Reading Your Mind Through Your Eyes: Using Eye Scan Patterns and Machine Learning to Predict Number Choice
John, Sharon G.
MetadataShow full item record
Eye tracking technology measures eye movements in real time. Studies have shown that eye movement patterns can express spatial cognitive thoughts, confirming the existence of the eye-cognition link. The purpose of this experiment is to determine a correlation between spatial cognition and number processing. If a correlation exists, features suggesting this correlation will be identified to predict a subject’s response using machine learning. Fifty subjects were asked to look at a screen and respond to the following auditory prompt: “Think of a number x/y/z and say it out loud”. These intervals were named Number Line, Pre-prompt, Mid-prompt, and Post-prompt. A predicted model was created by training a Random Forest (RF) Algorithm. Results indicate a strong correlation between saccade vector and participants’ response, suggesting subjects made saccades toward the direction of response in mind. A strong correlation was also identified between Fixation X and participants’ response as these fixations were made in relative locations of the given numbers on a number line. Test accuracy of 90.5% for Trial 1 was obtained with 50 trees. All ten trials resulted in a mean accuracy of 89%. Our future plans involve determining the existence of a vertical component of spatial cognition.