AB4Web : An On-Line A/B Tester for Comparing User Interface Design Alternatives
(2019) Proceedings of the ACM on Human-Computer Interaction — Vol. 3, n° EICS, p. 1-28 (2019)
Files
AB4Web-EICS2019.pdf
Open Access - Adobe PDF
- 2.67 MB
Details
- Authors
- Abstract
- We introduce AB4Web, a web-based engine that implements a balanced randomized version of the multivariate A/B testing, specifically designed for practitioners to readily compare end-users' preferences for user interface alternatives, such as menu layouts, widgets, controls, forms, or visual input commands. AB4Web automatically generates a balanced set of randomized pairs from a pool of user interface design alternatives, presents them to participants, collects their preferences, and reports results from the perspective of four quantitative measures: the number of presentations, the preference percentage, the latent score of preference, and the matrix of preferences. In this paper, we exemplify the AB4Web tester with a user study for which N=108 participants expressed their preferences regarding the visual design of 49 distinct graphical adaptive menus, with a total number of 5,400 preference votes. We compare the results obtained from our quantitative measures with four alternative methods: Condorcet, de Borda count starting at one and zero, and the Dowdall scoring system. We plan to release AB4Web as a public tool for practitioners to create their own A/B testing experiments.
- Affiliations
Citations
Vanderdonckt, J., Zen, M., & Vatavu, R.-D. (2019). AB4Web : An On-Line A/B Tester for Comparing User Interface Design Alternatives. Proceedings of the ACM on Human-Computer Interaction, 3(EICS), 1-28. https://doi.org/10.1145/3331160 (Original work published 2019)