Bone
Professor X
Philosophy of Science
March 2022

INTRODUCTION

The idea that explanations can be thought of as arguments has held much appeal in the philosophy of science. There are many refutations which seem to disprove the notion, but today I will be focusing on the asymmetry of explanations and arguments. It’s characterized most plainly, I think, as the notion that explanations are temporally constrained. This means that (in some views) an explanation, specifically a scientific explanation, only operates moving from antecedent conditions to subsequent events. In regular English, an explanation always points to the causes of an event. This criticism was most notably brought up by Wesley Salmon, who said,

“Although, by applying known laws, we can reliably retrodict any past occurrence on the basis of facts subsequent to the event, our intuitions rebel at the idea that we can explain events in terms of subsequent conditions. Thus, although our inferences to future events qualify as explanations according to the inferential conception, our inferences to the past do not."

Firstly, our intuitions going against this particular idea of explanation doesn’t necessarily mean that the idea is not valid. If common sense were the end-all be-all I find it hard to believe we’d be doing philosophy in the first place. This is surely not Salmon making an argument from intuition, but I’d like to point out that this is no reason to avoid going deeper. If there is something to be found, we should look for it, if not, a mental exercise.

A commonly used example which shows the perplexing nature of the idea is that of the lunar eclipse. If you were to use the position of the planets yesterday in conjunction with the law of gravity, you could explain why a lunar eclipse happened today. This is a very simple example, and one can see how this would work as an argument and explanation. Statement of antecedent conditions and known laws explain the subsequent conditions (the eclipse). I would like to note that the use of laws here is not essential, and I won’t be arguing about whether laws have validity in the philosophy of science, for now they are a logical tool to use. The idea behind Salmon’s argument here is that if you were to use the eclipse and the law of gravity to then retrodict the position of the sun and moon yesterday, it is a valid argument but not an explanation. This makes sense at first, it seems entirely illogical to say events in the past can be explained by things they have caused. The underlying reason for this perceived inconsistency is that explanation is deeply tied to causation. At least in Salmon’s view, when we explain things, we are trying to tell each other the causes of the events we observe.

THE LINK TO CAUSATION

The reason I take issue with this argument against explanations as arguments is that rearranging the argument such that past events are explained (rather than present or future events) does nothing to disrupt the causal nature of the events as they took place. By deriving past conditions which led to current events, you are not then saying the current events caused those in the past. On the contrary, you only find previous conditions because of the causal connection. To use an example that might make more sense, take a detective working on a case. The detective, hard at work, assembles clues and eventually figures out the identity of the perpetrator. Coworkers ask the detective “Why were there footprints leading away from the scene?” The detective explains that the suspect fled the scene after committing the crime, leaving prints. Uncontroversial, right? Now they ask “Why did they commit the crime?” Without a clear confession from the perpetrator, the detective has no answer to this question. Clearly though the detective has knowledge that works backwards temporally, they solved the crime (antecedent event) by working from subsequent evidence after all. Perhaps the better question which illuminates this would be “How do you know they committed the crime?”

To continue discussing this case, I will refer to the question of “Why did they commit the crime?” as Q1 and the question of “How do you know they committed the crime?” as Q2. In Q1 they ask for the general causes of the crime. By general causes I mean what we consider the most important events just prior to the crime in the causal chain. General causes could be socioeconomic conditions, motivations, etc. The details aren’t too important, the main point is that general causes are what we consider having impacted the occurrence of the event before the event happened and are often considered the subjects of our explanations. In Q2 they ask a more specific question, one framed through the lens of knowledge. The asker is entreating the detective to explain to them how they came to reconstruct past events in their mind using only present information. This asks the detective to illuminate the events of the past using subsequent conditions and background knowledge. This is a form of explanation as well. These two requests for explanation are markedly different in what they ask, though, and it would be an error to gloss over that when referring to them both as explanations. The most easily identifiable difference is the presence of the human actor in the question itself. Q1 asks for a story which has no detective, a retelling of the story arising purely from its antecedent conditions, and so on. Excluding direct involvement with the crime, the person asking Q1 does not care if the detective exists or not.

Q2 asks for a story of discovery. It is an epistemic question, which is valuable for many reasons, but may not seem to be the most important when we generically ask for an explanation of an event. It is often assumed that the general causes of an event are the real thing we are trying to get to, Wesley Salmon makes this case when he says causation and explanation are inextricably linked. I would generally agree that the causes of an event are what we’re after in science, using them to understand physical phenomena. What is often overlooked though, is the fact that we must answer these epistemic questions before we can build an idea of what the causes are.

