Choosing Prediction Over Explanation in Psychology: Lessons From Machine Learning

Abstract

Psychology has historically been concerned, first and foremost, with explaining the causal mechanisms that give rise to behavior. Randomized, tightly controlled experiments are enshrined as the gold standard of psychological research, and there are endless investigations of the various mediating and moderating variables that govern various behaviors. We argue that psychology's near-total focus on explaining the causes of behavior has led much of the field to be populated by research programs that provide intricate theories of psychological mechanism but that have little (or unknown) ability to predict future behaviors with any appreciable accuracy. We propose that principles and techniques from the field of machine learning can help psychology become a more predictive science. We review some of the fundamental concepts and tools of machine learning and point out examples where these concepts have been used to conduct interesting and important psychological research that focuses on predictive research questions. We suggest that an increased focus on prediction, rather than explanation, can ultimately lead us to greater understanding of behavior.

Memory consolidation reconfigures neural pathways involved in the suppression of emotional memories

“…Finally, we used a nonparametric testing approach to test for the statistical significance of the model by generating 1,000 surrogate data sets under the null hypothesis of r (predicted, observed) (ref. 68). The statistical significance ( P value) of the model was determined by measuring the percentage of generated surrogate data that are greater than the r (predicted, observed).…”

Building a Science of Individual Differences from fMRI

“…Correlation analysis relies on in-sample population inference and does not directly ensure the generalizability of the established relationship to out-of-sample individual subjects [133,134]. Shifting to a predictive framework is necessary to ensure generalizability and to interpret fMRI-derived statistics at the individual subject level [17,135] (for a more in-depth discussion of adopting a predictive machine-learning inspired framework, and the proper use of training, validation and test datasets, see [136]). To fully control for remaining confounds, fMRI-derived statistics should be included in a full model alongside other potential predictors, and the unique predictive power of fMRI features should then be assessed through their selective removal from the model (as demonstrated in [16]) (see Figure 1(e)).…”

“…While it might seem that the cumulative output from many underpowered studies should eventually converge to a reliable conclusion through meta-analysis, this is not the case because of the strong bias to publish only significant findings [136,141] (Figure 5c). The reporting of all results regardless of their significance in the null hypothesis significance testing (NHST) framework, which could be implemented through initial pre-registration to ensure the quality of the methods [142], would be a way for studies with small sample sizes to contribute unbiased information for meta-analysis.…”

Big data in social and psychological science: theoretical and methodological issues

“…The theory-based focus increases our understanding of causal relationships in psychological processes and underlying mechanisms of social phenomena. However, with the emergence of big data research where computer scientists often use data-driven methods such as machine learning, social scientists have started to adopt bottom-up data-driven approaches that favor prediction over explanation [35]. For example, Schwartz et al [31] proposed an open-vocabulary differential language analysis (DLA) approach to predict personality from Faceboook status updates.…”