A tool-supported approach to adapt web user interfaces considering the emotional state of the user

Abstract

The field of Human-Computer Interaction aims at developing computational systems that place the user at the center of the their development, so he/she can have the best experience when interact with the final solution. When it was acknowledged that emotions affect the relation of the individual with what surrounds him/her, scholars began to measure these emotions based on instruments that had been developed over the years. It should be emphasized that emotions can be seen as a composition of five components – physiological reactions, subjective feeling, motor expression, cognitive appraisals and behavioural tendencies. It is believed, therefore, that evaluating more than one component could bring results with greater correctness. Although some studies already measure the user’s emotional state to promote the adaptation of user interfaces and promote a better interaction experience, few worry about leading the user to achieve a desired emotional state. This work presents a new version of UIFlex (PROENÇA; NERIS, 2017), UIFlex 2.0. The previous version is presented as a Google Chrome plugin and it is responsible for providing interface adaptations to improve the accessibility of web pages. To do so, authors created rules of adaptation in JSON (JavaScript Object Notation) format that “injected” code into the web pages. The new proposed version brings together two major changes: (1) the architecture of the solution, which is now based on the MAPE-K model (Monitor-Analyse-Plan-Execute over a shared Knowledge) (KEPHART et al., 2003); (2) new user interface adaptation rules in order to provide color change in it according to theoretical studies published previously. Finally, a double blind experiment was conducted with 44 users in which two tasks were proposed – reading and transcript – on pages with the plugin enabled. Participants of the control group had access to UIFlex 2.0 as participants of the placebo group used UIFlex 3.0, which did not perform any adaptation. Both groups had the majority of participants reaching the desired emotional state – as can be seen in the generated incidence graphs. In addition, the Chi-Square statistical test was applied, which denied the alternative hypothesis. Thus, it is suggested that new rules be developed so that there are a greater number of changes to interface elements.