Web Design Prototypicality

The project tries to answer why some webpages look nice and are intuitive to use, while others look unappealing and are hard to understand. One reason for looking nice and being intuitive may be how typical a webpage design is: typical webpages follow a persons expectations of how a webpage should look like and behave. These expectations form based on the persons prior encounters with similar-type webpages, and, if they are not met, the person will struggle to like and interact with a webpage. For example, an average person would expect an online-newspaper homepage to list recent news stories, and may well feel lost if the homepage lists old, archived news, or distrust the entire news agency if the homepage tries to sell merchandize in addition to showing news. However, while it may be clear to the human eye whether a particular webpage is typical, conventional and expectation-following, it is difficult to define this precisely and to train a machine to know it. This project aims at solving these issues. Project researchers will collect a large number of webpages from the Internet, sort these webpages by genre (for example, as online newspaper VS online shop homepages), and involve a large number of people in viewing, using and evaluating these webpages. The researchers will then use these evaluations to study what exactly makes a webpage typical, how to teach an algorithm to detect atypical webpages, and how being typical affects webpage use and appreciation by people. Future web designers will rely on the project results to design better webpages webpages that are easy to interact with and look to fit a particular situation by better understanding what a typical person expects to find on a typical webpage of a particular-genre website. The designers will also understand better how to avoiding clichés and stereotypes in their design, and how to make their webpage designs stand-out, which is important when all alternative websites look alike, and the newly designed ones need to capture attention. Future software engineers and entrepreneurs may also rely on the project results to develop artificial intelligence systems for automatic web design. Such systems would either design websites fully automatically or actively help designers create better designs faster. Future entrepreneurs could commercialize such automatic-design systems, which would create jobs in the future economy in Austria and abroad.

The "Web Design Prototypicality" project has established that for each group of webpages - be it homepages of commercial banks, e-commerce websites, or university sites - there is a prototype that represents the most typically and ordinarily looking example for that group. Such a prototype is the page that a person imagines when thinking about a particular group of webpages. Looking similar to the prototype also makes a webpage appear more aesthetically pleasing and trustworthy, though this positive effect may diminish at very high levels of prototypicality, possibly because such webpages look cliché or dull. People do not necessarily perceive prototypicality consciously, but they are affected by it, seeing the organizations that are represented by prototypical websites as offering higher-quality services and being better employers than organizations with less prototypical websites. The project has developed 'prototypicality' as a concept applicable to webpage design, which was a novelty in that the vast majority of research on prototypicality has been conducted in cognitive psychology and its findings could not be applied directly to webpages. We have sampled the screenshots of over three thousand webpages and evaluated them in online user studies with over two thousand participants. In addition, users' perceptions of websites were investigated using eye-tracking. The resulting findings showed that prototypicality strongly correlates with people's perception of visual aesthetics and trustworthiness of webpages and significantly affects users' attitudes towards the represented organizations. These research outcomes are indicative for practice in the sense that, managers can decide, on the one hand, when designing corporate websites, to make them more prototypical (e.g., to increase trust, which could be particularly useful for small, relatively unknown organizations) or less prototypical (e.g., to make a website stand out, which may be an interesting option for well-known, already-trusted brands that want to differentiate themselves from the competition). The project has also developed a novel method for automatically estimating the prototypicality of webpages based on webpage features. We identified specific features (e.g., the amount of page text, the number of large images, or the average size of areas with non-white backgrounds) that can be used to describe a prototype for a category of webpages and estimate the distance of individual webpages to the prototype of the category. A larger distance corresponded, as hypothesized, to lower prototypicality, which was measured externally with the help of crowd-workers. Thus, the established new methodological approach allows to estimate prototypicality with algorithms in the future. Further research and development, including artificial intelligence research for webpage evaluation and redesign, could draw on the developed algorithms, features, and a large collected dataset of webpages published open-access as project results to develop automated tools to aid web design and inform management decisions.