A growing number of psychologists are learning how to construct and make use of large data sets, or “big data,” to gain new insights into human behavior. To do that, they’re learning analytic techniques and applying tools that go hand in hand with big data, in particular artificial intelligence (the simulation of human intelligence processes by machines) and machine learning (computers’ ability to learn from data without being explicitly programmed to do so).
For behavioral scientists, big data can come from a wide variety of sources, ranging from traditional large-scale databases to medical records to unstructured data gleaned from cell phones, social media, and wearable technology.
These data and related methodologies allow researchers to study more types of constructs and variables than ever before and to be much more exploratory than traditional methods allow—to take a speculative approach to generating hypotheses rather than testing given ones, said Sean Wojcik, PhD, a social psychologist and senior data scientist at the news media company Axios who uses and studies these techniques.
“There are a lot of benefits for researchers to become adept in these skill areas, because there’s so much to be learned from these data,” he said.
An example is the National Institutes of Health’s (NIH) Adolescent Brain Cognitive Development (ABCD) Studyopens in new window, the largest long-term study of brain development and child health in the United States.
The study “is collecting a vast amount of data from participants through self-report questionnaires, cognitive testing, brain imaging, and biospecimens,” said ABCD Director Gayathri Dowling, PhD, who works for the National Institute on Drug Abuse. “It is also bringing in data from sources like FitBit and cell phone apps as well as environmental data from existing data sets based on where participants live.”
“Together,” she said, “this rich data set will help us better understand the many different factors that influence developmental trajectories, from the individual to the family to the community and society.”
Studies like the ABCD Study exemplify how big data can be used with scientific and ethical integrity. ABCD overseers, for example, promote best practices for researchers, which include the use of appropriate statistical models, ethical considerations for interpreting findings, and more. But for other ventures, including those in the commercial realm, big data’s promise sometimes eclipses its proper use and understanding, and in general there’s a lack of sufficient oversight in the area, those working in the area noted. Unresolved ethical issues in the mental health realm, for example, include concerns about respecting patient autonomy and privacy and ensuring equity.
“The very reason that private information is private in the first place is why it is likely to be so robustly associated with mental health functioning and therefore why it’s so helpful to researchers and mental health providers,” said Benjamin W. Nelson, PhD, a clinical research scientist at Meru Health, an online health care provider that uses big data techniques to provide evidence-based mental health services. “So there needs to be strong regulations and ethics around the use of these data.”
Why the buzz?
Concerns notwithstanding, big data methods are gaining support because of their potential to help scientists, including psychologists, improve on the accuracy and breadth of existing findings and unearth new ones. With these techniques, researchers can take large data sets and reveal complex associations, including nonlinear and interactive associations, said Arizona State University professor Kevin Grimm, PhD, who studies machine-learning techniques. With machine learning, for example, “you don’t have to specify much [about the associations or variables that you want to test] ahead of time—the model essentially learns the nature of the associations from the data,” he said.
These techniques can also enhance traditional studies, Grimm said.
“Once you’ve analyzed your data using traditional techniques, you can use machine-learning techniques to discover additional associations that you might not have been aware of,” he explained. “And that might lead to new hypotheses and theories, which could then be tested with a new set of data.”
Another potential benefit: helping to address psychology’s problem of replicability, Wojcik noted.
“There’s been a lot of discussion about the validity of many psychology studies that relied on smaller sample sizes and about whether researchers are able to replicate some well-known effects,” Wojcik said. “With greater data volume, you can often have greater confidence that the effects you’re measuring are valid and reliable,” he said, “and that increases the likelihood that these data may generalize to larger populations or to new populations that you wouldn’t have studied otherwise.”
Seizing opportunities
Given the potential, many psychologists are finding creative ways to use big data and related methodologies. At Axios, for example, Wojcik is using machine-learning techniques to analyze language patterns, with the aim of spreading news content to more people. Similarly, he worked for several years at Upworthy, a digital “positive storytelling” platform. There, in collaboration with the Gates Foundation, he used social and data science to help stories on global health, poverty, and other significant topics go viral.
