We expect that causal laws will be the same across all experience. Hume famously claims that this expectation is grounded neither in pure reason nor in experience. Not pure reason: for one can posit a cause and deny the effect without being contradictory. And not in experience: for all experience can ever show is what we have observed in the past, and that information does not by itself tell us how to generalize upon it. We could generalize that causal laws will remain uniform; or we could generalize that the universe will go completely wonky from this date forward. Neither inference follows validly from what we have observed, and so they are in this sense equally nonstarters. Past performance is no guarantee of future results, as the saying goes.
Hume tries to find a way to explain why it is that, despite all that, we end up expecting causal laws to be constant. Strange as it sounds, the explanation he advances is itself causal. We become used to the causal patterns of the world, or conditioned by them through repeated associations, and so we come to subjectively expect causal patterns to continue. (This isn’t as paradoxical as it sounds. The salient fact about us, that we make causal generalizations, is also itself a generalization, and we expect to continue to generalize in the future as we have in the past. We are conditioned to expect continued conditioning.) We might well call Hume’s explanation the “Pavlovian” account of causality. It is meant precisely not to show that causal claims are grounded in any respectable, defensible process. It is only meant to explain the psychology behind our causal expectations.
Lord Kames, countryman and kinsman of David Hume, did not think this psychological account was good enough, and he raised a counterexample to the claim that constant connections breed causal associations:
In a garrison, the soldiers constantly turn out at a certain beat of the drum. The gates of the town are opened and shut regularly, as the clock points at a certain hour. These facts are observed by a child, grow up with him, and turn habitual during a long life. In this instance, there is a constant connection betwixt objects, which is attended with a similar connection in the imagination: yet the person above supposed, if not a changeling, never imagined, the beat of the drum to be the cause of the motion of the soldiers; nor the pointing of the clock to a certain hour, to be the cause of the opening or shutting of the gates. He perceives the cause of these operations to be very different; and is not led into any mistake by the above circumstances, however closely connected. (Kames 1751)
The child ends up smarter than his experience would suggest. How is he able to sort out the correlations from the causations? In reply to Kames, Hume could claim that the child is able to make the distinction because – once or twice – he has perhaps witnessed the drums beating without the troops mustering, or the gates opening or shutting at odd hours. And what if he hasn’t? Still, he might be able to see the events as only correlated because he has explored the barracks, the drum, the clock, and the gates, and he has found no mechanical links among them. This matters, because he has become otherwise accustomed to expect there to be spatially proximate, mechanical links between causes and effects, at least in events of this kind (“this kind” being correlations among bodies’ behaviors that are not alleged to be explicable through magnetism or gravity or (for us today) quantum spookiness). Indeed, in the Treatise, Hume insists that when we take ourselves to find a causal connection between events, we observe that the events “are contiguous in time and place, and that the object we call cause precedes the other we call effect” (1.3.14). The boy, perhaps, has found the correlated events to be spatially isolated – no links bridging them – and let’s throw in for good measure that perhaps he has also observed that the temporal relations are not as constant as one would otherwise expect among events that are really causally related.
But Kames, I expect, would have further complaints. Don’t we occasionally experience what sure seem like failures in mechanical explanation? We set up a perfect Rube-Goldberg contraption, push the first domino, and then what we believed must surely ensue does not. Indeed, don’t we encounter such causal disappointments just as frequently as we encounter correlated events that we are not supposed to think of as causal? The common course of life certainly suggests so. But if this is so, how on Hume’s account could we ever come to reliably sort out one kind from another? Why aren’t we far more confused than we are?
The upshot of this line of objection is that we end up knowing more about the world than we would if our knowledge were just a result of passive observation. Somehow, out of our experience, including our language and culture and education, we are able to form inner models of the world. In those models there are representations of what kinds of events are causally linked and which are not. Models can be mistaken, of course, and we can get causal explanations very wrong. But these models are not made automatically upon successive viewings of the passing show. Experience does not carve a model into our mind in the way a stream of water carves a canyon into rock. A model is an act of creative invention on our part, and it contains much more information than experience itself provides.
(Both Kant and Popper recognized this, by the way. But while Kant held that some components of the model are fixed, imparted to the model by the structure of the human mind, Popper regarded everything as negotiable.)