To highlight this, let’s return to the detective with some more detail. Imagine the investigator is a religious man in the 1800s. He is a Christian and values the truth wherever he can seek it out. When asked about the causes of the crime perhaps he could provide us a story which sees a young person influenced by evil forces to commit said crime. One could assume this is an incorrect (at least unjustified) explanation. It could check all our boxes for being an argument and intuitively make sense to whoever is listening, but it would be incorrect. When asked how he located the perpetrator though, he would give a similar retelling of events that a modern detective would, which end in the crime solved. Something very interesting happens here. The 19th century detective would work backwards to arrive at facts which are causally related to the crime. For instance, seeing a weapon at the scene indicates an important fact, the criminal brought a weapon. It sounds obvious, the criminal bringing a weapon is in fact one of the causes of the crime being committed in the way it was.

This is an essential aspect of what I’ll refer to as epistemic explanation. An explanation of this variety is meant to show the listener how the knowledge of an event was arrived at, which will show certain causes of said event. The causes revealed will of course be dependent on the techniques used to investigate the issue. Let us revisit the example of the eclipse. Using mathematical techniques can show us the previous positions of the sun and moon, but may not show us other important aspects, such as the atmosphere of the moon, or what the moon is made of (besides it being opaque). This highlights the limitations of the epistemic explanation, but it also helps to assert the link that still exists between the epistemic explanation and causation.

The extent to which we can epistemically explain an event is limited. It can illuminate causes and facts which bear effect on the phenomena we measure but cannot (by itself) show us the totality of general causes of an event. In many cases, it may not even be able to show us all relevant general causes of an event. Because our knowledge of events works in this way, testing things, changing variables, and looking backward, arriving at information requires epistemic explanation. Even if one never opens their mouth to tell another how they discovered something, we would assume that they could if asked. If they couldn’t, the validity of their newly found knowledge seems tenuous at best.

Regarding the interplay between cause and explanation, I want to make clear that I’m not holding that events actually cause their antecedent conditions, though that would be an interesting position to defend. I merely point out that the event can explain its antecedent conditions, and I do this to show that the link between causation and explanation is not so unbreakable, at least not in that a valid explanation must have causes as anterior conditions of the argument.

WHY REFRAME THE TOPIC

The reasons for reframing the issue in this way and acknowledging explanations as arguments which can work backwards have been laid out. Now I intend to expand on the concept of epistemic explanation. First, I want to show how an epistemic explanation is more than just the answering of a how-question. In a sense, they align in many ways, one can provide an epistemic explanation in response to a how-question, specifically the question “How do you know?” I would also hold that you can provide the epistemic explanation in response to the standard why-question. In the example earlier, I highlighted the epistemic explanation using the how-question asked to the detective because it makes the application much easier to see. This isn’t the only way it’s applied though, and it can even be seen answering the basic why-question of “Why did this event occur”.

Consider a situation with less clear causes. A scientist and her friend observe something neither of them have seen or heard of before. They see a mammal breathing underwater. They see, discuss for a bit, but since this is entirely new, the scientist can’t answer her friend’s question when she asks “Why can that mammal breathe underwater?” The scientist takes some time doing tests, using her power of reasoning, and developing a theory, but does not find a set of causes sufficient to explain in the standard way. However, when they meet back up, the scientist is not at a loss for words when the question comes up again. Instead of answering in the form F happened because Y happened before it, she answers in the form I have discovered facts A B and C about F, which point toward XYZ possible causes. When framed this way, the epistemic explanation I lay out seems more like an unfinished investigation trying to find a suitable ‘real’ explanation. I would hold that this is just the same as somebody’s answer to the how do you know question, merely with different prompts.

The reason this concept is important for consideration outside of appearing to be merely an incomplete answer lies in the purpose of scientific explanation itself. One could fully explain why a board is broken and has a shoeprint on it. Having witnessed it, if you were to say that the board was stepped on and it broke under the shoe-wearers weight, you would be correct in that explanation. If you were to look at the board and assert that given the evidence and what you know of boards, there was likely a stepping that occurred, you would also be correct. Despite having not seen the event and not having any incontrovertible proof, you provide an explanation which is just as useful and correct as that which a witness could provide. This is not to say that this line of argument justifies scientific reasoning, it’s meant to illustrate the validity of explanations which do not have the privilege of working from antecedent conditions to subsequent events.

CONCLUSION

In conclusion, the idea of explanations as arguments seems to hold up to this particular critique. I believe in general that rejecting explanations as arguments on the grounds that they are temporally constrained is misguided. For the reasons stated above, epistemic explanations are tied to causation in an interesting fashion, even if not immediately visible, and the role they play in explanatory endeavors as a whole is very important. While there remains many arguments against the idea of explanations as arguments, if nothing else, I’ve provided an avenue for further discussion of the topic.