Introductions
The EJP special issue on Behavioral Assessment in the Age of Big Data is out now and contains amazing and innovative work in this developing area of research! We had the opportunity to talk with authors of several papers about their interest in the research area and their own work.
Specifically, we talked with Alex Danvers about his paper titled, “Understanding Personality through Patterns of Daily Socializing: Applying Recurrence Quantification Analysis to Naturalistically Observed Intensive Longitudinal Social Interaction Data“, with Sang Eun Woo and Louis Tay about their shared first-authored work, “Psychometric and Validity Issues in Machine Learning Approaches to Personality Assessment: A Focus on Social Media Text Mining“, and with Ramona Schödel about her paper, titled “To Challenge the Morning Lark and the Night Owl: Using Smartphone Sensing Data to Investigate Day–Night Behaviour Patterns“.
Read more about their research interests and research below.
Q: Hello all, can you tell me a little about yourself and your research?
Alex: I am a social and personality psychologist who studies emotions and social interactions using advanced quantitative modeling. Recently my work has focused on measuring daily life outside the lab using smartphones and other sensors (like a Fitbit or Apple Watch). Finding out what people are doing “in the wild” is a great way to better understand human behavior and psychology.
Sang: I’m an associate professor of Industrial-Organizational Psychology at Purdue University. I am interested in pushing the boundaries of our field of science by incorporating multiple modes of research inquiries beyond what is considered conventional. I study the topic of openness, which entails both scientific (intellectual) and cultural (relational) willingness to try new things and to appreciate the diversity of perspectives and approaches.
Louis: I’m an associate professor of Industrial-Organizational Psychology at Purdue University. I love science, data, measurement, and analytics. Not surprisingly, my research focuses heavily on methods and data science. My substantive area of research is in well-being.
Ramona: I originally started studying psychology because I have always been interested in how and why people differ from one another. During my studies, I then discovered my passion for human-computer interaction. Fortunately, my Ph.D. studies with Prof. Markus Bühner at LMU Munich offered me the opportunity to work on an interdisciplinary project combining personality psychology, computer science, and statistics. This is where I received my Ph.D. in July of this year and where I have been working as a postdoc ever since.
My current research interests lie in the vast amounts of digital footprints such as smartphone usage data that people create every single day. More precisely, I use these new types of data to study individual differences in everyday lives. One behavior that is of particular interest to me is sleep-wake rhythms. Sleep is an essential human behavior that has diverse effects on psychological and physical wellbeing. New types of digital data allow me to investigate individual differences in sleep-wake behavior at an unprecedented scale. A second focus of my research is the exploration of new data types from a more methodological perspective, as they offer great opportunities, but also challenges such as the extraction of meaningful variables. In this context, I learned a lot from other disciplines, such as computational science in the past years. I think psychological science can benefit greatly from applying modeling approaches like machine learning to fully exhaust the informative power of these large amounts of digital data.
Q: What makes you especially excited about behavioral assessment in the age of big data?
Alex: The psychological literature is full of great ideas that have been tested on small groups of undergraduates in artificial lab situations. With the advent of the internet, wearable devices, and more behaviors moving online (e.g., conversations happening on Twitter or other social media sites), we can suddenly test our ideas on much larger samples—10 or 100 times as many people. This means that we can say with much more certainty whether our ideas are right, and we can say whether they apply to more general groups of people—all the people who might use the internet or smartphones, as opposed to only undergraduate students in psychology. Even more exciting, by capturing behavior over time, we can start to ask questions about patterns over time—how things change, and whether there are cycles to behavior. This hasn’t been feasible for the lab studies conducted in previous generations.
Sang: With the help of technology, we now have access to a number of intentional (active) and subconscious behaviors generated by humans. Such access to new types of data opens up countless possibilities of detecting new phenomena and subsequently developing new theoretical understandings of human behaviors.
Louis: Using big data for behavioral assessment is the frontier of psychological research and there is so much to explore, understand, and learn. In particular, Sang and I are both interested in examining (and finding new ways for examining) the reliability, validity, and utility of these new modes (e.g., social media, wearable sensors, videos) for assessing individual differences and well-being. In addition, we hope to see more systematic investigations into the issues of bias and fairness in algorithmic judgment and decision making applications used in employment and educational settings.
Ramona: In a nutshell, I am excited by the variety of possibilities to capture naturally occurring behaviors across longer periods. Let me give you a more detailed example: Every morning, I wake up to the sound of my alarm clock app. As a night owl, the first thing I do in the morning is to have a coffee while checking the news on my smartphone. Before I leave the house, I briefly reach for my smartphone to find out about the weather, and when the next subway is leaving. I could go on and on to tell you about my daily activities, most of which include some sort of digital devices. What I want to illustrate is that big data is ubiquitous and provides us with the possibility to conduct psychological research “in situ” in people’s everyday lives. This new form of digital assessment helps us to find out more about personality and its behavioral manifestations. Amongst the potential tools for behavioral data collection, the smartphone fascinates me the most. We all own smartphones and they accompany us everywhere we go and in everything we do because they have so many functionalities beyond mere communication. Consequently, these small supercomputers have so much to tell about their users’ behaviors across various situations.
Q: What is your study about?
Alex: Our study examines the way that patterning in people’s social lives over time is related to their personality. Participants in our study wore a smartphone that made recordings throughout the day, and trained research assistants coded whether they were talking in each file. We then borrowed a technique from statistical physics to capture the underlying patterns in talking throughout the day. Our results show that personality traits are correlated with these patterns of social activity. Earlier studies had shown that extraversion was related to how much you talk, but this study expanded that understanding to show extraversion is also related to how long you spend talking to people (unsurprisingly, extraverts have longer periods of extended socializing). Our approach also uncovered a surprising relationship: people higher in neuroticism (a tendency to experience more negative emotions) have shorter conversations in general. We think this might be related to social anxiety, with neurotic people feeling less comfortable in interactions and so cutting them short. It could also be related to the people talking to neurotic individuals, though: neurotic people might make others anxious, causing them to cut conversations short.
Sang and Louis: Researchers and practitioners increasingly use social media text mining (SMTM) to assess the personality of their users. However, there are many questions that are still unanswered in this practice. We systematically review past research and show that SMTM has some predictive accuracy for assessing personality, we discuss ways we need to evaluate SMTM predicted personality scores, how personality conceptually relates to language use and expression on social media, and whether different social media platforms are equivalent for assessing personality.
Ramona: Have you ever heard of “morning larks” and “night owls”? Morning larks represent early risers, who are immediately fit in the morning but go to bed early in the evening. Night owls, on the other hand, stand for evening-orientated persons, who take a long time to wake up in the morning, have their peak performance later in the day, and stay up late at night. These differences in sleep-wake timing constitute a trait called chronotype. This line of research is still dominated by the use of self-report measures about circadian preferences and sleep-wake timing habits.
However, the increasing digitization of our everyday lives opens up new ways for the investigation of chronotypes. Our exploratory study illustrated how researchers can access new types of data to complement the traditional self-report approach. We used a research mobile application called PhoneStudy to collect participants’ smartphone usage behaviors in the field for several weeks. Based on these smartphone sensing data, we derived behavioral variables such as the timing of the first and last smartphone usage events of the day. These variables helped us explore inter- and intraindividual differences in naturally occurring day-night patterns. More specifically, we focused on three research questions: We explored whether 1) "morning larks" and "night owls" manifest in day-night patterns of smartphone usage behavior, 2) how naturally occurring day-night patterns relate to personality traits, and 3) whether traits, as well as, day-night activity patterns during weekdays are associated with day-night activity on weekends.