The initial results were promising, with accuracy between 93.8 and 95 percent depending on the test. As Reddit users noted, though, the study only looked at 25 subjects — that’s not a huge sample to draw from. More work would be necessary to ensure the AI works reliably in flagging at-risk patients, and the team acknowledges there should be “further validation.” This also assumes patients would be comfortable knowing machines were studying their eye movements — those in the study were, but others might bristle at the thought.
If the accuracy translates well to larger studies, though, the AI could prove extremely helpful for the healthcare industry. Patients could recover at home while sharing their mental health, and psychotherapists would only have to focus on those patients with warning signs. In other words, the technology could improve the availability and quality of help for those that need it the most.