Perspectives on artificial intelligence (AI) often swing between the extremes of unbridled techno-optimism and existential fears for the survival of humanity. In the rush to integrate AI into products and services, companies sometimes provide value but other times stumble. In the realm of training and education, instructors grapple with the challenge of harnessing AI’s potential to enhance learning experiences.

To cut through the noise and find practical uses for AI in learning design, turn to those who have always been experts in adaptation: users with disabilities. Significant AI innovations are happening in the accessibility space, with profound implications for disabled learners and all users. By making educational content more inclusive and effective, AI holds the potential to be a powerful tool for creating accessible and equitable learning experiences.

Digital learning is particularly relevant for learners with disabilities. Given the flexibility inherent in online learning, these learners often prefer it over in-person options, as it allows them to access content using their own technology in an environment that suits their needs. As online learning continues to expand, there is an increasing legal and ethical obligation for content creators and platforms to ensure their offerings are accessible to all users.

Learners with disabilities who are faced with inaccessible learning content are often at an impasse. Some types of disabilities can bridge the gap with assistive devices, such as screen readers, voice activation devices, and switch devices. However, these devices rely on code within the application or site that is specifically written for accessibility.

Fixing inaccessible digital content comes with its own challenges. Remediation ideally would be done at-scale with the help of automated checkers. However, automated checkers are only able to flag 57% of issues under the best circumstances. The results of automated checking must be manually confirmed. Correcting accessibility issues can be difficult, time consuming and expensive.

How Can AI Improve Accessibility in Learning Environments?

• AI and learning platforms: Training content is only as accessible as the platform on which it is hosted. Learning management systems (LMSs) typically include content development tools for designers to build learning content. Much of the functionality found in an online course — the buttons, the interactive components, the media player — are native to the LMS. AI can assist in the development of accessible code for course hosting platforms. GitHub Copilot, for example (not to be confused with Microsoft Copilot), aims to assist developers with writing accessible code. AI can check code for accessibility. Axe DevTools (Axe DevTools), by Deque, is a set of accessibility testing tools that utilize AI; DevTools’  AI uses computer vision models for object detection, called Intelligent Guided Tests, to identify an element and determine if it is accessibly formatted. Content creators populating an LMS often do not have the ability to edit the accessibility of the digital learning platform, so it is important that these authoring tools are built with accessibility in mind, and there are clearly avenues where AI can help.

• AI and learning designers: AI can facilitate aspects of accessible content creation, such as creating alternative text for images and generating transcripts for audio content. Generative AI is uniquely adept at summarizing and categorizing content, tasks that are central to creating learning content and activities. Instructional designers (IDs) are using easily accessible tools like ChatGPT and Bard to create knowledge check questions. ID-Assist is one example of a tool that uses large language models (LLMs) to perform course design tasks like storyboarding, learning objectives and quiz creation.
• AI and learning paths: With the right instructions, AI can assist in creating personalized learning paths by analyzing individual learner behaviors, preferences and performance data. In the context of training for adult learners, this can be especially beneficial. AI-driven systems can tailor learning and development (L&D) initiatives to meet the unique demands of diverse adult learners by dynamically adapting content, pacing, and delivery methods based on real-time performance data. Learning pathways can also be adapted to accommodate a disclosed disability.
• AI and learners: People with learning, cognitive, or psychological disabilities can find the interpretive abilities of LLMs to be a huge time saver. Natural language processing, a capacity of AI-chatbots, can enable these students to ask for the extra help they may need to understand complicated topics. For those who are neurodivergent, AI can assist in interpreting the tone of a message or image.
• AI and assistive devices: If learning content, and the platform that hosts learning content, is not written with screen readers in mind, screen readers may not be able to access the course content. AI computer vision models can assist by interpreting the visual interface, such as through object character recognition (OCR), to provide an additional layer of support. The popular screen reader NVDA, for instance, offers an add-on called AI Content Describer (AI Content Describer) that can recognize and describe unclear screen controls and images. Be My Eyes virtual AI (Be My AI) is another huge advancement for people who are blind or vision impaired. Be My Eyes is a service that allows the user to take a photo or video and connect with a live assistant. With the addition of AI, Be My Eyes becomes much more effective, and allows users to be more independent without the reliance on a live human agent. AI can further enhance existing assistive technologies such as speech-to-text, predictive text, and switch access scanning.

Future Opportunities and Challenges

The power of AI in accessible digital education is its capacity for context-aware, customizable, and adaptive responses. The real impact of AI technology is still to be determined, but for users with disabilities, these tools are already having an effect.

The benefits of AI for accessible learning design must be balanced by an awareness of the issues and concerns surrounding this technology. AI-enabled technology does not solve the need for accessible code, accessible design, and accessible learning content. In fact, AI can create new challenges for accessibility.

LLMs are trained on data sets that include biases and harmful stereotypes. In tasks that involve drawing conclusions, such as analyzing performance data at-scale for a large organization, an AI model may unfairly stigmatize outliers. Another challenge of AI technology is that training LLMs demands a highly resource-intensive data infrastructure.

However, activists, technologists, and policy makers are working to mitigate these issues. Processor technology is becoming more energy-efficient, and the U.S. federal government has released a Blueprint for an AI Bill of Rights, including a right to safe and effective systems, data privacy, and protection from discrimination, among other things.

Amid these considerations, AI-enabled technology has the potential to reduce barriers for learners with disabilities. In fact, these tools are already making a difference for platform developers, instructional designers, and learners.