Computation of Interface Aesthetics Aliaksei Miniukovich University of Trento Via Sommarive 9, Povo TN, Italy miniukovich@disi.unitn.it Antonella De Angeli University of Trento Via Sommarive 9, Povo TN, Italy deangeli@disi.unitn.it ABSTRACT People prefer attractive interfaces. Designers strive to outmatch competitors, and create apps and websites that stand out. However, significant expenses on design are unaffordable to small companies; instead, they could adopt automatic tools of interface aesthetics evaluation, a cheaper strategy to good design. This paper describes an important step towards such a tool; it presents eight automatic metrics of graphical user interface (GUI) aesthetics. We tested the metrics in two exploratory studies – on desktop webpages (N = 62) and on iPhone apps (N = 53) – and found them to function on both GUI types and for both immediate (150ms exposure) and deliberate (4s exposure) aesthetics impressions. Our best-fit regression models explained up to 49% of variance in webpage aesthetics and up to 32% (if app genre is considered) of variance in iPhone app aesthetics. These results confirm past results and suggest the metrics are valid and reliable enough to be widely discussed, and possibly, to be embedded in our prospective GUI evaluation tool, tLight. Author Keywords Automatic metrics; GUI evaluation; user study; immediate impression; deliberate impression; tLight. ACM Classification Keywords H.5.2. Information interfaces and presentation (e.g., HCI): User interfaces – graphical user interfaces (GUI), evaluation/methodology. INTRODUCTION There is no doubt visual aesthetics matters in interface design. Surrounded with multiple offers of same-quality services and products, Web and app users have become selective and disregard apps and websites they do not like immediately [12]. A possible way to survive in such an environment includes carefully working out all details of visual design, making the design stand out [33]. However, small companies, start-ups and individual developers often cannot afford hiring a design agency and do their design themselves. In such cases, even well-detailed design guidelines are of limited help, since, to be applied properly, they require extensive training. Concrete GUI evaluation tools could exemplify abstract design guidelines, and drive and substantiate design choices. The tools would be based on specific quality metrics that represent specific GUI design aspects. In this paper, we extend earlier work [17, 16], and describe and test in two studies eight GUI aesthetics metrics: visual clutter, color range, number of dominant colors, figure- ground contrast, contour congestion, symmetry, and the new metrics of grid quality and white space. We based the metrics on the psychological investigations of what people see as complex and unappealing [23], and HCI investigations of webpage aesthetics [33, 14, 36, 25, 26]. The results of the present two studies replicated the results of past studies [17, 16], which suggests the metrics are solid enough to be presented to the larger CHI audience. In addition to this, we have replicated the phenomenon of consistent and lasting immediate impressions [14, 33] on two types of stimuli (webpages and mobile apps) using a between-subjects experimental design. In the rest of paper, we review related work on aesthetics in HCI and automatic aesthetics measures, and describe the eight GUI aesthetics metrics. We report Study 1, which tested metric performance on webpages, and Study 2, which tested metric performance on iPhone apps. Lastly, we summarize the results and discuss their implications for the automatic evaluation of interface aesthetics. RELATED WORK Past attempts [40, 39, 26, 25, 16, 17] to automatically account for visual aesthetics of GUIs consisted of two steps: gathering user scores of GUI aesthetics and matching them against computed scores of a set of automatic metrics. The first step reflects our understanding that beauty lies in the eye of the beholder, and involves conducting either carefully orchestrated in-lab user studies [14, 33, 21, 17] or large-scale crowdsourcing studies with thousands of participants [25, 25]. The second step uses the averaged user scores as the ground truth data, and tests how well various metrics and algorithms predict the scores. Collecting Aesthetics Scores The influence of aesthetics on the overall appreciation of GUI changes with time. Sonderegger et al. [29] conducted a longitudinal study and demonstrated the positive effect of aesthetics to almost disappear after the initial use phase. However, it is largely the initial phase that determines if a Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights for components of this work owned by others than ACM must be honored. Abstracting with credit is permitted. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. Request permissions from Permissions@acm.org. CHI 2015, April 18 - 23 2015, Seoul, Republic of Korea Copyright is held by the owner/author(s). Publication rights licensed to ACM. ACM 978-1-4503-3145-6/15/04…$15.00 http://dx.doi.org/10.1145/2702123.2702575