A Human-Centered Perspective of Intelligent Personalized Environments and Systems

In the current digital landscape, humans take center stage. This has caused a paradigm shift in the realm of intelligent technologies, prompting researchers and (industry) practitioners to reflect on the challenges and complexities involved in understanding the (potential) users of the technologies they develop. In this chapter, we provide an overview of human factors in user modeling, a core component of human-centered intelligent solutions. We discuss critical concepts, dimensions, and theories that inform the design of user models that are more attuned to human characteristics. Additionally, we emphasize the need for a comprehensive user model that simultaneously considers multiple factors to represent the intricacies of individuals’ interests and behaviors. Such a holistic model can, in turn, shape the design of intelligent solutions that are better able to adapt and cater to a wide range of user groups.

This chapter explores the principles and frameworks of human-centered Artificial Intelligence (AI), specifically focusing on user modeling, adaptation, and personalization. It introduces a four-dimensional framework comprising paradigms, actors, values, and levels of realization that should be considered in the design of human-centered AI systems. This framework highlights a perspective-taking approach with four lenses of technology-centric, user-centric, human-centric, and future-centric perspectives. Ethical considerations, transparency, fairness, and accountability, among others, are highlighted as values when developing and deploying AI systems. The chapter further discusses the corresponding human values for each of these concepts. Opportunities and challenges in human-centered AI are examined, including the need for interdisciplinary collaboration and the complexity of addressing diverse perspectives. Human-centered AI provides valuable insights for designing AI systems that prioritize human needs, values, and experiences while considering ethical and societal implications.

This chapter discusses the ethical complications and challenges arising from the use of AI systems in our everyday lives. It outlines recent and upcoming regulations and policies regarding the use of AI systems, and dives into the topics of explainability and fairness. We argue that trustworthiness has at its heart explainability, and thus we present ideas and techniques aimed at making AI systems more understandable and ultimately more trustworthy for humans. Moreover, we discuss the topic of algorithmic fairness and the requirement that AI systems are free from biases and do not discriminate against individuals on the basis of some protected attributes. In all cases, we present a brief summary of the most important concepts and results in the literature. As a conclusion, we present some ideas for future research and sketch open challenges in the field.

In this chapter, we discuss how to utilize human memory models for the task of modeling music preferences for recommender systems. Therefore, we discuss the theoretical underpinnings of using cognitive models for user modeling and recommender systems in order to introduce a model based on the cognitive architecture ACT-R to predict the music genre preferences of users in the Last.fm platform. By implementing the declarative memory module of ACT-R, comprising past usage frequency and recency, as well as the current semantic context, we model the music relistening behavior of users. We evaluate our approach using three user groups that we identify in Last.fm, namely (i) low-mainstream music listeners, (ii) medium-mainstream music listeners, and (iii) high-mainstream music listeners. We find that our approach provides significantly higher prediction accuracy than various baseline algorithms for all three user groups, and especially for the low-mainstream user group. Since our approach is based on a well-established human memory model, we also discuss how this contributes to the transparency of the calculated predictions.

Today’s data-driven business environment necessitates the adoption of business intelligence and analytics (BI &A) platforms for companies that want to remain competitive or even survive in the global market. On the other hand, as the volume and variety of data grow, the ability to quickly analyze data and reach actionable decisions becomes more complex. In response, BI &A vendors have developed features, to assist users with data exploration, analysis, and delivery of on-demand timely business insights using interactive data visualizations. These tools also enable non-expert analysts, such as directors, to quickly translate data into actionable items. However, the large variety of tools and data visualizations may become overwhelming and diminish the end users’ ability to quickly reach a decision. Furthermore, BI &A vendors adopt the on-size-fits-all approach, delivering the same output to all users for identical requests, neglecting to incorporate their unique characteristics and personalize their interaction. This chapter presents the user adaptation and personalization techniques employed by leading BI &A enterprise solutions, and discusses the effect of individual differences in understanding and processing data visualizations. Our goal is to explore opportunities and challenges of incorporating individual differences in the visual analytics process, and highlight the necessity of a human-centered model in BI &A platforms.

In this chapter we survey methods for inferring two types of characteristics for personalized systems: eudaimonia and hedonia (E and H). The rationale for focusing on these two characteristics is the potential to make good recommendations and the even bigger potential for creating good explanations. We first conceptualize the concepts of E and H for the purposes of personalized systems by disentangling the user preferences from the item characteristics. We proceed on surveying methods for inferring EH user characteristics from digital user traces. We follow with an overview of methods for inferring EH item characteristics from item content. Finally we provide an outlook into the future work.

