Mixed Feelings

In daily life, we tend to see an expression of emotion as a reflection of an individual's state (I feel), while in social settings, we are also regularly influenced by the emotions of others (we feel). This exchange of emotions largely occurs subconsciously and is commonly referred to as emotion contagion. In groups, it is believed that this exchange leads to emotional convergence, which can result in a collective emotional state. Consequently, the spread of negative emotions is also linked to certain behaviours of large groups, which can be dangerous or destructive and difficult to control. What if we could predict how emotions spread in groups? This dissertation focuses on whether multi-agent models can be used to simulate emotion contagion in groups. The ultimate goal is to predict emotional development in groups in real-time, which can serve as a tool for professionals to prevent or manage dangerous situations, as well as for training and planning purposes. Additionally, this dissertation aims to enhance scientific knowledge about emotion contagion in groups. Despite the risks in crowds, emotion contagion is in the first place an essential social process. It is therefore important to better understand and protect healthy exchanges of emotions in groups, especially given the rapid technological developments in how we communicate with one another. In the first half of this work, I contribute to several theoretical problems of existing models of emotion contagion. For example, I introduce a new model that provides a solution for situations where multiple emotions spread simultaneously within a group and explore the role of group identity. However, despite substantial theoretical work that has been done in the field, I find that empirical knowledge about how emotion spreads in groups remains limited, likely due to both practical and ethical challenges in collecting data. This limits model validation, while robust validation is essential to responsibly apply these models in practice. This conclusion motivates the second half of this dissertation, in which I focus on empirically studying emotion contagion in groups. As a first step, I introduce a new experiment in which the emotional reactions of participants are measured during group video calls. The results show that emotional expressions of those sharing a virtual space converge at the group level. Through eye-tracking, I also find at the individual level that a person's emotional expression becomes more similar to that of the individuals they watch. Additionally, I demonstrate how this method can improve validation by testing the assumptions underlying emotion contagion models. In the final part, the focus shifts to the effects of new technologies on emotion contagion. I investigate how well people can recognise emotional expressions from virtual humans. Yet, being able to recognise emotions is not the same as being influenced by them. To test this, I also repeat the earlier experiment, but this time the participants see virtual versions of the other participants mimicking their facial expressions in real time. Finally, I take a first step towards developing an application in the form of a virtual reality training programme for security personnel. I conclude that the technique of employing agent-based models for simulating emotions in groups is promising. However, simulating the psychological development of large groups remains highly challenging. Particularly for the real-time prediction of emotional development for important safety decisions, I see significant scientific and ethical concerns. In contrast, the use of emotion contagion models in virtual settings, such as for training, therapy, and planning, offers a realistic path to applications that can benefit society in the shorter term. This work has provided theoretical, methodological, and empirical contributions that form a step towards responsible application.