Science, for many philosophers and scientists alike, should be firmly committed to using experiments to demonstrate various cause and effect relations. Consider this proposition: “These boys were watching a violent movie and then they went out and acting violently towards others.”
Here we have a correlation: some event(s) happens, call this A, and then some other event(s), call this B, happens as well. But is there a causal connection between A and B? The fact that there is a correlation – A happens and then B does – doesn’t necessarily mean there is a causal relation: we might just have a coincidence. Thus “correlation doesn’t mean causation” is a slogan we often hear in the sciences. Typically scientists devise experiments to demonstrate whether or not some correlation, between violent movies and violent actions for example, is actually a causal relation. Until such a causal relation is established with empirical evidence there is no scientific knowledge.
So we see the importance of causality for understanding scientific knowledge. Indeed, causation might be the most general and fundamental concept upon which science is based. And, since science is taken by many to be the most legitimate and effective form of knowledge acquisition we have, causality would be at the foundation of many institutions that rest upon such legitimate knowledge (like the justice and education systems). Therefore any philosophical inquiry which reveals our belief in causality to be unjustified would be very dangerous to say the least.
But this is exactly what the Scottish philosopher David Hume (1711-1776) tried to do in his works A Treatise of Human Nature and An Enquiry concerning Human Understanding. Let’s take a look at the basic structure of his arguments which show that, if we are empiricists who believe that all knowledge must come from sense experience, our belief in causality is not rationally justified and must be based on nothing but habits and expectations.
Hume’s Conception of Causality
For Hume, to say that A caused B is to be committed to three things:
(1) A is temporally prior to B, that is, A comes before B in time; (2) A has a contiguous relationship with B, that is, A is either directly in touch with B or is indirectly in touch with B through a chain causes; and (3) There is a necessary connection between A and B, that is, if A occurs then B must occur.
This is certainly a commonsensical account of causality. However, all these conditions can be challenged. For example, it appears we can deny (1) by pointing out that a car which pulls another car is the cause of the pulled car’s motion and yet both are moving at the same time. We can also deny (2) if we consider various experiments in physics (for example, the various Bell Test Experiments based on the Bell Theorem put forth in 1964 by John Stewart Bell) which show that subatomic particles can affect each at distance despite the fact that they don’t directly or indirectly touch. And (3) can be challenged if, as some argue, causal relationships are a matter of probability not necessity. I will return to the last point below. For now, let’s continue with the exposition.
Hume’s approach to the issue of causality is framed by his distinction, which became known as “Hume’s fork,” between two ways of acquiring truth: relations of ideas and matters of fact. Relations of ideas are those assertions that are obviously certain since we cannot deny them without contradicting ourselves. For example, the statements ‘a bachelor is an unmarried male’ and ‘a triangle has three angles that add up to 180 degrees” are obviously true and to deny them is to contradict oneself. One cannot say a bachelor is a married male with contradicting oneself; one cannot say that a triangle has 182 degrees without contradicting oneself. And we quickly see why: the claims are about nothing other than our ideas and how we define certain terms. In these cases we need not investigate the world around us using our five senses – we can just examine the the meaning of the ideas themselves and make a judgment regarding their truth.
But matters of fact are statements that we can deny without contradicting ourselves like ‘it is raining outside’ or ‘that person is 45 years old’. To deny that someone is 45 years old doesn’t immediately reveal a contradiction. Thus we can’t grasp the truth of these claims by a conceptual analysis of the ideas involved. Rather, we have to take into consideration the facts and verify them through sense experience. Thus we are required to gain empirical knowledge: we have to go and see if it is raining or find some document that would verify the man’s age.
Once Hume establishes this two-fold path to knowledge he goes on to show how causality cannot be justified through either relations of ideas or matters of facts. Let’s take relations of ideas first.
Causality Cannot be Rationally Justified Through an Examination of Our Ideas
Cause and effect connections can’t be known just by considering relations between our ideas because, unlike the impossibility of thinking of a triangle with two sides or a married bachelor, Hume thinks we can easily think of a fire not making something hot, a fragile glass being thrown with force onto pavement and not breaking, the moon suddenly disappearing, and a cat suddenly appearing. Naturally, all these examples strike us as implausible. But, Hume argues, there is nothing contradictory about them. Thus we are not dealing with relations of ideas. If we don’t learn about cause and effect relations from a consideration of our ideas alone, then, since there is only one other option for Hume, we must turn to matters of fact.
