From Data to Discovery: Unsupervised Machine Learning's Role in Social Cognition

The study of how cognition and society interact is a complex endeavor that demands multiple methods and tools. Yet research in social cognition has only begun to capitalize on unsupervised machine learning (UML) tools that can uncover hidden patterns in data. In this tutorial we introduce UML as a complementary approach to traditional statistical methods. We illustrate four methods (K-means clustering, DBSCAN, PCA, and Market Basket Analysis) applied to data from Project Implicit and the Implicit Association Test. In the process, we show how UML can identify patterns and relationships that conventional methods might overlook. Throughout, we provide clear (and openly available) code and highlight important researcher decision-points in implementing UML in social cognition work. By bringing the advances of UML into social cognition we will be better equipped to tackle larger, more diverse, or multi-level datasets that reveal the complexities of our social world.