Predicting the Past

Absences pervade the historical record. The loss or destruction of material, redaction of documents, silence of participants, data embargoes, and poor record keeping present inherent difficulties to any understanding of the past. Gaps in the historical record pose significant challenges but they can also provide valuable opportunities. Instead of avoiding all gaps, historians can test the accuracy of some 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 hypothesized explanations and, in turn, help us to evaluate the quality of historians’ thinking about the unknown past. Given enough examples, we can begin to learn more about how to make better predictions about the past or what we term “retrodictions.”1 The importance of prediction has been well documented across the social sciences for decades.2 Economists, political scientists, and psychologists regularly extol the virtues of making testable predictions for good scholarly practice and improved judgement. Although historians and historically minded scholars in the social sciences routinely rely on their own expert judgement when studying the past, few have considered making and testing explicit predictions when conducting historical research. Even fewer have considered the methodological demands and theoretical implications of retrodiction.3 This article considers why historians and social scientists should attempt to predict the past and explains how they can make and test retrodictions in a more rigorous manner. These conceptual and methodological insights are of considerable significance. Retrodiction is not an abstract or hypothetical subject. While some scholars may argue that accurate prediction of the past is irrelevant or impossible, many historians and social scientists regularly make inferences about undocumented history. Educated guesses occur across the discipline and beyond. There is nevertheless a tension between the widespread but often implicit employment of retrodiction, and the rarity with which it is deliberately considered and applied. Thinking more explicitly about retrodiction offers the chance to assess and refine historical interpretation. By understanding why some predictions about the past prove more or less accurate than others, we can think more clearly about how to manage empirical gaps, better adjudicate between competing interpretations, and gain greater confidence in our own retrodictions. This article proceeds in three parts before concluding. First, we introduce the concept of retrodiction and then reflect on the rewards and risks involved in making predictions about the past. Second, we consider different ways of making and testing predictions about the past, giving special attention to the important role of retrodiction tournaments. Third, we draw on key insights from research in the social sciences to inform the practice of retrodiction and highlight future avenues of research. Historical claims are built on evidence. From archival research and oral histories to archaeological investigations and satellite imaging, we have assembled much of our knowledge about the past from available data. Yet there are many gaps in the historical record and the pool of data available to experts is often very shallow. Experts who study the past rely on imperfect or incomplete information. In developing a broader interpretation of history, they may therefore need to speculate or “predict” the sort of picture that would emerge if or when more evidence ever became available. This attempt to predict the past—to engage in retrodiction—is thus one of the basic techniques that historians and social scientists use when they construct an interpretation of the past. Efforts to predict the past are commonplace. Scholars often make assumptions or reach conclusions about the past in the absence of direct evidence. Many historians and social scientists therefore already make retrodictions, albeit sometimes unknowingly or implicitly.4 For example, scholars have debated the status of the United Kingdom’s considerable war debts to the United States government following World War I. At least one historian has concluded that the war debt issue was effectively laid to rest in 1940 as a new and highly personal relationship emerged between Winston Churchill and Franklin Roosevelt.5 In the absence of direct evidence, however, this conclusion represents an implicit prediction about what occurred in the past. Recent archival discoveries drawing on data from both sides of the Atlantic challenge this claim by showing that the United Kingdom’s war debts to the United States persisted as a source of diplomatic tension in the decades that followed and remain unpaid and outstanding.6 Some historians and social scientists advancing arguments about the past may seek to avoid making any explicit claims about gaps in the historical record and limit themselves to presenting the evidence they have found. Yet those scholars may have missed a valuable opportunity to test their historical judgement. Historians and historically minded scholars in the social sciences should instead accept that their interpretations of the past are often based on incomplete data and speculate about the sort of picture that would emerge if more evidence became available. They could then test their understanding of events by providing specific details about what we would expect to see if that missing evidence ever became public. Retrodiction does not alter the underlying logic of an historical argument but instead seeks only to make it clearer and more testable. Thinking more explicitly about retrodiction can therefore improve historical practice and our knowledge of the past. If historians and social scientists explain how they reach their conclusions, they and other scholars can better assess the quality of those arguments if new and important evidence ever comes to light. By making and testing enough examples, we may be able to determine how far the lessons of retrodiction can be generalized to other periods or fields of study. We accept that many scholars prefer to focus their attention on explanation rather than retrodiction. Analyzing what the