Understanding Errors. The Role of Mistakes in History and Future

In this post, I discuss what we can infer about the relative epistemic merits of scenario-work from the failures to formulate correct scenarios of the future. I argue that these failures are not qualitatively different from failures in other disciplines. Rather, the difference is in the way that the epistemic commitments become apparent and can be tested.

“I believe that it would be worth trying to learn something about the world even if in trying to do so we should merely learn that we do not know much. This state of learned ignorance might be a help in many of our troubles. It might be well for all of us to remember that, while differing widely in the various little bits we know, in our infinite ignorance we are all equal.”

Karl Popper, Conjectures and Refutations.


A common accusation against the building of future scenarios is that those scenarios often go wrong. This is probably true, but we have to remember that

(i) Our choices affect the future and sometimes the fact that a scenario did not actualize can be a good sign. If we are able to avoid an undesirable scenario, we have made the right decisions.

(ii) The future depends on so many factors that it is natural that we sometimes fail to model it.

(iii) Failures come in different scales. It is one thing to fail to predict that in 2020 a pandemic takes hold, but another to fail to understand the likeliness of a pandemic at some point of the (near) future.

These responses are all well known. Whether or not we consider the creation of future scenarios as worthwhile practice depends on how strong these responses are, given the problem at hand. The debate over the value of professional future scenarios continues.

However, I suggest that we take a different perspective on this topic by asking two questions.

1. Are futures scenarios especially fallible in comparison to other scholarly results?

2. Is it possible to produce knowledge without commitment to some future scenarios?

Discussing these two issues will deepen our understanding of the epistemic underpinnings of estimating the possible futures.

I begin the discussion about the fallibility of future scenarios by comparing scenarios with historiographical accounts of the past. I think this is a natural contrast since (i) both fields attempt to describe temporally located singular events, processes and structures, and (ii) not many people really think that historiography is a futile practice. However, I do think that what is said below applies also to historical, social and human sciences in general.

It is obvious that historiographical accounts have often failed to describe the past. This does not only concern some silly theories or propaganda but also serious historiography. For example, Efraim Wallach (forthcoming, [the version I have does not contain page numbers)] describes how

”Through most of the twentieth century, two schools of thought offered different and conflicting narratives to account for an important part of the biblical texts: The books of Joshua and Judges, which narrate how the tribes of Israel came to be settled in the land of Canaan [–] the Conquest narrative and the Peaceful Immigration [–] narrative.”

Wallach then goes on and describes how the debate between the two narratives proceeded and how the Immigration narrative gained support. However, “[a]s the Immigration narrative was gaining support, a new one was articulated to challenge it. The Revolt narrative rejected the scenarios of both the Conquest and the Immigration narratives. [–] [According to it] oppressed population united to attack and destroy the cities of their former overlords.”

There was a problem in the Revolt narrative:

“It could not point to physical evidence in support of one of its central claims. [–] It is perhaps for this reason that the Revolt narrative did not gain much support in the disciplinary community or the public audience. But one crucial element of it did [–] The Autochthonic narrative [which says] ‘While the biblical story of the Exodus relates that the Israelites came from Egypt, many archaeologists believe that they actually originated [here] in the land. They lived in semi-nomadic groups in the hill country and eventually began to build small, permanent settlements….’ This brief passage summarizes the current opinion of most archaeologists, historians, and biblical scholar.”

Should we conduct a pessimistic meta-induction from abandoned historiographical accounts to the conclusion that the current accounts are probably false? I do not think so. Our approaches to the past have become more sophisticated – we know more about how to know the past – and therefore we are allowed to think that historiographical narratives have become more truthlike or accurate (at least in some cases). It seems that we are willing to accept failures from historiography and that failures do not make a scholarly field worthless. Moreover, historiography can be said to make progress even if there is disagreement about the details. As Wallach’s points in relation to the case-study, “Historians may disagree over whether the background of the people who lived in the hill settlements was pastoral [–] or agrarian [–] and on a few other things. But they share the narrative [–].”

However, the fact that we forgive errors and disagreement over details in historiography does not automatically mean that failures in scenario-work can be forgiven. One could argue that the fact that repeating failures of scenarios indicate that scenarios cannot be improved and therefore scenario-work is not intellectually sound practice (or, more provocatively, that it is a pseudo-science). I think this line of accusations should be taken seriously and we should diagnose the source of the failures.

I chose Wallach’s paper as an example for the reason that it describes in a clear manner how complex process a historiographical investigation is. It takes time to achieve a convincing account of some past process. For example, Wallach writes:

“Even though the (generalized) historiographic narrative is underdetermined by evidence, it is constrained by it. When the former changes, either by new discoveries or as the result of a novel analysis, the latter is constrained to changes by adding, modifying or omitting some elements while striving to preserve its core rationale.” [Emphasis added.]