Causality Cannot be Rationally Justified Through an Examination of Matters of Fact
But cause and effect connections cannot be known through an investigation of facts given in sense experience either. Why? Hume’s answer is quite simple: no matter how many times we see event A happen and B follow from it, we will never see, hear, touch, taste, or smell a necessary connection between A and B. All we ever experience through our sense is that one thing follows after another or is correlated with another. To be sure, certain things, like heat from a flame or the sun “rising” everyday, consistently follow one another and this consistency allows us to lead the lives we lead. But this consistency is still nothing but a correlation since there is no way to establish a necessary connection through sense experience. So if, like Hume, one is an empiricist who thinks all knowledge must come from sense experience, then there is no way to justify the existence of cause and effect relations through matters of fact. And most scientists tend to be empiricists.
Induction and PUN
For those committed to objective cause and effect relations in the world this is troubling to say the least. One way to try and respond to Hume has been to invoke induction. To oversimplify, the method of induction requires us to gather data collected in sense experience and then cautiously infer a general law from this data. So, for example, if we see that water boils at 100 degrees Celsius over and over and over, then we will cautiously make a generalization: all water boils at 100 degrees Celsius. This generalization would then by a law that can be used to predict events in the future.
Now clearly induction, along with deduction and abduction, is a widely embraced form of rational justification. So can’t we be rationally justified in believing that two events are causally connected if they have always been causally connected in the past? Hume says no. If we can’t establish necessary connections through either relations of ideas or matters of fact then we have NO reason to believe the future will be the same as the past: the relationships between events can change at any time. This is the problem of induction: how can we know that the future will be like the past?
At this point one might appeal to PUN or the Principle of the Uniformity of Nature. This principle states that the laws of nature, such as gravity, electromagnetism, and the weak and strong nuclear force, are necessary, constant, and universally applicable in their respective domains of influence. So if PUN holds, then perhaps induction will work: the future will be like the past because the laws of nature are necessary.
But Hume insightfully sees that this move is a form of invalid circular reasoning. He asks: how did we come up with PUN? How do we know the laws of nature are necessary? We can easily imagine that the laws of nature could be different: indeed, science fiction has always shown us non-contradictory possible worlds in which certain laws of nature are different from our own. So the PUN did not come from relations of ideas, like ‘all bachelors are unmarried males’ or ‘a triangle has three angles that add up to 180 degrees’ which we cannot logically deny. What is left? According to Hume’s fork, just matters of fact: the PUN must be the outcome of empirical investigations. But this reveals that the PUN is itself the outcome of induction: we have empirically assessed that there are certain regularities in nature and have, after accumulating a great deal of data, cautiously come up with generalizations that we conclude are necessary. But if the PUN is the outcome of induction then it can’t possibly provide necessary laws: Hume has already shown us that no empirical investigation of nature can reveal necessary connections. So we end up with the following vicious circle: the method of induction needs necessary laws of nature (PUN) to provide a way to solve the problem of induction, namely, why the future will be like the past; but the so-called necessary laws of nature (PUN) are themselves the outcome of the method of induction that can’t give us any necessity.
Since this vicious circle is absurd, we see that induction and the laws of nature cannot help the empiricist escape Hume’s conclusion that causality, the most general and fundamental ground of science, is indeed rationally unjustified if we are empiricists.
It is important to note that Hume doesn’t argue there is no causality in the world. What he claims is that we cannot rationally justify our belief in it, that is, we must be causal skeptics or people who admit that we will never know if there are necessary causal relations in nature. However, he does provide a competing account of what causality for us. He argues that when we speak of cause and effect we are really speaking of psychological expectations that arise from custom and habit: we are in the custom of being burned by fire and seeing the sun rise every morning. The only necessity to be found in this form of causation is psychological necessity which is a feeling of inevitability. But this psychological necessity cannot be projected out into the objective order of things with any empirical justification. Thus Hume offers us a naturalistic account of how we come to believe that there are necessary connections based on how our minds get habituated to expect certain things after experiencing forms of constant conjunction.
