Performance on Activities of Daily Living and User Experience When Using Artificial Intelligence by Individuals With...

Area of Science:

• Assistive technology research
• Human-computer interaction
• Ophthalmology and vision science

Background:

• People with vision loss (PVL) can benefit from assistive technologies.
• Artificial intelligence (AI) offers potential advancements in tools for PVL.
• Objective data is needed to guide the selection of AI-powered assistive tools.

Purpose of the Study:

• To assess the objective performance, usability, and acceptance of AI by PVL.
• To provide evidence-based data for selecting AI technologies tailored to individual vision loss and needs.
• To compare the effectiveness of different AI implementations for PVL.

Main Methods:

• A cross-sectional, counterbalanced, cross-over study was conducted with 25 PVL.
• Performance was compared across two smart glasses (OrCam, Envision Glasses) and two AI apps (Seeing AI, Google Lookout), termed assistive artificial intelligence implementations (AAIIs).
• Task completion and timing were measured for text, text in columns, and searching/identifying tasks. Usability was assessed using the System Usability Scale (SUS).

Main Results:

• Assistive artificial intelligence implementations (AAIIs) significantly improved text task completion compared to baseline.
• Performance in "Searching and Identifying" tasks varied by AAII, with Seeing AI and Envision Glasses showing broader improvements.
• Participants reported high satisfaction with the evaluated AAIIs.

Conclusions:

• Despite task-specific performance variations, overall high satisfaction suggests acceptance of AI as an assistive technology for PVL.
• The study highlights the developing potential of AI in vision assistance.
• Findings offer guidelines for clinicians recommending AAIIs to PVL.