- February 2, 2024
- Research
Can the eyes be portals to our inner thoughts? That seems to be the case, thanks to the recent development of an advanced AI-based system that interprets eye movements to predict a person’s next actions.
A team from the University of Maryland, consisting of a researcher and two of his colleagues, used eye tracking and advanced artificial intelligence with deep learning to predict the choices made by people comparing products on a website. The architecture of this AI, named RETINA (Raw Eye Tracking and Image Ncoder Architecture), has demonstrated an impressive ability to predict a person’s choices even before they consciously make a decision.
“Artificial intelligence excels at making predictions based on data,” explained Michel Wedel, Robert H. Smith Distinguished Professor at the Robert H. Smith School of Business and Chair of PepsiCo’s Department of Consumer Sciences. Moshe Unger of Tel Aviv University and Oleksandr Tuzhilin of New York University collaborated with him on RETINA. Their findings were documented in the scientific journal Data Mining and Knowledge Discovery.
Eye movement research has traditionally involved summarizing data, potentially missing certain details and types of eye movements. Wedel and his team’s sophisticated machine-learning technique allowed them to make full use of the detailed eye-tracking data. An interesting aspect of this algorithm is its ability to process the raw data from each eye separately, which Wedel says involves “huge amounts of data—hundreds of thousands of data points with millions of parameters.”
The potential of the algorithm is huge and has implications for various companies. A retailer like Walmart could use this technology to enhance the shopping experience in the metaverse, where many virtual reality devices equipped with eye-tracking technology can enhance the visualization of the virtual space. RETINA could give Walmart the ability to customize virtual displays based on consumers’ preferences, deciphered by previous eye movements. “Before making a choice, we can often predict a product’s preference based on where people are looking,” Wedel said. “Marketers can use this to reinforce incomplete choices or promote an alternative product.”
The benefits of RETINA extend beyond marketing, as eye tracking is gaining traction in many industries such as healthcare, mental health, user interface design, art, literature, finance, and more—virtually any sector that relies on visual decision-making. Big tech companies, including Meta and Google, have begun to combine eye-tracking firms as they look at multiple applications. Even with today’s ubiquitous front-facing cameras on devices like smartphones and computers, consumer eye tracking still lacks the accuracy of sophisticated research equipment, Wedel noted. In addition, privacy issues remain paramount; companies must obtain user consent.
The team is now looking to commercialize their algorithm and refine the research to improve the decision-making process. “We expect eye tracking to become extremely affordable,” Wedel noted. “Traditionally, eye movement data processing has been time-consuming. This algorithm greatly mitigates this, paving the way for potential applications that we cannot yet imagine.”