available evidence says is a more familiar practice than speculating explicitly about what presently unavailable evidence might show.7 Yet explanation and retrodiction are cognate concepts.8 As an interpretation is developed—and historians and social scientists develop an interpretation by drawing on theories about how the world works—predictions are generated as a matter of course. “At any given point in an historical argument,” one of the authors of this article has previously explained, “the historian can ask, given what was said up to that point, whether it would be possible to predict how things would develop. This provides a useful test of the power of the argument: a strong interpretation should have a certain predictive force.”9 Retrodiction also provides the historian with a way to assess the validity of the interpretation taking shape in their mind. As historians or social scientists proceed with a project, they could ask themselves what might turn up when they read new material—whether material that has already been released, but which they have not seen yet, or material that is released later—if that tentative interpretation is correct. Those retrodictions about what new sources would show might be borne out, in which case the historian would have more confidence in the validity of that interpretation. Alternatively, the new sources might not be in line with what had been expected, in which case the historian might have to rethink their emerging interpretation, and maybe even question the assumptions or arguments on which it was based. We argue that it is important for scholars to consider and apply retrodiction consciously and thoughtfully, by explicitly stating what they would expect to see in as-yet-unexamined or unavailable sources. Historians and social scientists have advanced broader interpretations about what drove the collapse of the Soviet Union, for example, with a small number of scholars making educated guesses about the unknown past and what they would expect the archives to eventually reveal.10 The absence of evidence therefore serves as a way in which to test and potentially bolster a specific historical interpretation. Transforming explanations of past events into explicit, testable predictions enables historians and social scientists to shift from rationalizing events after they have occurred to formulating hypotheses in advance. Emphasizing a priori theorizing over post-hoc analysis can potentially lead to sharper, more robust historical narratives and help to strengthen the inferential power of interpretations or theories regarding causes of past events and historical trends.11 Retrodiction can also help scholars avoid hindsight bias—namely, the common tendency to incorrectly believe in retrospect that one correctly guessed how events would unfold.12 By making falsifiable claims, before new historical data comes to light, historians and social scientists can combat this tendency and increase accountability, thereby fostering scholarly rigor.13 Attempts to predict the past, however, do carry some risks. Retrodiction requires a degree of uncertainty about what has happened in the past. If we make educated guesses when the required evidence is unavailable to us, there is always the possibility that the missing data will never come to light. Assessments of historical judgement require the discovery of new, relevant evidence after a retrodiction has been made. Some predictions about the past, no matter how specific or well formulated, may be unverifiable. Crucial archival records have been destroyed in wars, for example, just as some individuals or organizations may have never written much down. Historians and social scientists should therefore try to focus their attention on areas where there is a reasonable possibility of testing their retrodictions. New data about the past can be anticipated in some cases. Some archives use “timed releases,” such as the National Archives in the United Kingdom, which embargoes many government documents for twenty years before releasing them to the public. We tend to know which government documents have been embargoed even if we do not know what they say. Not all government documents are released within twenty years, to be sure, but even in sensitive cases a release date is often agreed. Venturesome scholars can test their historical judgement by offering educated guesses about what these files contain and help us learn from their retrodictions by sharing how they reached these conclusions. Moreover, new data continues to emerge at a remarkable pace. Archives around the world regularly release new material. Previously closed archives have reopened, even if only temporarily, and archivists remain hard at work recovering lost data.14 Archival data represents one of many types of evidence that can emerge over time. Witnesses may eventually come forward, for example, just as leaked documents can shed light on the unknown past.15 Many exciting testing opportunities therefore exist for historians and social scientists. Some scholars may nevertheless the of making and testing predictions about the past. We three challenges in First, may argue that retrodiction is is and In few accurate predictions about the past are possible and, even if some the lessons of retrodiction would not from one case to This of to history, however, into We do not know if some or all historical are or the to which some possible insights are Some scholars may be able to make more accurate predictions about the past than others, which is what social scientists have in their of and testing predictions about the past would therefore us to determine whether retrodiction is and whether it can be many scholars already make predictions about the past, whether or these retrodictions to light, understanding how they and making to test them could help to improve historical Second, some may consider more explicit or at retrodiction as of to They might argue that scholars have already their and their own work in the light of new We do not that historians and social scientists sometimes work but that such are and often which any If we rely only on the of history, we can only