Perception and Manipulation of Game Control

Abstract

Abstract. Brain-computer interfaces do not provide perfect recognition of user input, for similar reasons as natural input modalities. How well can users assess the amount of control they have, and how much control do they need? We describe an experiment where we manipulated the control users had in a keyboard-controlled browser game. The data of 211 runs from 87 individuals indicates a significant linear correlation between users' sense of control and the amount of control they really had in terms of mutual information (not accuracy!). If users know what they put in, they can assess quite well how much control they have over the system. In our case, from an amount of control of above 0.68 bits in mutual information (a 5-class accuracy of 65%), this aspect of control no longer seems to be the critical factor for finishing the game. Deliberate manipulation of perception may offer a way to make imperfect, uncertain input modalities more acceptable, especially in combination with games.

Automation Accuracy Is Good, but High Controllability May Be Better

“…Both Kay et al and Katsuragawa et al noticed that different recognizers are perceived differently by users [8,9]. However, acceptable accuracy was only investigated in the context of automation with which user is passive [9], like notifications, or input recognition [8,16]. Automated tasks were not considered, and usability only in the context of input recognition.…”