“As time progressed and evidence accumulated, the competing narratives tended to be successively less streamlined, more complex, less “elegant” and in this sense, less comprehensible. [–] This tendency to progress from simple to more complex and detailed structure as relevant data accumulate is a common feature of scientific hypotheses in the historical sciences”.

We have to formulate many hypotheses, analyze their coherence and assumptions, and gather evidence (and discuss what counts as evidence).

This is important because a fundamental difference between the scenarios of the future and the narratives of the past is that we move away from the past and towards the future. We can improve our understanding about the past in the future while we cannot improve our future scenarios once the future is here. The future will come, and it will tell us whether the scenarios were correct. On the other hand, the correctness of a historiographical account can only be evaluated in comparison to other accounts. There will never be a moment when historians have to say “we knew nothing at all about the past process P”. Even in the case where the existing accounts of P turn out to be wrong, there will be some evidence E which tell that those accounts are all wrong. And given E, something can be known about the past and some alternatives excluded. There will never be a historiography-independent way of telling whether historiography is correct or not. However, the future will tell whether scenarios are correct independently of us.[1]

The differences in fallibility between historiography and scenario-work seem to be that

a) historiography has more time to improve its accounts of the past

b) the correctness of a historiographical account never faces a direct, historiography-independent test.

The point (a) should make us humble when comparing scenario-work and historiography. We are better off, epistemologically speaking, in historiography than in scenario-work. However, the point (b) reveals that it is difficult to tell the exact distance between accuracy of historiography and scenario-work. As we noticed, the past will never tell what really happened.

Point (a) also indicates that good scholarly work takes time. We have to formulate many hypotheses, analyze their coherence and assumptions, and gather evidence (and discuss what counts as evidence). In order to keep the epistemic process going, it has to involve critical virtues that are popularly known as self-correction, critical debate, willingness to falsify etc.

However, it is perhaps a bit ironic that virtues that are often associated with sciences and other academic fields require a future-oriented process that takes time and still we are skeptical about our ability to say something about the future. Why is this ironic?

First, in order to appreciate a field with the “scientific virtues”, one must assume that a future scenario where these virtues define research will lead to a preferable outcome is plausible. We do not mind errors in our search for knowledge because we think that, in the future, those errors will be corrected if we proceed in the adequate manner. For example, historians can offer many competing narratives of the past in order to advance their collective effort to capture the past. The hope is that, in the future, the merits of these narratives can be assessed. Mistakes are allowed and even considered as an essential part of a process that leads to an epistemically preferable future.

Secondly, the inability to tell the future and the fallibility of our knowledge are the two sides of the same coin. We cannot tell the future because our current knowledge is fallible. If the knowledge-base on which we build a scenario is erroneous, then the scenario is likely to go wrong. On the other hand, we cannot know whether our current beliefs are true because we do not know if there will be, in the future, evidence against them. This is the whole point of fallibility. The scientific virtues are considered as virtues because we know how difficult it is to gain knowledge. However, we are not willing to say that we know nothing. Rather, we would like to say that, because our present knowledge is a product of a virtuous process, we can trust it. If we are not willing to commit to our future scenarios on the basis of that knowledge, then it becomes unclear what the function of the scientific virtues is.

The second point generalizes. Knowledge should enable us to navigate in the world. Surely, there are two different senses in which knowledge (historical knowledge, for example) could enable us to navigate in the world. First, historical knowledge obviously makes historical records understandable and makes suggestions about what to expect from unexamined sources. We can call these internal expectations. In internal expectations, we expect to find something that does already exist. Secondly, knowledge makes it possible to expect things that do not exist yet. We can call these external expectations. These expectations rely on inferences on how things will develop from what we know about their current state. To illustrate the distinction, think of a doctor who thinks that a person has a tumor. She may expect that a biopsy will indicate certain changes in the tissue. This is an internal expectation. The doctor can also expect that without a treatment the cancer will spread. This is an external expectation.

One could argue that the difference between scenario-work and other disciplines like historiography is that the former is dedicated to external expectations and the latter to internal expectations. One could also argue that this indicates a difference in the humbleness of the disciplines. While other disciplines focus on investigating the accuracy of their accounts and on confirming internal expectations, scenario-work has the much more ambitious goal of formulating external expectations. It is this lack of epistemic humility, the argument continues, that makes scenario-work so vulnerable.

This argument is deceiving. Historians are future-oriented in their search for evidence and alternative accounts of the past, but their aim is to understand something that happened or existed in the past. Historians are not committed only on confirming internal expectations and making sense of the evidence[2]. Rather, a historical account of some historical H is committed to the claim that H happened in a certain way. Historians are, in essence, making a causal inference from evidence to the past. This is no less “external” inference than a scenario of the future. The difference between historiography and scenario-work is that, in the latter, the external commitments will come back to haunt the ones who made them. A historian will never face the events she describes. It is easy to appear humble when the commitments are blurred.

