Predicting the Past: Testing Expert Historical Judgement in the Humanities and Social Sciences

This project pilots a novel method for testing historical judgement that has the potential to transform how scholars and policy-makers research, assess, and learn from the past. Gaps in the historical record present significant challenges but also valuable opportunities. Instead of avoiding gaps, scholars and practitioners can test the accuracy of inferences about the past by carefully outlining their assumptions and explicitly predicting what they believe to have occurred in the absence of evidence. Subsequent discoveries or declassifications can then be used to assess the accuracy of these hypothesised explanations and, when collected and analysed, enable better predictions about the past, or “retrodictions.” The importance of making and testing predictions about the future has been well documented. Yet there has been limited effort to make, collect, and assess predictions about the unknown past. Our interdisciplinary pilot project addresses this gap, drawing on the expertise of political scientists, historians, psychologists, and relevant practitioners to establish a proof of concept for how to identify, assess, and improve the practice of retrodiction. As a pilot study, it will test the viability and power of retrodiction, aiming to answer three questions: what are the most effective ways to collect and assess retrodictions? What lessons can be learned from accurate and inaccurate predictions about the past? How far are these lessons generalizable? By answering these questions, we will provide a springboard for a new research agenda on retrodiction. We will work with our Project Partner (Moody’s Analytics), key stakeholders (Forecasting Collective [FC], HM Treasury), and three supporting Labs (the Rights Lab, the Wisdom and Culture Lab, and the Decision Neuroscience Lab) to collect and test retrodictions in three different subject areas: human rights, nuclear proliferation, and sovereign debt. These choices reflect salience and relevance, availability of data, and our own expertise. As few testable examples exist, we will identify and test up to 300 retrodictions. Our methods for data collection include novel retrodiction tournaments. We will assess the accuracy of historical judgement by using existing forecasting criteria and standards supported by industry experts. Our findings will generate a novel dataset for testing and comparison. We will identify lessons from our findings, examine the rationales for predictions, and share findings with forecasting experts to establish broader generalisations. This project’s contribution and value reflect the creation of a new mode of inquiry to generate original data and novel insights, moving beyond the constraints of only predicting the future. Using retrodiction to test existing hypotheses about why forced labour persists or how states manage sovereign debt, for example, may help to improve understanding of both subjects and refine thinking about enduring real-world problems. By providing innovative methodologies, we aim to assess the efficacy of previously untestable work, promote better scholarly practice, and improve historical judgement across multiple disciplines. We believe this pilot can set a new research agenda by promoting a more rigorous, testable set of methodologies that will benefit scholars and policymakers who seek a clearer, richer, and more usable past.