A blink can reveal a lot about what we’re thinking. Now, researchers have developed an AI algorithm called RETINA that can analyze raw eye tracking data to predict consumer choices with uncanny accuracy.
Instead of relying on aggregated eye movement metrics, RETINA leverages deep learning to process granular, millisecond-by-millisecond data points from both eyes. This retains more nuanced insights into decision making that are typically lost by traditional analysis methods.
In tests comparing four laptops, RETINA achieved over 70% predictive accuracy after viewing just 5 seconds of eye tracking data per person. The AI was often able to foretell someone’s eventual choice well before they consciously committed to it.
“Before people have made a choice, we can say it’s very likely they’ll choose a certain product based on their eye movements,” explained lead researcher Michel Wedel. “Marketers could use this to reinforce that choice or nudge them to a different one.”
The uncanny predictive power of RETINA stems from its ability to recognize subtle patterns in the frenzy of data points from rapid eye movements. As Wedel put it, “It’s a lot of data – several hundred thousands points, with millions of parameters – and we use it for both eyes separately.”
The applications extend far beyond marketing. RETINA’s streamlined yet incisive analytical approach could prove useful in medicine, psychology, design, finance, and more as eye tracking becomes increasingly widespread.
Some reasonable privacy concerns exist, especially as front-facing cameras on phones and AR/VR headsets enable eye tracking on personal devices. Still, Wedel believes that as the practice becomes more commonplace, RETINA will unlock invaluable consumer insights across industries.