Moreover, whatever the main commitments of historiography are, all research requires auxiliary assumptions about the principles that govern the world. In previous posts, I have described how these principles and assumption govern historiographical counterfactuals that are necessary for historiographical explanations. However, they are also necessary when the past is merely described. First, we need to have all sorts of “everyday” causal understanding when describing the past. This is a trivial point. “Caesar died because he was stabbed” is based on such understanding. Often, we need to be more sophisticated. For example, once we learned that Darwin’s theory of evolution was not accepted on the basis of the Origin but later when he applied the framework in more detailed work, we are able to describe the scientific community of the time in more detail. We can specify its exact values (e.g. humbleness) and epistemic standards (e.g. novel discoveries). We are able to understand a historical event/process/structure by extracting its features from a causal series. In essence, this requires causal understanding about the world and the correctness of the historical description depends on the correctness of the causal understanding.

Once we notice that not only historiographical explanations (which some scholars think are clearly out of our epistemic reach) but also historiographical description depends on understanding causal processes, the difference between historiography and scenario-work seems to diminish. Surely, there are differences in the availability of evidence (as indicated here) between the two disciplines, but the commitment to causal patterns and structures is shared both in estimating the future and investigating the past.

I would like to go even further and ask the classical question “Why can’t we tell the future on the basis of the past?”. Surely, this is due the complexity of the world. However, the problem is not that we know the past but cannot infer the future from it. The problem is that we do not understand the world. Because we do not understand the world, we probably make wrong inferences about both the past and the future. Our failures to understand the past are less obvious than our failures to estimate the future. Yet, this is hardly a reason to give up estimating the future or explaining the past. There is no safe haven from errors and mistakes. We can be safe from errors only if we do not even try to understand the world.

Perhaps there are such safe havens. Some philosophers have suggested that historiography is not a representation of the past (see Pataut 2009) and it has been suggested that the truth of a scenario does not determine its value but its utility in thinking about the future (Staley 2006). I doubt that we can make sense of the concept utility without reference to the truth. Surely, some scenario-work W1 can serve as a good background for further work W2, but if each step the chain W1-W2-…-Wn is always evaluated only in terms of its utility with respect to next steps, we have an infinite regress. At some point, the result must be something more substantial than a useful way of thinking in further scenario-work. We have to assume that, at some point, the chain will involve an accurate description of future possibilities. I think the same is true of historiography. Even if we can understand historiography as constructing views of the past on the basis of current evidence when we focus on some local part of that practice, the practice as a whole (as a process) does not make sense if it is not assumed that the results are connected to the past.

Moreover, the problems in our knowledge of the past, the future and their connection should not lead to the denial of the value of the ideal of achieving such knowledge. It is quite astonishing that some people think that, because we fail so often, the humble thing to do is to define the goodness of a scholarly result with respect to our knowledge. Turning away from the world is a rather egoistic thing to do. It would be quite sad if academic research on societies and human life did not commit to any external expectations; if it only studied how it can describe, model and explain what happens now and what happened in the past. Not only would this be impossible (see above) but also frustrating. Every explanation and every account of the world has implicit consequences for the future because they commit to certain causal patterns. It is cowardly to remain silent about these implicit futures. The implicit futures might be all wrong but so what? We can never know the future if we do not take all the potentially relevant considerations into account.

Summing up, I suggest that there is nothing embarrassing in the failures to estimate the future. That we can know we were wrong is a good thing. It makes us to rethink our knowledge and inferences. It is much more difficult to get caught with systematic errors when studying the past or the present, as most disciplines do. However, it is not obvious that avoiding errors is the right thing to do. Or even a possibility.


Pataut, Fabrice (2009). “Anti-realism about the Past”. In Tucker (ed.) A Companion to the Philosophy of History and Historiography.

Staley, David, J. 2006. History and Future.

Wallach, Efraim (forthcoming). Historiographic narratives and empirical evidence: a case study. Synthese.

[1] Of course, there are some case where we have to make inferences about the present situation in order to evaluate whether some previous scenario about it was correct. For example, if there was a scenario which said that sporting events will have strange outcomes this summer due to Covid19, we have to estimate (a) whether Man City losing to Lyon or O’Sullivan winning the Snooker WC is a strange outcome and (b) whether these were affected by Covid19.

[2] It has to be admitted that some philosophers argue that historiography does not explain the past but the current evidence (see below). I think this is a mistake.

2 thoughts on “Understanding Errors. The Role of Mistakes in History and Future

Leave a Reply

Your email address will not be published. Required fields are marked *