People in close relationships are closely and inextricably interconnected—sharing in their
moment-to-moment emotional, physiological, and behavioral reactions, both to each other and to their shared external environment. While psychologists often consider how individuals’ thoughts, feelings, and behaviors affect each other, I am interested in how our own thoughts, feelings, and behaviors impact people with whom we are close, and vice versa. Relationships are made up of complex, intricate, and fast-moving patterns of interaction that can become reinforced, locked, or escalate, impacting our individual and interpersonal functioning and physical health. Therapists who specialize in relationships often target these patterns, attempting to change way couples and parent-child dyads speak and react to each other, in order to improve relationship functioning. Although an important focus of intervention, I am often left wondering exactly where and when to intervene in a chain of events. Couple and family processes typically unfold across time in everyday life and often involve seemingly mundane or insignificant events, such as small-scale insults or a lack of positive interactions. As such, I am interested in capturing and modeling time-based, naturalistically-occurring processes unfolding in couples’ day-to-day lives, in order to better understand what factors impact real-life interpersonal functioning, and to one day develop interventions based on these methodologies. In particular, I hope to identify how small-scale, everyday events and interactions culminate into later outcomes of interest, with a particular focus on how these processes relate to relationship aggression. The four papers that follow are the first products of the Couple Mobile Sensing Project, a collaboration between the Family Studies Project in Psychology and the Signal Analysis and Interpretation Laboratory in Electrical Engineering. The project grew out of my NSF fellowship application, which proposed measuring couples’ electrodermal activity (EDA) in everyday life.
With input from many people, the project has grown from there, and now includes many other measures, including heart rate (HR), GPS, and audio recordings. Collecting, processing, and analyzing the various data streams, collected over a 24-hour period, proved to be a difficult task, requiring close collaboration with our engineering colleagues, and has resulted in exciting new directions for our project. The first paper included here is a methodological, proof-of-concept article that is in press for the special issue on new developments in research methods at Social Psychological and Personality Science (Timmons et al, in press). This proof-of-concept paper introduces the methodology of the Couple Mobile Sensing Project and discusses the utility of “ambulatory big data,” a term I use to denote intensive, high volume, heterogeneous data streams collected in real-life settings. In the paper, I describe our methodology and present two mini-illustrations of how these methods can be applied to study relationship processes. I also discuss the challenges associated with collecting these data, such as data processing, analysis, security, ethics, and privacy. In addition, I discuss how these methods can be applied across multiple contexts, to test theory-based hypotheses, as is often done in psychology, and to use exploratory techniques, such as machine learning. The papers that follow provide examples of how this can be done, with Paper 2 employing machine learning and Papers 3 and 4 using multilevel modeling to test psychological theories. The second paper included here is the result of a collaboration with electrical engineers Theodora Chaspari and Dr. Shrikanth Narayanan (Timmons et al., 2017). Our goal in this paper was to use machine learning methods to automatically detect couple conflict in daily life. In this paper, which was published in IEEE Computer in March 2017, we used audio, GPS, EDA, and HR data to correctly identify when romantic partners expressed annoyance to one another in approximately 80% of cases, depending on the combination of features examined. These results, though preliminary, are the first step towards developing just-in-time adaptive interventions to improve couple functioning. The third paper uses Linguistic Inquiry and Word Count Software (LIWC) to examine how word usage fluctuates according concurrent relationship processes, such as feeling annoyed or close with one’s partner, and tests whether these patterns of covariation are associated with the amount of aggression in the relationship more generally. This paper is the first to our knowledge to investigate how language use recorded in real-life settings fluctuates across the day according to ongoing interpersonal dynamics. In the fourth paper, we examine how physiological reactivity during naturally-occurring periods of irritation between partners relates to family-of-origin aggression and dating aggression. This paper tests patterns of physiological reactivity as risk factors in the intergeneration transmission of aggression and also examines gender differences in these processes. It is notable that we find several significant links with family-of-origin and dating aggression, given that we measured small scale, naturalistically occurring periods of annoyance in the context of daily life. Finally, in the supplemental materials, I include data comparing EDA collected using ambulatory monitors to EDA collected with standard in-lab devices.