Researchers have revealed how artificial intelligence (AI) could help people who are blind or have low vision create their own personalised assistive technology, offering a glimpse into a future where accessibility tools can be tailored to individual needs rather than relying solely on commercially available solutions.
Published by researchers from institutions including Carnegie Mellon University, the University of Washington and Tufts University, the study explored how five blind and low vision participants used an AI-powered programming tool to design and test camera-based accessibility solutions. Over the course of the research, participants created more than 37 personalised tools to assist with everyday tasks.
Some of the tools addressed challenges not currently met by mainstream assistive technology. Examples included identifying the correct rideshare vehicle when several similar cars were nearby, recognising hand gestures during conversations, and creating customised solutions for specific environments and routines. Rather than replacing existing accessibility technology, the AI acted as a collaborative assistant, helping participants describe what they wanted and generating software that could then be refined through testing and feedback.
The researchers also identified important limitations. Participants still encountered challenges when describing their requirements, resolving conflicting instructions and working within the current capabilities of AI models. The study recommends future AI development should include better conversational guidance, community sharing of accessibility tools and support for personalised datasets to make custom assistive technology easier to create.
For Australians with disability, this research really shows how AI could one day make assistive technology more flexible, personal and responsive to individual needs, particularly where existing products do not fully address everyday accessibility challenges.