With a Flick of the Eye: Assessing Gaze-Controlled Human-Computer Interaction

Gaze-controlled user interfaces appear to be a viable alternative to manual mouse control in human-computer interaction. Eye movements, however, often occur involuntarily and fixations do not necessarily indicate an intention to interact with a particular element of a visual display. To address this so-called Midas-touch problem, we investigated two methods of object/action selection using volitional eye movements, fixating versus blinking, and evaluated error rates, response times, response accuracy and user satisfaction in a text-typing task. Results show significantly less errors for the blinking method while task completion times do only vary between methods when practice is allowed. In that case, the fixation method is quicker than the blinking method. Also, participants rate the fixation method higher for its ease of use and regard it as less tiring. In general, blinking appears more suited for sparse and non-continuous input (e.g., when operating ticket vending machines), whereas fixating seems preferable for tasks requiring more rapid and continuous selections (e.g., when using virtual keyboards). We could demonstrate that the quality of the selection method does not rely on efficiency measures (e.g., error rate or task completion time) alone: user satisfaction measures must certainly be taken into account as well to ensure user-friendly interfaces and, furthermore, gaze-controlled interaction methods must be adapted to specific applications.