speculate about how interpretations have We therefore learn about what retrodiction and how to improve historical judgement. In might claim that when new data does it already into the work of other historians or social scientists, who can then assess whether predictions about the past correct. We accept of this Some of the authors of this article have used previously unavailable archival material to challenge or refine interpretations of the are often and in Many predictions about the past remain Moreover, there is on where or why retrodictions Although it may be to about the guesses that the study of the past, we are much more to them if we or these Third, of retrodiction may be of making predictions about the past. We these but believe them to be Many scholars already make or implicit inferences about missing data. and testing retrodictions seeks only to make historical arguments clearer and more Some predictions about the past may be but this would be historians and social scientists regularly challenge the claims of their and from them is key to our historical judgement. We should not scholars for making but that a to make arguments that might turn to be is a key of the by which historical interpretations of this is to the and methodological challenges by the testing of historical We accept that not can be or and that can challenge historical the difficulties involved should not scholars from to make and evaluate when challenges have to the of prediction and testing in other for testing the rewards on not least to and in testing can also which can improve and in By scholars with competing interpretations of the past to on the of specific retrodictions and the for future archival discoveries can be used more effectively to scholarly The insights generated by testing on could therefore be The and testing of claims about the future have been a of understanding and the and of If we seek to whether accurate predictions about the past are and whether some scholars or might better than others, historians and social scientists also to make and test retrodictions in ways that the and interpretation of data. We introduce to retrodiction for scholars to First, historians and social scientists could a the of an article or an for example, providing specific and testable predictions about the past, offering of confidence in these predictions and how they could then to these retrodictions to learn more about the power and validity of the underlying historical The political scientists and research the of Soviet at the of the for example, provides a of specific predictions in of the release of more archival them and other scholars to assess their broader interpretation of events as more evidence Although this of is in that it does not test any retrodictions, it scholars by them to think carefully about their assumptions and more by future This also that making explicit and testable retrodictions need not a shift or to research Second, scholars could and a Historians or scientists to test specific hypotheses or advancing broader interpretations of would begin their research as drawing on available evidence and their own but before conducting key archival or research. At this point, they would a that their predictions about the past their confidence and this of was historians and social scientists would to the archives or to test their This would help other scholars assess the quality of a arguments by how well the retrodictions generated by those arguments up in the light of evidence that have important in the and social sciences by and considered thinking about the unknown also many of the with post-hoc Third, historians and social scientists can retrodictions by other scholars in work and then assess their historical judgement when missing data eventually comes to light. This is with to of and the to new evidence can and are with testing opportunities. an however, Even if retrodictions can be and they may be making them to evaluate and to with post-hoc it may also be to historians’ or confidence their predictions about the past. Moreover, and relevant also to the testing a number of retrodictions may help us to assess the work of a historian and their specific claims but broader about a or will be historians and social scientists can retrodiction tournaments. of what social scientists have about prediction from of in the and In these experts to predict the of a of to which by their Scholars have to on these in the following more about the and of expert predictions and their underlying Retrodiction would historians and social scientists to make predictions about data from the past which are unknown to them but are to the of the way of such an would be for the to a of released archival documents, for example, and then provide with to those would then be to use their historical to predict what the documents would could provide historians and social scientists with a small number of to help their retrodictions or the more on what they to The would record their and confidence assess their and learn more about what make some predictions about the past more than Retrodiction could or and or as special within or These events could be or The might claims by different historical In to other of making and testing retrodictions, are more in of and rely on they would to manage and the testing across Retrodiction would therefore make and more In they can a of participants, different types of social scientists, and and our understanding of to predict the past. In all we would historians and social scientists to think carefully and explicitly about the confidence with any retrodictions, whether in of their own claims or those of The predictions a with in or with many ways to be and only one to be that come with a degree of confidence can be more useful than predictions they valuable about the or of a specific explanation or and testing predictions in the discipline of but is more common in fields and political The lessons from decades of research in other the social and can be by historians and historically minded scholars to better and refine how they make and test their own predictions about the past. Scholars have