“The Eyes Are the Window to the Soul” is a commonly used phrase that refers to the intricate relationship between a person's eyes and their underlying thought processes and affective states. Indeed, researchers in perceptual and cognitive psychology have long studied a user's eye movements in order to understand attention patterns and emotional responses to stimuli. More recently, researchers in Human–Computer Interaction and Artificial Intelligence have started to study the potential of using eye tracking to infer a user's underlying human factors, in order to drive personalized systems that adapt to an individual user's abilities, traits, and states. This chapter provides an overview of eye tracking technology, as well as a review of the various eye movement features that may be calculated from eye tracking data. This is followed by a review of various modeling techniques that use these features to infer human factors, as well as a range of application scenarios and examples. Lastly, the chapter concludes with an outlook on the use of eye tracking technology for human factor inference, including both challenges and opportunities.

For many people, their first encounter with learning programming happens online. However, it is likely that these first-time learners will encounter obstacles that they cannot overcome on their own, especially as they progress to more complex concepts. Keeping these online learners engaged with the content is essential for them to learn programming, as their experience could have long-term implications for how they view computing. One way to address this issue is to detect when a learner is having difficulty with a concept, provide them with automated assistance and encouragement, and offer more opportunities to practice. For struggling learners, customized encouragement may be just what they need to re-engage with the task, and additional practice may help them better understand the concept(s) and prepare them for future topics. Many recent technologies advocate for the use of machine learning techniques to customize and personalize educational content for their users. However, less resource-intensive methods (and those not requiring time and other resources to train and test models) utilizing users’ historical interaction data may provide enough support to provide just-in-time customization and personalization to help learners. This chapter describes two studies where we created a simple, coarse-grained (instead of using complex machine learning detection methods) frustration detector to provide customized content for the user. We modified an existing programming game, providing encouraging messages and hints in the first study, and providing extra game levels for more practice when necessary in the second study. Based on our results, we show that simple, coarse-grained detection methods are sufficient to trigger adaptive interventions to benefit struggling learners.

Studies in psychology have shown various ways how humans can be influenced in their choices and behavior. Many of these persuasive strategies and nudges are now also used online, affecting how digital choice environments are designed. In the sustainability domain, these strategies have been used to promote specific pro-environmental behaviors, such as through green energy defaults and social norms (e.g., ‘75% of people re-use their towel’). Most of these nudges are, however, evaluated in one-size-fits-all interventions, not reflecting to an extent to which today’s digital environments are personalized. Not only does this call for smarter, personalized nudges, it also overlooks the fact that various nudges would be applied in tailored choice environments. In particular, recommender systems have become ubiquitous, directly tailoring advice to end users, which might deem nudges to become superfluous. Hence, it remains an open question whether nudging is still effective if the advice is also tailored. This chapter explores the effectiveness of different (smart) nudges in the context of tailored choice systems for household energy conservation. We have developed an approach for a psychology-informed recommender system that presents personalized, attitude-tailored energy-saving advice to end users. Our approach comprises an algorithm and interface nudges that are both personalized and operationalized through smart default and social norm interventions. We present the results of multiple studies performed with our energy recommender systems, providing evidence for the limited effectiveness of interface nudges in a personalized advice context. We discuss the design implications and what nudging and persuasion mean in a world in which most decisions are digitized and content is personalized.

Persuasive technologies for health and wellness are designed as interventions to promote desired health behaviours using various persuasive techniques. In recent years, a growing number of persuasive interventions have been used in various health domains, including mental health, physical exercise, and healthy eating. Most of these interventions take a one-size-fits-all approach. However, studies have shown that a persuasive strategy that is successful for one person may not be effective for others, indicating that persuasive interventions should be tailored to the individual and be adaptable to the user and their specific circumstances. Research indicates that an individual’s responses to persuasive strategies can be influenced by different factors, including personal characteristics (such as personality traits, age, and gender) and contextual factors (such as time, weather, and location). This chapter discusses personalizing persuasive health interventions. It first investigates the theories deployed and the personalization dimensions used in personalizing persuasive health interventions. Then, the chapter investigates how these theories are applied in practice and the different techniques used to deploy personalized persuasive health interventions, followed by a discussion about the challenges and directions for future research in this area.