An explanation that is successful, in any sense of the word, must increase the extent to which we understand what is being explained. Obviously not all explanations follow the same pattern, but there is a common thread that links failed explanations: they either explain by drawing on something that is equally in need of explaining or they are unable to satisfactorily answer the question “how do we know that is the case?” regarding the claims of the explanation. Often in fact a failed explanation will suffer from both these problems.
Before I describe some real failed explanations allow me to illustrate the principle described above with some contrived, and hence uncontroversial, examples. Suppose then we explained what an elephant was by saying “an elephant is anything that possesses the quality of elephant-ness”. Obviously this is true, by definition, but it is not an explanation. Elephant-ness is itself unexplained, and certainly cannot be explained itself as “the property that defines being an elephant”, since elephant is at this point still unexplained. In order to escape this problem some might describe the property of elephant-ness as a non-physical “universal” that adheres to elephants (let us set aside whether that is an explanation for a moment). This version now cannot satisfactorily answer the question “how do we know that elephants have elephant-ness?” Since it has been separated from the elephants themselves in order to explain it we no longer have any access to “elephant-ness”, and hence cannot know that such elephant-ness is actually had by elephants.
Let’s consider a second explanation of what an elephant is, in this case “an elephant is not a leopard”. This example is slightly more interesting, partly because the explanation can be read in two ways. One is as stating what elephants are not. Obviously that can’t be an explanation of what elephants are, it’s not even in the right form. If it is to be an explanation then it must be claiming that elephants are “not a leopard”. But that is false, many things are “not a leopard”, and not all of them are elephants. So at best this is an informative description of elephants, but it is not an explanation of what they are. Thus it is not even a failed explanation, but rather not an explanation at all.
How about “elephants are zebras that have been cursed by a witch, with curse #45”? Ridiculous, I know, but it illustrates, again, how to fail at explaining by using something equally in need of explanation, in this case magic. The reason I bring it up is because I think most people in situations such as this immediately skip to asking how we can know that elephants are cursed zebras. The problem with that line of inquiry is that by presuming that magic works as an explanation it allows the believers in magic to manufacture whatever they need. For example, they could invent the idea that looking like elephants do is one of the signs of a curse. This is not as circular as it seems, similar reasoning can be used in connection with eye color and genetics. We explain eye color in terms of genetics, but we can equally use eye color to guess at a person’s genes. Of course there are other ways to have knowledge about genetics, but who is to say that there aren’t other ways to have knowledge about curses? No, the correct response is to ask for an explanation of magic. Once that is given it can either be shown definitively to be false (say by making bad predictions) or to be something we can’t properly have knowledge about.
Now it may seem like that line of inquiry may allow us to invalidate every explanation, and I should say a few words about how we can escape the dilemma it seems to pose. The dilemma being, specifically, that every explanation seems ultimately like it is going to be some kind of description. And to answer how we can know that description holds will require another description (elephants are explained in terms of elephant DNA, but DNA must in turn be explained by molecular biology, and so on.) Admittedly every explanation is a description of events, most involving a number of theoretical entities which are theorized to interact in certain ways, and thus explain events (theoretical meaning, in this context: something posited as part of an explanation). But there are two ways to know a description actually holds, not just one. One way is by “looking”, the theoretical entities involved in the description are identified with some state of an entity(s) that is part of another description, such that we can know about the entities involved in the explanation through our existing knowledge. This is the case with DNA, where DNA is itself explained in terms of chemistry, and thus we can, using our knowledge of chemistry determine facts about DNA, to see if it is in fact behaving as our theory claims it does. But there is another way we can know that a description actually holds, by virtue of it making accurate predictions. This is how we know that the descriptions of fundamental physics are accurate, not because they are identified with something else so that we can validate that electrons and quarks are really doing what the theory says they are (because the only way to know what they are doing is by consulting how the theory itself says they interact with the macroscopic world), but because the theory correctly predicts what will happen using them. I am obviously interested in philosophical explanations. Since philosophy does not make testable predictions this means that every philosophical explanation will involve explaining what is not understood in terms of what is better understood.
Obviously this makes a lot of philosophical explanations bad explanations. But so much the worse for them. Here are five of them: god or magic invoked as an explanation, good and beauty described as primitive and non-natural properties, dualism about anything. God and magic obviously can’t be used as an explanation for anything (cause of the universe, near death experiences, etc) because they themselves are desperately in need of explanation. Note also that to describe god as good or magic as a fifth force is not to explain them, to explain them would be to say completely what they are, not just to give an incomplete list of their properties (similar to our situation with “elephants are not leopards”, giving one property isn’t enough). Describing good and beauty as non-natural primitive properties runs into the same problem as describing what an elephant is in terms of its “elephant-ness”. Giving rules saying where goodness and beauty are found doesn’t help, because how do we know that goodness and beauty actually obey these rules? Finally dualism about anything (which technically the preceding explanations of goodness and beauty are an example of) can’t be a satisfactory explanation because to say that something is non-physical is not to say what it actually is. Take dualism about the mind. If it isn’t physical then what is the mind? Mental-stuff? Clearly that doesn’t explain anything about what the mind is, it is equivalent to saying “the mind is mental”. Now if we are discussing a dualist theory with content, that gives rules describing how this non-physical stuff is supposed to work (such as often are given for beauty and goodness, rules saying what is beautiful and good), then those rules may be a partial explanation, all that is needed is a way of verifying that the mind, or whatever, actually obeys those rules. And of course to do that we need to make it material (because obviously the explanation itself can’t itself justify how we know of such things, say through direct experience, because that would be circular, a theory can’t justify itself). What these claims all have in common, really, is that they substitute one kind of ignorance for another. Obviously we never completely remove our ignorance through any explanation, but we expect it to at least diminish to some extent. But all these explanations do is simply put a label on our ignorance, which makes them simply linguistic illusions, which only give the appearance of explaining.