that predicting the future is have that predictions are not of the accuracy of social predictions regarding how of and political would in the of the that expert no more accurate than those of a of We should not therefore of accuracy in the work of historians or social scientists. past of expert prediction point to that can the accuracy of The that a of experts and to predictions as new became In the of to improve In a have that making more accurate had relevant used and based predictions on some of these or would help historians or social scientists to improve the accuracy of their retrodictions an exciting of future research. from the social and sciences also the importance of our own in which provides an opportunity to that can improve historical such is which would making an prediction about the past and then on why it may be whether by competing assumptions or different thereby historians and social scientists to in their and to make historical from a different can increase about assumptions and of other interpretations, thereby to more accurate These of from historians often work or in small research there can be in Scholars have that many and then taking the tendency more accurate than any This may occur individuals often such that when enough are other out, but of the insights from decades of research and testing the accuracy of predictions is that what experts think far less than how they this point, social scientists often turn to between and more in different and to explain events on what they think are who use one or to and explain the Although it might be more to these as on a rather than a the serves a useful These types of experts to do better than some scientists have that this has more important than or political to predictive or are better at predicting the past and therefore represents a line of future research for historians and social scientists, those that and are to their and their more those that prefer interpretations or theories of the interpretations or theories can an by understanding key of the they can and many gaps in the historical record. Many scholars studying the past have work by The historian for example, was an that to the past. have instead that historical valuable way to about historical thus learn more about the judgement of be to and assess retrodictions in those on which key arguments may As one political has assess any one can ask, well does it explain the We that a good or interpretation should be able to explain both the and unknown past, the us to limit many of the with post-hoc In to the accuracy of key we should also of historians and social scientists from to in retrodiction which would us to about the power of their in a more The of decades of previously unavailable evidence should us to on many predictions about the past. on a line of thinking when on the of the after the of the Soviet archives in the that all the of the much was and of and but of what happened can be is about historians’ accuracy and that of the that that will have to be in sources and the of better or more data are they the of poor and incomplete about to predict the past is with social the of predictions of the is reasonable to believe that many predictions about the past will be when they are eventually better or more data. and testing retrodictions, in a valuable than research and we can instead this data to learn from predictions about the past. By studying many of retrodiction, whether or we can begin to learn more about whether certain or can improve historical judgement. been more in predicting the past than for example, or have certain of methodological more than we may be able to of prediction of the past and, in turn, useful insights for future that can help to improve the practice of retrodiction. Retrodiction many valuable opportunities for historians and social scientists. explicit claims about the unknown past and then testing them can help us to improve historical judgement by better understanding the to which prediction is possible and whether some or better than Retrodiction can also inform scholarly as we expect scholars who study the past to reflect on the and interpretation of evidence, we should also expect them to think carefully and clearly about how they will gaps in the historical record. and explicit claims about the unknown past can help to confidence in and arguments by and in their can the work of historians and social scientists but also The more predictions about the past that we can and the more we can learn about the and of retrodiction. We have and different ways to make and test predictions about the past. Scholars can to their retrodictions and to test them or and the research Alternatively, they can retrodictions in work and use new evidence to assess their Some scholars may instead prefer to retrodiction many and data from all these can help us to improve our understanding of retrodiction. The of this are Scholars in many other fields have in understanding and predictions about the These have in turn good practice and to improve scholarly judgement. and testing retrodictions will historians to improve their own historical judgement and our understanding of the past. is of at the of has research in the fields of history, and with in the and has written and the and The of and has one an historian by is a of political at is the of a number of and about a with on The of was by in is an of at the of research in historical and research to individuals in the has in in the of and the of of of and the at the of with at and the of and work the of political and and such as accountability, and work on and their to improve analysis and to on as well as is a in at has written and and with the on areas of and is The on the and and the on the between and the is of at work has in of the National of and of the is an of at the of a to and with work in some of the in the fields of and of the National of and is of at the of research and to study and the and the to challenge of of the of the of and of from the and also the the and such as and is of and and of the at the of of the The to and the and by

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.