1 Theory-grounded user modeling for personalized HCI

Abstract

Personalized systems are systems that adapt themselves to meet the inferred needs of individual users. The majority of personalized systems mainly rely on data describing how users interacted with these systems. A common approach is to use historical data to predict users’ future needs, preferences and behavior to subsequently adapt the system to cater to these predictions. However, this adaptation is often done without leveraging the theoretical understanding between behavior and user traits that can be used to characterize individual users or the relationship between user traits and needs that can be used to adapt the system. Adopting a more theoretical perspective can benefit personalization in three ways: (i) relying on theory can reduce the amount of data required to train compared to a purely data-driven system, (ii) interpreting the outcomes of data-driven analysis (such as predictive models) from a theoretical perspective can expand our knowledge about users and (iii) provide means for explanations and transparency. However, in order to incorporate theoretical knowledge in personalization a number of obstacles need to be faced. In this chapter, we review literature that taps into aspects of (i) psychological models from traditional psychological theory that can be used in personalization, (ii) relationships between psychological models and online behavior, (iii) automated inference of psychological models from data, and (iv) how to incorporate psychological models in personalized systems. Finally, we propose a step-by-step approach on how to design personalized systems that take users’ traits into account.

1.1 Introduction

Personalization is performed by adapting aspects of systems to match individual users’ needs in order to improve the user experience by making it more easy for the user to reach their goals in the system. Examples are recommender systems that make it easy to find relevant content in a library [94], or adaptive interfaces that make it easier for users to achieve their goals [91]. Current personalization strategies are mainly data-driven in the sense that they are based on the way users have been and are interacting with a system, after which the system is dynamically adapted to match inferred user needs. The more theory-driven counterparts of personalization are often designed based on general knowledge about how user traits influence user needs, and how these needs influence the requirements of a system. Systems are adapted to individual users based on a set of rules. Although both strategies are used separately, combining the knowledge gained from both strategies could be used to achieve greater personalization possibilities.

In order to provide approaches to personalization, current research has primarily focused on using historical data that describes interaction behavior. Using this data, personalization strategies are developed that predict users’ future interactions. The prediction of these future interactions is often done without leveraging the understanding of the relationship between user behavior and user traits. In other words, predictions are made without considering the root cause of certain behavior that users are showing. A prominent direction using this approach is the field of recommender systems in which historical behavioral data is used to alter the order of items in a catalog (from highest predicted relevance to lowest predicted relevance), with the goal of making users consume more items or helping them to find relevant items more easily [73].

By adopting a more theoretical perspective (often based on psychological literature), the root cause of behavior can be identified and thereby benefit personalization opportunities. Using a theoretical perspective can benefit personalization in two ways: (i) a large body of theoretical work can be used to inform personalized systems without the need of extensive data-driven analysis. For example, research has shown that it might be beneficial to adapt the way course material is presented to match students’ working memory capacity [45], and (ii) including theory can help to interpret the results gained from the data-driven perspective and thereby meaningfully expand our knowledge about users. For example, research on music players has demonstrated that different types of people base their decisions on what to listen to on different sources of information [31].

Although personalization has been shown to benefit from adopting a more theoretical perspective by considering the relationship between user behavior and user traits, this theoretical perspective comes with theoretical and methodological challenges. A first challenge is to identify and measure user traits that play a role in the needs for personalization (e. g., cognitive style [82], personality [16] or susceptibility to persuasive strategies [21]) and to capture these traits in a formal user model. A second challenge is to infer the relevant user traits from interaction behavior (e. g., inferring user preferences from historical ratings or inferring a person’s personality from the content they share on social media). A third challenge is to identify the aspects of a system that can or should be altered based on these user traits that can improve the user experience. In certain cases, this is straightforward (e. g., altering the order of a list of items based on predicted relevance), while in other cases the required alterations can be more intricate and require more thought to implement (e. g., altering the way in which information is presented visually to match a user’s cognitive style).

While the aforementioned challenges are interconnected, they are often addressed in isolation. The current chapter provides an overview of work that relied on user traits for several (system) aspects:

• introduction of psychological models that are currently used in personalization

• psychological models that have been linked to online behavior

• automatic inference of psychological models from behavioral data

• incorporating psychological models in personalized systems or systems for personalization

The literature discussed throughout the chapter can serve as starting points for theory-grounded personalization in certain applications (e. g., e-learning, recommendations) and content domains (e. g., movies, music). Finally, the chapter concludes with a blueprint for designing personalized systems that take user traits into consideration.

1.2 Psychological models in personalization

Psychological models serve to explain how aspects of the environment influence human behavior and cognition. Since these models provide information on how people react to their surroundings, they can also be used to anticipate how people will react to aspects of technological systems and can thus provide insight in people’s needs in technological contexts. The proposition to use psychological models for personalization is not a new concept. Rich [95] already proposed in 1979 the use of psychological stereotypes for personalizing digital systems. A more recent overview provided by Lex et al. [78] similarly elaborates on how psychological theory can be applied in specifically recommender systems, or personalized systems that aid people in finding relevant items based on their interaction behavior.

While the concept of incorporating psychological theory into personalization is not new, the current abundance of available user data have made personalization strategies adopt more data-driven approaches and move away from incorporating theoretical knowledge. While the availability of user data obviously benefits data-driven approaches, there are opportunities for theory-driven approaches as well to exploit the available data (e. g., the implicit acquisition of user traits). In the following section, we will lay out different models that are currently used in personalization. We will then continue with providing an overview of prior research that has focused on the relationship between psychological models and online behavior, then continue with work that has looked at the automated inference of psychological...