Risk and Reward: Artificial Intelligence in K–12 Education
Dr. Stacy Hawthorne is a member of CoSN’s board of directors as well as the Chief Academic Officer for Learn21, a non-profit organization based in Ohio that provides educational technology support to schools. Currently, Dr. Hawthorne and her team are assisting school districts in Ohio to implement Artificial Intelligence (AI) tools in their schools. CoSN got in touch with her to discuss the implications of AI as the means to enhance accessibility in education.
Opportunities and Hurdles Using AI for Accessible Education
One of the most common promises of AI in education is its potential to personalize materials for students. From readability to content structure, teachers could be able to adapt their class to any student’s needs. Regarding accessibility, this means an enormous step forward, as appropriate personalization could ensure that students can access and engage in all the learning opportunities available in their school.
However, Dr. Hawthorne highlights two challenges to achieve this goal: “Where I’m skeptical is in professional development and current curriculum mandates being able to adapt and turn around fast enough to take advantage of the promise of AI in education.”
Is policy design critical at the local level?
After the U.S. Department of Education issued a report around AI and education, there has been an urge from other education institutions to follow suit and create AI policy that aligns with curriculum standards. However, there is a growing concern about the extent to which policies regarding AI at the local level are a priority at this moment. Currently, AI-integrated tools are constantly transforming technology solutions while educational actors are still in the process of learning what AI is, how to implement AI tools in their practice, and how the latter can benefit their students the most.
By creating a narrow policy that mandates a specific use of AI, there is a risk of limiting the educators’ possibilities to explore the potential of AI-integrated tools for only the specific needs in their classroom. In addition, there might already be policies in place at the district level for the common concerns in AI use, “ […] we might not use the term AI in those current policies, but we have data privacy policies, we have cybersecurity policies. They are going to apply in an AI world”, Dr Hawthorne says. By being aware of the latter, leaders can focus on creating flexible guidelines and manuals of best practices that can promptly adapt to the future uses of AI as an Assistive Technology.
The possibilities and limits of AI in the classroom
Dr. Hawthorne suggests that at this moment it is necessary to prioritize providing teachers with practical examples of scenarios of students with different learning needs and how AI can become a useful form of Assistive Technology for them. For instance, there are AI tools that create real-time captions and notes of a class or media content, which can help students focus on a single task and engage in discussion rather than multitasking. Although this is beneficial for all students, it could specifically improve the learning experience of students with Attention-Deficit/Hyperactivity Disorder (ADHD), who tend to find multitasking more mentally and physically draining (Ewen et al., 2012).
Another extensively debated use of Generative AI is its potential to personalize Individualized Education Programs (IEPs) while reducing the teachers’ paperwork burden. While current AI tools can generate documents like IEPs for students with disabilities, there are limitations to consider, especially regarding the curricular content that the algorithm can incorporate into the IEP.
“The “I” stands for Individualized. So, I appreciate the set of knowledge that AI has, and it could […] help prompt a special education teacher’s thinking about some accommodations, but asking it to flat out write the IEP? Sure, it’s going to save time, but we lost the “I” in that process.”
At this moment, the algorithm cannot recognize the diverse needs of all students. Consequently, unique accommodations and modifications for students cannot be easily standardized. For example, students on the autism spectrum exhibit a wide range of interaction behaviors, speech patterns, restricted interests, and other traits (National Institute of Mental Health, n.d.) that cannot be easily quantified or described in a prompt. If educators prompt a GenAI tool to generate an IEP for all students on the autism spectrum, the tool will produce a document that overlooks these specific characteristics.
Adam Garry, the president of StrategicEDU consulting, proposes an alternative. In the near future, school districts should establish hybrid models that ensure human involvement in the IEP design process. Through this model, teachers can prompt a GenAI tool to do 80% of the main IEP outline while they fill the remaining gap on their own with details from the learner that ensure the plan is specific to their needs.
Looking into the future
Through ongoing professional development and support in implementing AI in schools, identifying specific challenges affecting a student’s progress and providing alternatives to enhance their learning will benefit not only IEP-eligible students but all learners. Dr. Hawthorne illustrates the limited access to IEPs by comparing the experience of her son and her daughters, “I wish his sisters could have had an IEP. They weren’t eligible for one. Not because I wanted them to have a learning disability, but because I wanted somebody to care as much about personalizing instruction for them as they did for my son.”
Flexible policies at the local level are crucial to accommodate the evolving nature of AI tools and ensure they meet diverse student needs. As we move forward, incorporating AI into education should be done with careful consideration of both its benefits and limitations, ensuring that human oversight and personalized approaches remain central to effective teaching.
References:
Ewen, J. B., Moher, J. S., Lakshmanan, B. M., Ryan, M., Xavier, P., Crone, N. E., Mahone, E. M. (2012). Multiple Task Interference is Greater in Children With ADHD. Developmental Neuropsychology, 37(2), 119–133. https://doi.org/10.1080/87565641.2011.632459
Author: Fernanda Pérez Perez, CoSN’s 2024 Blaschke Fellow
This is the second blog in a series by Fernanda Pérez Perez; this blog serves as a companion to the forthcoming comprehensive report on AI and Accessibility, slated for release this Fall. Read the first blog here.
Learn more about the Charles Blaschke EdTech Fund.