Others are applying these techniques to long-standing areas of study, both for commercial purposes and to ramp up their findings on solutions to important issues. University of Oregon clinical psychology professor Nick Allen, PhD, a youth mental health researcher, for example, used big data analytics to address a common problem he saw in mental health care: that kids often experience mental health crises in between therapy sessions, when they’re most in need of help and providers are least available. With a team of programmers and data scientists, he figured out how to capture, via teens’ phone sensors, potential indicators of mental health problems between sessions, such as changes in sleep patterns, physical activity, and social interactions.
Eventually, that work led to the creation of Allen’s company, Ksana Health, which offers products that allow researchers and mental health clinicians to integrate continuous monitoring of behavior into their projects and practices via cell phone data and big data analytics. The research software is already being widely used by researchers across the globe, including in the ABCD Study. The clinical software, being tested with select clinical partners, will be available early this year.
Nelson, meanwhile, is helping Meru Health use big data techniques with a team of research and data scientists to provide evidence-based care to thousands of people experiencing depression, anxiety, and burnout. Similar to Allen’s platform, Meru Health collects individualized program-engagement data and other self-report metrics to gather information on patient symptoms and treatment app use. In turn, this information helps Meru Health’s licensed therapists understand when to intervene with patients and provides patients with digital reminders related to their conditions, Nelson explained. Given big data’s role in this enterprise, using big data to improve daily patient interactions directly supports Meru’s ambitious goal “to treat 10 million lives by 2027,” he noted.
Basic researchers, too, are making the most of big data. Child and adolescent psychiatrist Armin Raznahan, MD, PhD, chief of the Section on Developmental Neurogenomics at the National Institute of Mental Health (NIMH), taps into large data sets such as the UK Biobankopens in new window, the Healthy Brain Networkopens in new window, and the ABCD Study to analyze behavioral, genetic, and neuroimaging data to better understand how brain organization in healthy individuals varies as a function of sex, genotype, and behavior. He then uses those insights to shed light on risk factors for mental health difficulties.
Thanks to the NIMH’s Intramural Research Program, which provides statistical help to NIH researchers on big data analytics, Raznahan has been able to gather and study in-depth, multi-modal data on approximately 10,000 young people up to age 17. Recent findings from an analysis of more than 2,000 brain scans, for example, found strong evidence for sex differences in the volume of certain regions in the human brain—a finding that had been previously shown only in mice (Proceedings of the National Academy of Sciences, Vol. 117, No. 31, 2020). Those findings have implications for our understanding of well-established sex differences in cognition, behavior, and risk for psychiatric illness, Raznahan noted.
One reason the findings are so trustworthy, Raznahan added, is that the researchers were able to compare their findings with thousands of images from an unrelated data set from the UK Biobank and found them highly consistent—another potential plus of big data.
Getting on board
Whereas some psychologists are becoming more familiar with big data, for others it can seem like a dauntingly high mountain of new concepts and techniques to master. Besides learning to work with big data sets, big data analysis involves learning multiple techniques related to acquiring, managing, and analyzing those data.
Fortunately, there’s lots of help available for learning these techniques. For one thing, colleges and universities are offering more programs, initiatives, and even degrees in these areas. And although psychology departments don’t always provide such courses, students can often take them in other departments “and really expand the potential of what they can do, as well as the tools they have available to them,” Wojcik said. Learning programming languages such as Python, R, or SQL, for example, can be foundational support for this work, allowing for extremely rapid data collection, he noted.
Meanwhile, the online universe offers a wealth of training opportunities, including summer camps like Neuromatch, Statistical Horizons, and Stats Camp, as well as trainings and certificate programs. You have to be ready to do a lot of learning on your own, as well, Grimm added—but there’s plenty of community support on online forums to help troubleshoot problems that arise.
Whether psychologists take the plunge to learn these methods on their own or work in tandem with data collection experts, it’s smart to take an interest because of big data’s potential to improve research and the field at large, Wojcik said.
“As a psychologist, I discovered that there are really rich sources of meaningful behavioral data that are just as interesting, if not more so, than conventional survey data that I might have collected in the laboratory.”
