User interaction (UI) designs often cannot adapt to changes in user contextual conditions. While users usually expect UI designs to change based on changes in the conditions of external contexts, most current UI designs propose only a single design rule for all user actions in their context. Users must either accept the suggested action or modify their actions to adapt themselves to a new design-based condition. Each alteration may cause changes in the structure of human activity. This study proposes an emotion-centered design method that enables designs to adapt to a user’s contextual conditions. Changes in contextual conditions often cause users to experience different emotional states, and, the proposed design method chooses among most likely UI designs based on identified user emotional states. The emotion-centered design method is applicable to a wide variety of different human-computer interaction fields, including mobile and wearable computing, contextual computing, the Internet of things, affective computing, personal computing, etc.

The design method is implemented through user verbal interaction design methodology. Two user studies were run: In user study one, 19 participants viewed 14 examples of TV content, while in user study two, 27 participants interacted with 9 examples of multimedia content. Users then created text messages and reported their emotional states using mobile applications. Performances of Bayesian networking classifiers, used for detecting users’ affective states, were evaluated using two methods: 10-fold cross validation and leave-one-person-out. These two test studies achieved emotion recognition results approximately 80% of the time. These results show that learning classifiers can detect user emotional states in users’ present contexts. Two learning classifiers were also tested with user behavior features obtained from each other’s studies. Comparison studies demonstrated recognition rates of about 30%. Because UI designs reflect the assumption of action independence with respect to natural context, the dependency between user actions was set based on users’ requirements at that present time. In other words, context, as an entity connecting actions to each other, was created to help users with their activities. Whenever the users were done with the activity, the context was broken up, and users were not able to transfer results of previous experiences into the new context.