Emotions contribute to essentially every human behavior. We decide what to do based partly on our emotions, we learn from our emotional experiences, and we interact successfully with other people by recognizing their emotions. If we could predict emotional responses accurately, it would have implications for a broad array of domains, ranging across (but not limited to) close relationships, education, parenting, business management, consumer behavior, work performance, health behaviors, conflict resolution, and negotiations. Unfortunately, so far we cannot predict emotions accurately, mostly because they emerge in complex ways across time and involve multiple changes across short time spans, including subjective experience (e.g., feeling angry), behavior (e.g., frowning, hitting someone), and physiology (e.g., blood pressure, heart rate). In addition, most strong emotions occur during social interactions, making it necessary to simultaneously predict the emotions of all the people involved. Dr. Emily Butler (University of Arizona) and her colleagues refer to these multi-person emotion systems as "Temporal Interpersonal Emotion Systems (TIES)."
Social psychologists have developed sophisticated theories about how TIES operate, but using those theories to predict actual behavior has been thwarted by a lack of mathematical modeling expertise. This research brings together expertise in TIES and Bayesian modeling, which is a powerful mathematical approach for predicting the behavior of complex systems. For example, Bayesian modeling has been used successfully for problems as diverse as weather prediction, controlling missiles, and detecting credit card fraud, but has rarely been applied to human emotional processes. Bayesian methods and existing data from three prior studies will be used to predict health behaviors and outcomes (e.g., weight gain, cancer recovery) from interpersonal emotional processes in the context of adult romantic relationships. To ensure broad impact, the iPlant collaborative, a large NSF funded project to create shared computational infrastructure for 2020 science, will provide computational capability in the form of web services, whereby researchers choose TIES models, submit data, fit models using software running on a super-computer, and retrieve results over the web. By providing support to other research groups to make their models available through iPlant, a wide range of powerful TIES models will become available, broader web dissemination will be achieved, and it will become easier to compare models and methods. The results of this work will inform family-level health-promotion interventions. More broadly, the larger goal is to foster cumulative research that supports pragmatic applications. Information about TIES is important for developing parenting classes, family counseling programs, interventions for health behaviors, managerial and negotiation training, reducing bullying in schools, and promoting constructive international relations. Together these efforts will provide the broad scientific base needed for transforming theoretical knowledge of TIES into pragmatic applications.
