69传媒 need to take a harder look at how the evolution of AI could help or harm students with special needs because use of the technology in K-12 education is expanding faster than many experts predicted, suggests a
鈥淎I & Accessibility in Education"鈥攔eleased Sept. 16 by the Consortium for School Networking, which represents school district tech leaders, and CAST, a group that advocates for universally accessible educational materials鈥攑oints out that most AI technologies are not developed based on the needs of students with disabilities. That could put those students at a disadvantage academically as the use of AI in education continues to expand, experts say.
鈥淎I is spinning so fast, it鈥檚 important to pause and think about specific areas and considerations, like accessibility,鈥� said Christine Fox, the vice president of operations for CAST.
That鈥檚 why district leaders are 鈥渓ooking for a roadmap,鈥� said Stacy Hawthorne, a CoSN board member and the chief academic officer for Learn21, a nonprofit that provides educational technology solutions to schools.
Looking at what we could do with artificial intelligence a year ago and what we can do with artificial intelligence today has been an enormous leap.
The report highlights ways AI could be used more effectively to help students with special needs. For example:
- AI has 鈥済reatly improved鈥� assistive technologies, such as text-to-speech software and speech recognition systems, which enhance the learning experience for students with diverse needs, the report notes. Bruce Alter, an assistive technology consultant for the Tigard-Tualatin and Woodburn school districts in Oregon, has started testing an AI-integrated 鈥渁ugmented and alternative communication鈥� tool to help him interpret the speech of a student with cerebral palsy.
- AI could help alleviate teacher burnout by automating tasks such as grading, attendance tracking, and lesson planning. Those time-saving measures could give teachers more opportunities to work one-on-one or in small groups with students and provide more meaningful academic and social support.
Previous Education Week reporting shows that some special education practitioners have already started experimenting with generative AI to speed up some of their administrative tasks, such as generating Individualized Education Program goals.
But the problem is that many generative AI tools are not built with diverse learners in mind, according to the report. The datasets that AI tools are trained on tend to lack diversity in their samples, the report said. For instance, if an AI grading tool is designed to evaluate written responses based on standard grammatical and stylistic norms, it might unfairly penalize students with disabilities such as dyslexia, who may struggle with spelling and grammar. This could result in lower grades for these students, even if they understand the material.
Another big challenge is that AI tools often require personal data to function effectively, including sensitive information about students鈥� disabilities, which schools do not want to fall into the hands of hackers or other bad actors online, the report said.
School districts need better AI planning for the present and the future
The report鈥攚hich notes that human oversight of AI tools is crucial鈥攐ffers short-term, medium-term, and long-term actions that districts and policymakers can take to create more inclusive and safe learning environments with AI.
- In the short-term, educators at all levels need to , how it functions, and how to use it responsibly. This shouldn鈥檛 just be a one-and-done event, Hawthorne said. It should be ongoing.
- In the medium-term, district leaders should seek expert guidance on how to use AI effectively and appropriately, whether that expertise comes from within the district or from outside sources, such as researchers. They should also ensure that special education is part of the discussion, instead of just general education.
- In the long-term, districts should cultivate a culture of accessibility and inclusivity. Often, the strategies teachers use for special education students are also helpful for general education students, Hawthorne said. Those transferable strategies should be shared across the district.
Chris Smallen, the chief technology officer for Lenoir City schools in Tennessee and a CoSN board member, said his district has already started implementing some of these recommendations. For instance, district officials have been meeting with a scientist who has done a lot of work in AI; they鈥檙e also working on educating staff about how to think about the role of of AI in education.
The challenges, Smallen said, include not having enough staff and time. 鈥淓verybody wants time with our teachers,鈥� he said, so it鈥檚 hard to squeeze in professional development on AI.
Still, he said, he鈥檚 excited about the possibilities for AI.
鈥淟ooking at what we could do with artificial intelligence a year ago and what we can do with artificial intelligence today has been an enormous leap,鈥� Smallen said. 鈥淚鈥檓 looking forward to what that looks like five years from now.鈥�
