OSEP Tech Meeting

How to make AI accessible for people with a range of disabilities?

The critical shortage of Speech-Language Pathologists (SLPs) is hindering timely intervention for millions of children with speech and language disorders, particularly in rural, remote, and underprivileged areas. This disparity in access to essential services is a significant equity issue.AI-powered screening tools can bridge this gap by providing an initial assessment for parents, guardians, and teachers, enabling early identification of children who require further evaluation. Additionally, AI can streamline SLP workflows, allowing them to focus on high-impact, individualized care for students with complex needs.

• Human

• Hybrid

• AI

How to do you ensure privacy and accuracy of data?

To ensure privacy when implementing AI, begin with data minimization by only transmitting the data essential for completing the task. Anonymize any personally identifiable information. If feasible, introduce an API layer to act as a buffer between the end-user and the AI service, allowing for validation of data in both directions. Maintain logs of all submissions and AI responses, regularly reviewing them to enhance model training. Additionally, establish channels for users to provide detailed feedback.

• Human

• Hybrid

• AI

Ways to further change instruction. Based on what AI is now able to do, what are our general and special education teachers' expectations?

Opportunity to focus on growth and personalization of learning over delivery of content. I worked with too many educators who said "I taught it." We sit at a point in time where all our talk about personalizing learning is a tangible possibility. There is work to be done, but the light at the end of the tunnel is coming into focus. Understanding students starts with holistic view of their data in a way where it is not solid or separate.  Old'school AI (analytics) can recognize patterns and surface areas of need to help education be proactive and catch students who are at risk of becoming at risk. 

• Human

• Hybrid

• AI

What are the recommended metrics for testing reliability and validity of AI output?

When assessing the reliability and validity of AI outputs in supporting students with disabilities, it is essential to consider several key metrics and methodologies. As with any content or tool these metrics ensure that AI systems are effective, fair, and trustworthy. Reliability metrics include consistency, accuracy, and robustness. Validity metrics: Construct validity, content validity, and criterion validity. However, ethical and practical considerations include bias and fairness, accessibility, transparency and explainability, data privacy, and security. These things should be implemented through ongoing monitoring and stakeholder feedback. 

• Human

• Hybrid

• AI

How to help students with disabilities walk the tightrope of using AI as a tool without it compromising or overshadowing learning?

This resembles discussions about accommodations: when to phase out supports to encourage growth versus recognizing if a support or intervention has become a crutch. The question now is how we assess in the era of AI and what skills we need to develop to achieve mastery.

• Human

• Hybrid

• AI

As states begin to share this guidance with districts and schools, how can we ensure accessibility is included in those conversations?

As states begin to share AI guidance with districts and schools, ensuring accessibility is included in these conversations requires mandating the integration of accessibility standards in all educational policies and procurement processes. States should involve accessibility experts and advocacy groups to provide comprehensive input, ensuring AI products are compatible with assistive technologies (AT) before adoption. Training educators and administrators on accessibility best practices and establishing accountability measures can help guarantee that students with disabilities are not excluded from the benefits of AI due to overlooked needs.

• Human

• Hybrid

• AI

How do we support educators in understanding the ethical and legal implications of using student information in these tools? What kind of guidance might we provide for developers of these products around student data privacy, especially for students with disabilities? How do we support students in understanding how their data is being used in these tools? 

This question touches on three interrelated topics. For resources, everyone should consult the U.S. Department of Education's Office of Educational Technology (OET). The National Educational Technology Plan (NETP) offers good initial guidance for any technology. Additionally, the new OET AI Developer Guide provides a solid foundation for AI development. The Office of Special Education Programs (OSEP) also offers valuable guidance on assistive technology. Educators and developers should receive comprehensive professional development on relevant laws (e.g., FERPA, COPPA, ADA), ethical considerations, and best practices for data privacy. Open communication with parents and students about how technology gathers, uses, and protects data is essential. Furthermore, we must ensure students receive understandable and age-appropriate education about data privacy, provide them with transparency and control over their data, and ensure their voices are heard in the selection and use of technology.

• Human

• Hybrid

• AI

What skills might students and teachers need to be able to evaluate the AI they’re using? What recommendations might we make to developers about data model transparency? How can we ensure the AI products used in schools are transparent about the risks, flaws, and potential dangers of AI products related to their data models?

Given that AI tools are only as reliable as the data models they are built on, it is vital to equip students and teachers with the skills to critically evaluate AI outputs. Educational programs should include training on identifying biases and misinformation in AI models. Developers must be urged to maintain data model transparency, clearly disclosing potential risks, flaws, and biases. This transparency ensures that AI products used in schools are scrutinized for their ethical implications, promoting a safer and more informed use of AI in education.

• Human

• Hybrid

• AI

Will AI be considered as a stand-in for paras or other related service providers when hiring becomes difficult? How can we encourage cautious adoption of AI that could take the place of a human being in a school, especially for professionals who interface directly with students?

I hope not. These are the questions we should ask, and we should ensure we are innovating, skilling, and working in a way where the answer is always NO. Removing administrative burden does not remove the need for educators or paras, but opens more ways to connect and build a positive rapport with students. 

• Human

• Hybrid

• AI