AI as a Catalyst for Universal Accessibility: Bridging the Gap in Inclusivity

A few weeks ago, during a visit to my brother in Winnipeg, he took me to the Human Rights Museum. As we walked through the corridors of this architectural masterpiece, which is as beautiful as it is educational and poignant, my project management hat wasn't just on—it was buzzing. Each design element, from the wide pathways to the ASL interpretations and multilingual captions accompanying every exhibit, spoke of a deep commitment to accessibility. It was impossible not to marvel at the meticulous orchestration behind such inclusivity and acknowledge the substantial resources—time, effort, and finances—that must have been invested.

The Current Landscape of Accessibility in Canada

In Canada, the legislative framework supporting accessibility includes the Accessible Canada Act (ACA) at the federal level, and various provincial laws like Ontario’s Accessibility for Ontarians with Disabilities Act (AODA). These laws mandate that organizations meet specific standards to ensure their services are accessible to all Canadians, regardless of their physical or mental abilities.

Despite the existence of comprehensive guidelines like the Web Content Accessibility Guidelines (WCAG), which offer multiple levels of compliance (Levels A, AA, and AAA), accessibility too often remains an afterthought for many organizations. This oversight is not always intentional. Factors such as constrained resources, inadequate planning, and a lack of representation and advocacy at the design stage can lead organizations, both large and small, to only meet the minimum required standards. While this approach meets legal requirements, it fails to fully embrace the spirit of inclusivity these guidelines are designed to promote. In my experience overseeing numerous digital communication projects, I've seen how these challenges manifest in real-world settings, leading to inconsistent implementations across different outputs. It underscores a broader issue: a systemic lack of prioritization of accessibility, which hampers our ability to cater to the diverse needs of all citizens.

How AI Can Transform Accessibility

Reflecting on my recent visit to the Human Rights Museum, I couldn’t help but wonder: What if we could achieve this level of accessibility universally? What if AI was the tool that could make such accessibility accessible to more organizations, each grappling with their own set of challenges, from small nonprofits to large corporations?

AI-Enabled Design and Testing:

AI might be able to act as a low-cost accessibility consultant during the design phase, like a new kind of team member who specializes in these considerations. This AI could:

• Analyze designs for potential accessibility issues, offering suggestions that could help avoid costly post-launch modifications.
• Recommend accessible features such as color palettes or text layouts to enhance usability.
• Simulate various user experiences to identify potential barriers that might not be obvious initially.

This proactive approach might allow organizations to consider accessibility from the start, potentially reducing the need for extensive revisions and helping them aim for comprehensive standards like WCAG AAA from the outset.

Automation of Accessibility Features:

As AI technology evolves, it's starting to more accurately handle tasks that typically required a lot of time and effort:

• Generating Subtitles and Audio Descriptions: AI might help make videos more accessible by providing subtitles and audio descriptions.
• Creating Descriptive Text for Images: Automating the creation of descriptive texts, such as alt text for images, helps make content more accessible to those who are visually impaired, ensuring they can understand visual elements through screen readers.
• Translations: AI can also translate content into multiple languages, broadening the reach and making the information accessible to a more diverse audience.

Integrating AI to manage these tasks could streamline some of our operations and help reduce some of the overhead costs. This efficiency would allow teams to redirect their energy or budgets—perhaps previously spent on outsourcing these tasks—toward more strategic and creative projects. Freeing up staff from repetitive tasks would also allow them to focus on work that utilizes their skills and creativity more effectively, which we know is one of the recipes to increase job satisfaction, employee engagement, and the cultivation of a workplace where innovation and strategic thinking are at the forefront.

Consistency and Scalability:

With AI, the application of accessibility features across all digital content might become more uniform, offering:

• Uniform Standards: Content could meet set accessibility standards with less need for manual oversight, reducing variability that often arises from different team members' styles and interpretations.
• Scalable Solutions: As an organization expands, AI’s scalability could help ensure new content and platforms automatically include these accessibility protocols. This consistency is crucial when managing diverse and growing amounts of content across various media.

One particular area where AI could make a significant difference is in the "last mile" of content creation—developing descriptions, alt text, metadata, and translations. Traditionally, these tasks have varied significantly from channel to channel and even from one marketer to another, often leading to inconsistencies that dilute the effectiveness of accessibility efforts. Now, with AI capable of generating all these elements quickly once specifications are provided, there's a compelling argument for raising our standards.

Practical Applications of AI in Accessibility Across Industries

Companies with specialized expertise in Computer Vision, Large Language Models, and Machine Learning are leading the charge in using AI to enhance accessibility for people with disabilities. Here's a look at how firms around the world are applying their skills to break down barriers:

• Microsoft: Utilizes Computer Vision with its Seeing AI app, helping visually impaired users by narrating the visual world around them.
• Google’s Project Euphonia: Leverages Machine Learning to enhance voice recognition systems for people with speech impairments, adapting AI to understand varied speech patterns.
• Voiceitt: Employs Large Language Models to translate atypical speech patterns into clear speech in real-time, facilitating everyday communication for those with speech disorders.
• IBM: Developed AI Fairness 360 using Machine Learning to detect and mitigate biases within AI models, promoting equity and inclusiveness.

While these examples highlight advancements from well-resourced companies, they also set a precedent that can inspire smaller organizations and startups. Last year in 2023, the Canadian Minister of Innovation, Science and Industry announced that 11 organizations in communities across the country will receive funding as part of the $5.8 million investment in the second phase of the Accessible Technology Program, the majority of which are using AI on their projects. By demonstrating what is possible with AI in accessibility, these innovations encourage a wider array of businesses to explore how they might leverage existing AI technologies or partner with tech leaders. This trickle-down effect can broaden the impact of AI on accessibility, making it more universal not just through direct implementation but also through inspiring scalable and cost-effective solutions adapted to different contexts and resources.

Navigating the Limitations of AI in Accessibility

While AI offers huge potential to enhance our accessibility efforts, it's important to approach its integration with caution and awareness. AI systems are not foolproof; they make mistakes and can inherit biases present in the data they learn from which may inadvertently exclude or misrepresent certain groups (side note, if you want to learn more about algorithmic biases and how present they are in the tech we take for granted, I cannot recommend enough Author and PhD Dr. Joy Buolamwini 's book Unmasking AI). This is why continuous human oversight and advocacy is essential to monitor and correct these errors and biases, and to make sure that our accessibility solutions are genuinely accurate, inclusive, and fair.

As we embrace AI, we must also recognize that not every team member may view these changes with the same enthusiasm as me. The introduction of AI into our workflows can be met with apprehension, particularly around its impact on job roles and responsibilities. It's important to navigate these transitions with empathy and openness, providing support and training to help everyone adapt and feel included.

Parting Thoughts

Inspired by the seamless integration of accessibility at the Human Rights Museum, it's clear that as more of us begin to use AI, we have the opportunity to embrace this specific use case to transform our environments. With this new tool making universal accessibility more accessible to all of us, we have no excuse not to elevate our practices and expectations.