A device, method, and system for inferring user intent to perform an action on a computing device includes monitoring the eye movement patterns of a user and determining the action to be performed based on the eye movement patterns. Signals relating to the eye movement of the user while observing a display screen of the computing device are processed to produce at least one eye movement feature. One or more classifiers are generated based on a training set of data of the eye movement feature over a time interval in which the user observes the display screen with an intent to activate the action. Thereafter, an eye-movement-pattern of the user may be analyzed, and features may be extracted, as the user is observing the display screen. The intended action is automatically activated if the user intent is inferred using the one or more of the classifiers.