Applying Artificial Intelligence to Expand the Measurement Tool Kit in Clinical-Psychological Science: Moving Beyond Self-Reports.

Abstract

Research exploring correlates of, precursors to, and consequences of psychological disorders has often relied on designs wherein both predictor and outcome are measured by self-reports. In this article, coauthored by a clinical psychologist (C. E. Fairbairn) and a data scientist (N. Bosch), we offer information surrounding an evolving class of machine-learning models as these inform an expanding measurement tool kit in clinical-psychological science. Specifically, we note the development of deep-learning applications for image analysis, language analysis, and the analysis of physiological time-series data, reviewing implications of these advances for measurement in behavioral research. We weigh strengths and limitations of these automated methods in comparison with self-reports, including the specific form of error likely yielded via each (random vs. systematic), with the aim of fostering a replicable, sustainable, and reputationally strong field of clinical-psychological science.