Smartphone-based suicide risk
Passive sensing — screenshots, keyboard input, screen-time patterns — to study suicidal ideation and behavior as they unfold in daily life. Recent work in JAMA Network Open, npj Digital Medicine, and JMIR Mental Health.
ML in clinical psychology
How measurement quality and analytic choices shape the reliability of machine learning applied to psychological data. Influential commentaries in Clinical Psychological Science and Perspectives on Psychological Science.
Regularized structural equation modeling
A decade of work extending regularization (lasso, elastic net, stability selection) to latent-variable models, beginning with the original 2016 paper and the regsem R package.
Research Program
Digital phenotyping and suicide risk. I lead a program of research that uses smartphone-based passive sensing — including high-resolution screenshots, keyboard input, and screen-time patterns — to study suicidal ideation and behavior as they unfold in daily life. Recent work includes studies in JAMA Network Open on nighttime smartphone use as a marker of next-day suicide risk, npj Digital Medicine on screen-time captured through screenshot data, and JMIR Mental Health using vision-language models to predict momentary suicidal ideation from on-device screenshots.
Ecological momentary assessment and intensive longitudinal data. Much of my methodological work concerns the analysis of intensive longitudinal data collected from clinical populations — handling momentary missingness, zero inflation, continuous-time dynamics, and computerized adaptive testing for in-the-moment risk assessment.
Regularized structural equation modeling. Beginning with the original Regularized Structural Equation Modeling paper (2016) and the regsem R package, I have contributed to a line of work extending regularization (lasso, elastic net, stability selection) to latent-variable models, and to understanding how measurement quality shapes the conclusions of machine learning applied to psychological data.
Machine learning methodology in psychology. A second methodological strand examines the reliability of machine learning claims in clinical psychology — including a frequently-cited commentary in Clinical Psychological Science on inflated prediction performance in suicide risk modeling, and a paper in Perspectives on Psychological Science on the often-overlooked role of measurement error in machine learning applications.
Book. Machine Learning for Social and Behavioral Research (Guilford Press, 2023), with Kevin Grimm and Zhiyong Zhang — part of the Methodology in the Social Sciences series.
Background
I joined the Center for Healthy Minds at the University of Wisconsin–Madison in 2024. From 2017 to 2024 I was Assistant Professor of Psychology at the University of Notre Dame. I received my PhD in Psychology from the University of Southern California in 2017.