Cognitive Computing

The Impact of AI on Accessibility

Share this post:

At the recent Boston Accessibility Conference, Dr. Ruoyi Zhou, Director of IBM Accessibility Research, hosted an exciting session on the possibilities of AI, machine learning and virtual reality to benefit people with disabilities.

The panelists, who included leaders and innovators from MIT, Open Access Technologies, Rendever, Aira and the Massachusetts Association of the Blind, all spoke enthusiastically about the potential of AI technologies to enhance and extend the independence, quality of life, and experiences of people with physical or cognitive impairments.

Future of Accessibility

Facial and object recognition, AI algorithms to trigger sensory responses, virtual reality and the ability of machine learning to infer context and guide people through an online or physical experience are just a few areas where they see AI making an impact in accessibility today.

For example, consider screen readers. By leveraging AI and machine learning capabilities, screen readers could infer, determine and summarize what’s being presented on the screen, and guide the user through the individual components, like a push button or check box – potentially correcting coding lapses, creating bridges where no API or standards yet exist, and eliminating the “accessibility lag” found with many solutions.

However, all of the panelists cautioned against viewing AI as a panacea for accessibility. They cited several issues and challenges ahead, including:

  • Reliability & Availability.  If an object is misidentified or the system goes offline, someone relying on the technology while crossing a street or using an assistive device could be in danger.
  • Data. AI needs lots of data in order to do a good job. If organizations accumulating large stores of data share it, we will achieve better systems, faster.
  • Integration. While AI systems can operate where APIs don’t exist, they will work more effectively if they operate and integrate with existing systems and standards.
  • Governance & Accountability. At present, AI and machine learning systems are making relatively simple decisions but these decisions will rapidly become more complex. Who governs these decisions, and who will be accountable for their impact?

Overall, the panel saw tremendous opportunities for AI technologies to deliver new cost-effective, adaptable and personalized solutions and services that will benefit people of all ages and abilities.

At this point, however, AI is no substitute for good accessibility guidelines and best practices. In the words of one panelist, “It’s not magic…..don’t wait for the AI guys to solve the problem!”

To learn more, please view the entire session.

More Cognitive Computing stories
By Tim Powers on March 14, 2018

IBM Accessibility Research at CSUN 2018

Artificial intelligence (AI), automation, compliance, and best practices in enterprise accessibility make up the IBM Accessibility Research lineup for the 33rd annual CSUN Assistive Technology Conference in San Diego next week (March 19-23, 2018). CSUN is the world’s largest event dedicated to presenting and exploring new ways technology can assist people with disabilities. At the […]

Continue reading

By Erich Manser on March 7, 2018

IBM Employees with Disabilities Volunteer Time to Create Accessible Technology

Who better to confirm the accessibility of solutions before they are released to market than people with disabilities? Being visually impaired, I rely considerably on magnification and altered colors to see what’s on my device screens. One day, not long ago, I sat in a room full of developers in a meeting and it was illuminating to […]

Continue reading

By Michael Gower on February 8, 2018

Simplifying the New WCAG 2.1 Guidelines

Recommendations for new accessibility guidelines were released on January 30, 2018 by the World Wide Web Consortium (W3C). In this new draft version of the Web Content Accessibility Guidelines (WCAG), none of the current 2.0 Success Criteria have been altered. Instead, 17 additional criteria have been recommended for 2.1. This article is intended to help […]

Continue reading