Resumen
Tasks that imply engagement of visual-spatial working memory (VSWM) are common in interaction with two-dimensional graphical user interfaces. In our paper, we consider two groups of factors as predictors of user performance in such tasks: (1) the metrics based on compression algorithms (RLE and Deflate) plus the Hick?s law, which are known to be characteristic of visual complexity, and (2) metrics based on Gestalt groping principle of proximity, operationalized as von Neumann and Moore range 1 neighborhoods from the cellular automata theory. We involved 88 subjects who performed about 5000 VSWM-engaging tasks and 78 participants who assessed the complexity of the tasks? configurations. We found that the Gestalt-based predictors had a notable advantage over the visual complexity-based ones, as the memorized chunks best corresponded to von Neumann neighborhood groping. The latter was further used in the formulation of index of difficulty and throughput for VSWM-engaging tasks, which we proposed by analogy with the infamous Fitts? law. In our experimental study, throughput amounted to 3.75 bit/s, and we believe that it can be utilized for comparing and assessing UI designs.