5.5 Usability Criteria for Eye-Gaze Media

On the basis of the preceding discussion, we will propose some usability criteria for eye-gaze media that can be used to asses the usability of an eye-gaze based application. When evaluating a specific system, one must, however, remember to take the specific circumstances into account. If, for example, an application is to be used in a noisy environment, it is unlikely that integration of eye tracking input with speech recognition input will be beneficial.

Some of these criteria are wholly or partly fulfilled by existing eye-gaze tracking systems, but some pose such profound paradoxes that probably no system will ever be able to match all criteria. We suggest that an ideal eye-gaze media should offer a high degree of

Involuntariness,

i.e., wherever possible, users' natural eye movements should be used, not requiring the user to perform more voluntary eye movements than absolutely necessary.

Tracking data utilization,

i.e., as many "dimensions" as possible of the eye tracking data should be utilized, both short and long term eye-gaze patterns, saccade direction, blink rate, pupil diameter etc.

Modality integration,

i.e., integration of tracking data with data from other modalities (e.g. movement, speech, gesture), to determine from what reaction the user would benefit the most, should be maximised.

Customisability,

i.e., as many idiosyncratic aspects and preferences as possible should be recordable (by the equipment, more or less with the user's knowledge), settable (by the user) and utilized (by the equipment).

The motivation for this criterion is the observed fact that people do have their individual ways of looking at their surroundings (cf. ), and they certainly have their individual interests and preferences for what they want their computers and televisions to do.

Technical transparency,

i.e., Hallett's (1986) technical requirements in the list should be satisfied as much as possible so as to make the actual eye-tracking as transparent as possible for the user. In particular, task-integrated recalibration should be maximised; though explicit user-conscious recalibration procedures might be short and simple, they do distract the user from his real task.