Emotion as contextual information is increasingly recognized to be essential variables to web personalization but is still an underresearched topic. Poor assumptions are made about user preferences as current personalization technology only gather historical data, without considering real-time information besides factors like time, location, and device. As a consequence, consumers have a bad user experience because the content is deemed to be irrelevant. The researcher introduces contextual personalization as a system-initiated process that considers the user's current emotional state as contextual information to address the individual user with relevant content.
This master's thesis investigates how providers can offer a better user experience by utilizing emotions as part of contextual information in the personalization process. By employing a mixed-method design, combining both qualitative and quantitative methods from user-centered design and affective computing, user needs and user behavior during specific emotional states were identified. The following investigation is twofold; the first part involves identifying emotional states with current emotion detection technology on the market, as well as empirically validate the effectiveness and reliability of the software. The second part of the project investigates whether the user's emotional state affects the user experience and the user's perception of content relevance. By using methods adapted mainly from previous studies, participants were induced with a specific emotion before interacting with a website. The results in the first experiment suggest that current face analysis technology can detect the intended emotion from even microexpressions. The results in the second experiment did not show significant differences between the two groups on perceived content relevance or the user experience. The data still yields indications that there is an emotional affect in interaction.