Questions for Reflection
If we find this conclusion unacceptable, then how might we avoid it? What assumptions did Hume make that we might question? Could we define causality in a way different from how Hume did thus opening up new possibilities? We already saw some powerful counterexamples to all three aspects of his definition of a cause. Might these open up ways around Hume’s argument? One, that an effect can flow with probability from a cause, might do so. Perhaps causes, rather than necessitating their effects, can dispose or tend towards their effects. Stephen Mumford and Rani Lill Anjum, in their book Causation: A Very Short Introduction (Oxford: 2013), explain that there is another alternative besides, on the one hand, effects flowing necessarily from their causes and, on the other hand, having nothing but the pure contingency of constant conjunction that can’t provide knowledge at all:
“The third alternative, that Hume does not consider, is that causes tend or dispose towards their effects, where this is short of necessity but much more than pure contingency. We should not say that any possible outcome is just as likely. There is a definite tendency to one in particular. With tendencies and their possible interferers, we can see how over a large population there need not be constant conjunction between causes and effects. Smoking can indeed be a cause of cancer where, for the dispositionalist, this mean that it has a real causal power towards it. But this power is ‘only’ a tendency towards cancer, even though it is a pretty significant one.” (102-3)
So on a dispositional account of causality we shouldn’t even expect the pure regularity Hume assumed we should expect from causal relations. Indeed, if we find such necessary regularity perhaps we can be confident we don’t have causality! If we find regularity, say in how all salt is NaCl, then perhaps it is because we have two descriptions of the same thing. Or if we find that all whales are mammals then perhaps this regularity is a function of our classification system. As Mumford and Anjum put it, “Being a whale does not cause something to be a mammal. It is simply that the class of whales is subset of the class of mammals” (104).
This dispositional approach works well with naturalistic approaches to causality (but it can also be connected to accounts of formal causality so popular in the ancient/medieval world and very much alive today). But we might also consider a theistic addition, explored first by Descartes, which maintains that God is the all good, all powerful, and all knowing foundation which guarantees that our empirical inquiries can, if we inquire responsibly and thoroughly, get us to the truth more often than not. According to this approach the laws, regularities, and/or dispositional powers that exist are planned and maintained for the best and thus the problems that face an empiricist of the Humean sort might be avoided. In “Meditation One” of his Meditations Descartes observes:
“Nevertheless, in whatever way it be supposed that I reach the state in which I exist, whether by fate, or chance, or by an endless series of antecedents and consequents, or by any other means, it is clear (since to be deceived and to err is a certain defect ) that the probability of my being so imperfect as to be the constant victim of deception, will be increased exactly in proportion as the power possessed by the cause, to which they assign my origin, is lessened.”
That is, if we believe that we emerged from a process not directed by intelligence, goodness, power, and so on then the probability that we are deceived will be drastically increased; if we and the world emerged in accordance with God’s plan then the probability of our mental faculties getting to the truth will rise. Thus the way is open to a theist to claim that Hume’s issues with causality and induction are non issues given God’s providence.
In any case, if some or all of these approaches are plausible then one side of Hume’s fork, that side that says we can’t obtain knowledge of causality from empirical investigation, might be unconvincing.
And what about the other side of the fork? Hume argues we can imagine things appearing out of nowhere. This is supposed to show that an investigation of our ideas cannot show that causality is necessary since we can imagine scenarios where causality is suspended without contradiction. But is this really convincing? To be sure, we can, for example, imagine a ball suddenly popping into a room. But can we really conceive that scenario in a way that does not imply a cause? Can we really think about that and make rational sense of that? Wouldn’t we find that a full intellectual grasp of that situation would require us to think that, while we may not know where the ball came from, that we do implicitly believe it came from somewhere? Immanuel Kant argued that to have an experience at all is to already presuppose the existence of cause and effect relations. Kant, agreeing with Hume that causal relations must be necessary, agreed that we cannot obtain knowledge of them from any empirical investigation. But he thought we could nonetheless have knowledge of causality since (1) we have knowledge of something that did not come from sense experience, namely, the workings of our minds and how they forge necessary cause and effect relations that allow us to have experiences in the first place; and (2) that this knowledge we have, far from being a mere relation of ideas like ‘all bachelors are unmarried males’, tells us something about the world beyond our minds. Moreover, we should consider the fact that accepting Hume’s insight about uncaused events appears to violate the principle of sufficient reason which, in one formulation, states that everything that comes to be has a cause for its existence. Are we prepared to accept this implication? Shouldn’t the fact that we do not see random things popping in and out of existence lead us to an inference to the best explanation, namely, that our world is not a place where such things can happen? Approaches along these lines might show that the other side of Hume’s fork can be questioned as well.
In any case, these are just some of the questions and issues one might explore in dealing with Hume’s powerful challenge to those who would embrace both empiricism and the view that cause and effect relations need to be necessary.