Detection and Integration of the Big Five Personality Traits for the Development of User-Centered Systems

Affecting how people behave in everyday life, personality has been demonstrated to correlate with user preferences and search habits in several contexts. For this reason, personality constructs find compelling applications in information systems, where they can be used to develop user-centered approaches. However, the modeling and integration of personality into adaptive tools are still in their early stages. This thesis explores the adoption of the Big Five personality traits (OCEAN traits) in adaptive systems, aiming to tackle some of the main issues related to this topic. The use of these psychological dimensions in personalization tools requires both a method to assess users’ personality profiles and a strategy to adjust a system’s content or interface based on the available information. Before delving into these aspects, we introduce the research domain by providing background knowledge on personality theories, recognition techniques, and various types of psychologically aware information technologies. At first, this thesis focuses on studying techniques to acquire the OCEAN traits. Finding novel methods to automatically detect personality without bothering users presents a significant challenge. We consider the possibility of utilizing aesthetic preferences for visuals to predict individual scores for the traits. The proposed approach involves employing a Siamese Neural Network to simultaneously extract features from multiple images liked on social media, which serve as input for a regression component. Then, practical scenarios in which personality information can be used are discussed. In many systems, personality profiles are processed to derive further knowledge on individual choices for specific items. However, further research is needed to investigate other possible applications. To address this challenge, we explore the adoption of the Big Five model to predict preferences for contextual factors, which represents a topic that still needs to be fully explored. In particular, this thesis iii illustrates the outcomes of a survey-based study analyzing correlations between the OCEAN traits and preferences for contextual factors in itinerary recommendation. To contribute to the development of recommender systems that integrate the latter, we also propose a genetic algorithm-based architecture that creates tours personalized for several time-related aspects. Finally, we investigate how personality can be used to design adaptive conversational agents, exploring the relationships between individuals’ Big Five traits and their perception of artificial personalities. A qualitative and quantitative analysis discusses the results of a user-based experiment in which participants interacted with and evaluated three chatbots, each modeled with diverse personality profiles. The work offers new perspectives on the development of psychologically informed conversational agents. The thesis concludes with some reflections on the studies presented, discussing critical aspects related to their application in real-world scenarios