The purpose of a theory is to explain some aspect of the world. Thus many kinds of explanations are rightly called theories, not just scientific ones. I contend that philosophy too, if it is to be of any value, must consist of theories, about subjects that are not captured by any science (here). And by improving our understanding of what makes a theory good we can thus improve our philosophy.
A good theory must have content, it must assert that things operate in one way and rule out other possibilities. Now just because the theory seems like it is saying something doesn’t mean it actually has content. The simplest, and best, way to determine if a theory has content is to see whether it can be refuted. If a theory can’t be refuted then it implies that it either does a very poor job of explaining, or that the objects of the theory are disconnected from the real world. And both of these faults imply that the theory only seems to explain, only seems to have content. So, even if we don’t plan on systematically testing the theory, whether it can, in principle, be falsified is one way to judge the quality of the explanation it provides.
A good example of a theory that isn’t able to be falsified, even in principle, is Freudian psychology. Freudian psychology fails to be falsifiable not because the theory is unclear but because the objects of the theory don’t seem to correspond to anything real. Freudian psychology posits an ego-superego-id structure to explain human behavior. How the ego superego and id interact is pretty clearly spelled out in the theory, so the problem is not that the theory doesn’t say anything definite about the theoretical objects it posits to explain phenomena. The problem lies in the way judgments about the ego superego and id are made. The psychologist observes the patient and then, on the basis of the patient’s behavior, creates a story about the interaction of their ego superego and id that seems to explain those observations. It would seem then that once the doctor has arrived at a hypothesis about the patient’s ego superego and id that Freudian psychology would then be testable, by comparing future behavior to the kind of behavior the theory says should be displayed by someone with that combination of ego superego and id. And this is where problems arise, because if the patient does act in unexpected ways it does not prompt the psychologist to question Freudian psychology. Instead they simply alter their story about the patient’s ego superego and id to fit this new behavior, and Freudian psychology can provide a story for any combination of behavior. The problem then is that the ego superego and id themselves don’t correspond to anything definite; there are no observations we can make that allow us to say something definite about the state of the patient’s ego superego and id, we can only infer facts about them by deduction from their behavior and psychological theory. But if future behavior doesn’t fit we are led to modify our deductions, not challenge the theory. And this is why I say that the ego superego and id are disconnected from the world, because the theory doesn’t allow us to make any definite observations about them.
Of course this is a bit of a problem for every psychological theory. Psychological theories in general posit some facts about the internal life of a person based on their behavior, and then predict future behavior on the basis of those internal facts. So we might think that, like Freudian psychology, they would be free to simply revise their guesses about internal facts in light of disagreements between predictions and actual behavior. This problem has not gone unrecognized by psychologists, and has prompted some rather extreme reactions, such as behaviorism, which attempted to deal with the problem by assuming people had no inner life, and that all of psychology could be handled by statistical predictions of behavior alone. But behaviorism isn’t the only possible solution. Another possibility is to make the revision of guesses about a person’s internal behavior harder. For example, our psychological theory might posit that if a person systematically prefers one kind of thing to various alternatives then they have a conscious or unconscious desire for it. And our theory has, in addition to this, rules that govern how desires can change over time or be suppressed by other desires. Furthermore let us suppose that we have observed a group of people, all of whom show a preference for chocolate ice cream. Thus our theory motivates us to say that they all desire chocolate ice cream. Now let us also suppose that one day a large percentage of this group suddenly started preferring a different flavor. Our theory says that if this happens then their desire must have changed or been suppressed, in one of an enumerated number of ways. And so we go through that list. One by one we rule out the possibilities, including chance, since too many of them changed their preferences at once. If we rule out all the possibilities then we are indeed forced to admit that our theory doesn’t satisfactorily explain this situation. Either the theory was wrong in assuming that they had a desire for chocolate ice cream, or the theory was wrong by omitting some way in which desires can change.
So to be falsifiable a theory must have two properties. It must provide a way to determine the status of its theoretical entities (whatever is posited by the theory to explain events that isn’t directly observable; egos, trade, liberties, ect) with a fair degree of certainty, and it must say how those theoretical entities interact and explain various events. Obviously good science meets both these criteria easily. Physics, for example, ties its theoretical entities to reality by holding that certain measurements on certain devices indicate their presence. And physics predicts how those entities interact very precisely, by describing them mathematically.
Now let’s apply this standard to philosophy. Let us first set aside the truly analytic philosophy, the work that primarily investigates what we mean by the use of a term. As I have argued (here) such work is best understood as a precursor to a philosophical theory that serves to clarify how that theory relates to our common sense understanding of the world. Consider a possible philosophical theory of ethics. Let us say that ethics is defined ostensively in the context of this theory as “that which we have reason to do regardless of the nature of our other desires” (the actual ostensive definition of ethics is probably more complicated than this). The ethical theory then will itself describe certain kinds of actions as ethical and unethical. And this theory is falsifiable, assuming that we have an independent understanding of when people have a reason to do something. To falsify it all we need to show is that in some circumstances the theory recommends some action that we do not have reason to do, modulo our other desires. Note that possibility of falsification does not consist in disputing whether the theory begins with a proper ostensive definition of ethics. Such a dispute cannot falsify the theory, all it can show is that the theory is really a description of something else, and not what we commonly call ethics. Another example of such a falsifiable philosophical theory is the description of society I provided a few days ago. In that description I outlined five functions that society must perform well if it is to fulfill its purpose. And thus the theory can be falsified by showing that society can best fulfill its purpose in some situations by ignoring or going against one of the functions I described. In contrast the philosophical theory that all the world is an illusion (either a dream or a simulation of a world) is not falsifiable, and hence a bad theory. (Although its opposite, the theory that the reality we know is not a dream or a simulation is falsifiable, and hence a decent philosophical theory.)
Of the two ways in which a theory may fail to be falsifiable I am concerned more with the possibility that our philosophical theories may not connect the theoretical entities they posit to the world in a sufficiently determinate way. Certainly some philosophy does consist in muddled and vague theories, but the people who like such theories certainly won’t be reading this. There is a real danger that some of our philosophical ideas, such as universals and intentionality and qualia, become disconnected from the world, such that we don’t have a way of telling when they are and aren’t present that is separate from our theories. Or, in other words, that we might fall into the Freudian trap; we have a theory that we think explains reality, but instead of checking the theory against reality to see if they agree we instead check reality and then arrange our theoretical entities so that they do agree, and so our theories can never be falsified.