How AI is transforming accessibility: expert opinions from TechShare Pro

In addition to having chats with so many inspirational people working in digital accessibility, including Sarah Herrlinger from Apple (pictured, right, with me and my guide dog, Hugo), one of my top highlight themes from Day 2 of AbilityNet's TechShare Pro 2024 conference was around the intersection of artificial intelligence (AI) and accessibility.

Artificial intelligence (AI) is transforming many aspects of technology, particularly when it comes to accessibility. For both disabled users and developers creating accessible software, AI is proving to be a game-changer.

From real-time transcription services to tools for automated code compliance, the potential for AI in accessibility is vast and growing.

Here’s how AI is making strides in this domain, supported by practical examples.

Empowering people with disabilities through AI

AI-powered tools are becoming essential for disabled individuals by helping bridge gaps in everyday activities and communication. Here are a few examples:

• Be My AI: Building on the popular Be My Eyes app, Be My AI leverages generative AI to interpret images and provide detailed contextual information. For example, a visually impaired person can photograph ingredients and ask the AI what dishes they could prepare.
• Seeing AI by Microsoft: This app uses a smartphone camera to read text, identify objects, and recognize handwriting, aiding blind users in navigating their surroundings and tasks independently.
• Ava and HeardThat: Apps like these transcribe conversations in real time, using advanced speech recognition models. HeardThat, in particular, isolates voices from background noise, a boon for those with hearing impairments.
• AI-Powered Hearing Aids: Devices such as the Orka Two dynamically adjust settings to optimize sound clarity, prioritizing voices over ambient noise.
• Goblin Tools: An AI-powered to-do list that breaks complex tasks into manageable steps, helping users with executive dysfunction stay organised.
• Google's Project Relate: Designed for individuals with non-standard speech patterns, this tool converts their speech into comprehensible text or spoken responses.

There’s no doubt that AI is unrivalled in assisting people with disabilities in their daily lives, but it’s also important that the apps and websites they need to use are accessible and inclusive.

Assisting developers in accessibility

AI is not only aiding end-users but also empowering developers to create more inclusive software and ensure compliance with accessibility standards. Key applications include:

1. Automating accessibility testing

Tools like Google’s Lighthouse and axe by Deque Systems: Automated testing suites use AI to scan for WCAG (Web Content Accessibility Guidelines) compliance, identifying common errors like poor colour contrast or missing alt text. Other common elements such as JavaScript, however, currently remain outside the abilities for AI to assess for accessibility.

2. Code generation and refinement

GitHub Copilot: This AI tool assists developers by generating code snippets, including accessible components. While it streamlines coding, developers must verify its outputs for accessibility compliance since AI can still make errors.

3. Creating personas for testing

AI has begun to be used to simulate user personas, predicting potential barriers for individuals with specific disabilities. For instance, it can model the experience of someone with low vision or limited mobility and highlight potential areas of concern.

These tools help reduce the time and effort required to identify accessibility flaws, enabling a more inclusive development lifecycle.

Challenges and ethical considerations with AI

Despite its potential, AI in accessibility comes with challenges:

• Accuracy and Trust: AI models like ChatGPT sometimes confidently produce incorrect outputs, underscoring the need for human oversight – especially where the user in question isn’t able to verify the data themselves due to their disability.
• Data Privacy: Using AI often involves processing user data, raising concerns about security and consent.
• Ethical Testing: Simulating disabilities raises questions about whether AI can truly replicate lived experiences without oversimplifying or misrepresenting them.

The road ahead

Discussions from TechShare Pro 2024 included that AI's capabilities are evolving rapidly, promising even more innovative solutions for accessibility. Tools are becoming better at recognising nuanced user needs, from generating more contextually aware responses to creating fully accessible digital experiences. 

For developers, AI offers a path to more efficiently integrate accessibility into every stage of the design and testing process, ultimately resulting in technology that works for everyone.

By continuing to address ethical considerations and leveraging AI responsibly, we can ensure that it serves as a powerful ally in building a more inclusive digital world.

We look forward to you joining us at next year's TechShare Pro conference! If you'd like to catch up with all the sessions from this year's event, get an archive ticket to access all the recordings. 

Accessibility training courses