Metrics-Based Evaluation of Graphical User Interface Aesthetics

Aesthetics has been shown to play a considerable role in system acceptability and perceived usability, hence the importance of studying the aesthetic aspect of user interface design. Computable aesthetics evaluation measures have been proposed that are based on the extraction (or segmentation) of the visual elements that compose the interface layout. In this paper, we observe that many past studies have bypassed the segmentation problem by either using artificial designs, where individual elements are all known a priori, or by doing the segmentation manually. We then report on an experimental study, which demonstrates the impact of different segmentation methods (i.e. manual versus automatic segmentation) on the calculation of the aesthetics evaluation metrics. The segmentation problem remains a bottleneck issue, which impedes the full automation of computational aesthetics evaluation.