together AI researchers and education researchers, this interdisciplinary research partnership is setting the foundation for cutting-edge AI technologies to be created that solve real-world problems in our schools. A recent example of this is in the creation of flash cards for targeted intervention. It is common practice for an SLP to use flash cards that depict a noun or a verb in their interventions, but finding or creating the exact set of flash cards to target a specific learning objective for each child is very time consuming. Here is where AI comes into play. The Institute’s team of researchers is leveraging the power of AI to help SLPs identify optimal sets of flash cards to target the learning objectives of each learner while also creating the flash cards in real time. To do this effectively, the AI researchers are working hand-in-hand with speech and language researchers and SLPs in the iterative development process, ensuring that the final product is aligned with sound educational practices. This one AI solution can help SLPs optimize their practice and reduce time wasted in creating materials.
Adapting a popular math curriculum to support students with reading disabilities
Another example of how partnerships can strengthen cutting-edge research using AI to improve outcomes for students with disabilities is a 2021 grant to CAST to partner with Carnegie Learning to improve their widely used digital math curriculum, MATHia. The goal of this project is to develop and evaluate reading supports that can be embedded into the adaptive program to improve the math performance, particularly with word problems, of students with reading disabilities. CAST is known for its research and development in the area of universal design for learning (UDL) and technology supports for students with disabilities. Carnegie Learning is well known for their suite of curriculum products that apply cognitive science to instruction and learning. The researchers in this partnership also rely on a diverse team of special education researchers who have expertise in math and reading disabilities and an educator advisory council of teachers, special educators, and math/reading specialists.
It has taken this kind of partnership—and the inclusion of relevant stakeholders and experts—to conduct complex research applying generative AI (ChatGPT) and humans to revise word problems within MATHia to decrease reading challenges and support students in understanding the semantic and conceptual structure of a word problem. Rapid randomized control trials are being used to test these revised versions with over 116,000 students participating in the study. In 2022-2023 the research team demonstrated that humans can successfully revise word problems in ways that lead to improvements in student performance, including students with disabilities. The challenge is in trying to train generative AI to reproduce the kinds of revisions humans make. While generative AI has so far been unevenly successful in making revisions that similarly lead to improvements in student outcomes, the researchers are not ruling out the use of generative AI in revising word problems in MATHia.
The research team is now working with their expert consultants on a systematic reading and problem-solving approach as an alternative to revising word problems. Instead of text simplification, they will be testing the effect of adding instructional support within MATHia for some word problems.
The promise of AI
AI technologies may provide an opportunity to optimize education for all learners. With educators spending large amounts of their day planning and doing paperwork, AI technologies can be leveraged to drastically decrease the amount of time teachers need to spend on this administrative work, allowing more time for them to do what only they can—teach children. Developers and data scientists are invariably going to continue developing AI technologies, many with a specific focus on solutions to support students with disabilities. We would like to encourage special education researchers to exert their expertise in this development work, to partner with developers and interdisciplinary teams to apply AI to create innovative and novel solutions to improve outcomes for students with disabilities. For AI to lead to lasting advances in education spaces, it will be imperative that this development is inclusive of the special education field.
This blog was written by NCSER Commissioner, Nate Jones (Nathan.Jones@ed.gov) and NCSER program officers Britta Bresina (Britta.Bresina@ed.gov) and Sarah Brasiel (Sarah.Brasiel@ed.gov).