Machine-Assisted Social Psychology Hypothesis Generation

Abstract

Social psychology research projects begin with generating a testable idea that relies heavily on a researcher’s ability to assimilate, recall, and accurately process available research findings. However, an exponential increase in new research findings is making the task of synthesizing across the multitude of topics challenging and could result in potentially overlooked research connections. In this research, we leverage the fact that social psychology research is based on verbal models and train a natural language model to generate hypotheses that can serve as an aid to psychology researchers. Our hypotheses-generation model uses a pre-trained language model to generate text that is meaningful and coherent. We then fine-tune the model on thousands of abstracts published in more than 50 social psychology journals, in the past 55 years as well as on pre-print repositories (PsyArXiv). Finally, social psychology experts rated model-generated and human-generated hypotheses on the dimensions of clarity, originality and impact.