École Nationale Supérieure de
Techniques Avancées

Abstract

Keyboard shortcuts are an efficient method to execute commands, but they remain difficult to discover and to memorize. Multiple shortcut adoption methods have been proposed in older studies. These studies have relied on iid command sequences following some distributions to evaluate user adoption of shortcuts. In this article, we have 2 main goals. We first propose a new method to model command sequences for these experiences and compare it to older methods. Next, the goal is to create TeacherCut, a model-based, adaptive shortcut teacher that recommends shortcuts based on user characteristics (namely the estimated memory state, as well as adversity to learning of the user), the task characteristics (notably the history of modalities and executed commands) as well as the teacher characteristics (via the estimated interruptibility of the recommendation). then, we perform multiple simulated evaluations, varying the user type (fast/slow learner, adverse/not adverse), the distribution of sequences commands, and the assistance (recommendation policy). Model comparisons suggest that adversity of learning is an important factor for the success of the shortcut teacher and the lack of theoretical foundations to estimate the parameters. This project emphasizes the need to conduct further empirical studies to explore in priority the effects of user adversity and interruptibility in order to elaborate an adaptive shortcut teacher.