I wonder, though, why Hume was so attracted to such a simplistic view of our understanding. It may be that he could not see a way to contribute anything more complicated to the mind without bringing on the worry that he was making the mind supernatural. Nature as he knew it could produce an organism that is rudely shaped by experience in the way he describes. But how can nature produce a model-creating mechanism? Today we don’t worry about that question – not as much as we should, I think – but perhaps in Hume’s day the ability to create complex inner models that went beyond the elements of sensory experience had to be seen as something supernatural. Before you know it, there would be talk of souls, and Hume did not want to see talk drifting in that direction. Better an overly simple mechanism that nature can produce than a fancy one nature can’t, if what you’re trying to do is build a broadly nature-bound epistemology. Then you can hope that custom, habit, and culture will fill in any missing structure.
Or maybe I’m wrong to think that individual minds generate models, and Hume is right to look to larger cultural entities and traditions as the generators of models. When Hume claims that custom or habit is what leads us to expect causal regularities, he might be saying that our expectations – our models – are results of training and education and not results of individuals’ abilities. A humean Adam, with no one around to teach him, would have no expectations for the future. It takes a society for there to be individuals with some kind of shared model of the world that goes beyond each individual’s own experience. That’s an interesting idea.
I’m not convinced that Kames’ critique stands up to scrutiny. I think he unintentionally set up a straw man here. What if we never saw anyone at the gate, but the gate always opened at six am? One might reasonably suppose that the same mechanism that drives the clock is also connected mechanically to the system that opens the gate. I think that correlation can still be the answer here, but its the closest viewable level of correlation that matters. Lets take the oft quoted aphorism that any future technology, sufficiently advanced, will appear to be magic. If a man from the future arrives and commands the sun to stop shinning and the sun becomes black, people may believe that he is magic (or a god) and that the sun responds to his command. It could be that his voice command relays a signal to a giant fan that opens up in space and blocks the sun out, but the place where we attribute cause and effect is the closest place we can scrutinize, which in this case is the verbal command.
Lets take this idea of the closest level of scrutiny and apply it back to the gates opening. A person who knows of biochemistry and physics has different available levels of scrutiny. She or he might say that the opening of the gate is caused by a chain reaction of the electromagnetism force that causes the gate to open when a wheel turns when a rope winds when a lever is pulled, and that this happens when the concatenation of brain chemistry and signals fires in the brain of the person that pulls the switch.
Similarly, we could zoom out, rather than in. If we saw a conversation exchanged between the leader of the garrison and the gate operator in which the commander threatened a court martial if the operator did not open the gate precisely at six, our closer available level of scrutiny would help use see a different point at which we’d want to apply cause and effect, and its still one based on observation and correlation. In this way, two people could see the same event, the opening of the gate, and based on observation and correlations in the past, rather than an internal model, they would assign different causes.
Drawing this all back together, my main belabored point is that I don’t see a need to turn to an internal model and that observations and correlations could be enough of a reason for people to want to assign cause and effect. Whether someone never sees someone pull a lever, someone sees the operator pull the lever, or someone sees the operator talking to his commanding officer, the closest available level of scrutiny will be sufficient for attribution of cause/effect. Kames’ critique falls apart because it is simply too cursory an examination. Rather, it relies on all of us having a closer available level of scrutiny because we’ve all seen, these other interactions and they are necessarily a part of our history of observations and therefor inseparable from our reasoning. That Kames’ argument makes use of these observations while acting as though we can simply forget about them is why I say he unintentionally set up a straw man.
I think the real test of Kames’ critique would be to develop a novel situation in which none of us had any previous experience (read: observations) beforehand and in which we had to try to apply cause and effect. My guess is that many of us would observe a correlation and that’s as much as we’d have to go on and therein would lie our attribution.
Good to hear from you, Mabey! Your examples point out that what seem to be correlations are in fact causal, when you take into account all the various mechanisms involved. That’s an interesting point. It leads me to wonder why we don’t regard the clock’s pointing at a certain hour as the cause of the gates opening or closing (or do we?). I guess it’s because there are so many intermediary steps, each of which could fail under very slightly different circumstances. The causal chain is just too long and tenuous.
I’m tempted to try to come up with two constantly correlated events that don’t have any such long causal links – say, a deaf and blind goat randomly bleats when the gates open/close, with no causal chain linking the events. But if that happened reliably and regularly, we’d probably conclude there was in fact some mysterious causal link we couldn’t explain (magic, as you say).