Characterizing non-visual target acquisition tasks with the aid of a tactile display: investigating factors beyond the classical Fitts' theorem

TLDR: The results show that Fitts’ law accurately predicts performance for the visually impaired, even though the average movement time exceeded the 3-s window typical of rapid-aimed movement, and a model to predict overall MT in non-visual environments with the aid of haptic reference points is proposed.

Abstract: Interaction in most non-visual environments is limited to discrete units of input and feedback, such as keyboard and sound. The purpose of this study is to use the classical Fitts' law to characterize target acquisition in a non-visual environment with continuous input devices and haptic feedback. Two groups of 21 severely visually impaired volunteers were recruited to perform a series of reciprocal tapping tasks based on three experimental conditions: Fitts' index of difficulty (ID), types of tactile feedback, and the presence of haptic reference points. Types of tactile feedback included continuous vibration, enclosure, and surface friction. The results show that Fitts' law accurately predicts performance for the visually impaired, even though the average movement time (MT) exceeded the 3-s window typical of rapid-aimed movement, particularly when the Fitts' ID is greater than 3 bits. Interestingly, actual distance to the target has a larger effect on MT than target width. Enclosure haptic feedback (i.e., a sudden bump when the mouse pointer crosses over the object boundary) allows participants to rapidly acquire even small targets with a width of only 4 pixels, while varying constant degrees of surface friction are ineffective at conveying distance. Adding a grid of equidistant haptic reference points significantly reduced the number of missed targets and the standard deviation of end clicks around the target center, but increased the overall MT. This paper proposes a model to predict overall MT in non-visual environments with the aid of haptic reference points. These results have implications toward the design of interfaces and input devices that provide continuous control in non-visual environments.