Reason and causation

Greg Detre

Monday, February 12, 2001

Lucy Allais, History of Philosophy V

Hume found his own sceptical claims to be distasteful, yet could find no means to refute them. Likewise, we feel sure that somehow every philosopher in the last 250 years has missed a crucial and obvious rebuttal to Hume�s attack on the notion of causation � but what? Essentially, Hume is asking whether there is anything more to A�s causing B than B�s regularly following A, and concludes that there cannot be any such thing that we can know, thus relegating our causal laws to the status of laws of regularity, or �constant conjunction�.

In order to understand the importance of causation is in Hume�s system, it is necessary to first examine his theory of mind. Hume outlines a bipartite picture of knowledge based on his identification of seven different kinds of philosophical relations: resemblance; identity; relations of time and place; proportion in quantity or number; degrees in any quality; contrariety; and causation. Resemblance, proportion, degrees and contrariety are all relations of ideas, which reason employs in a priori knowledge. Identity, relations of time and place (or �contiguity�) and causation are derived from experience-based impressions from the senses of contingent (i.e. non-demonstrative) matters of fact.

Knowledge of matters of fact can be seen to be based almost entirely on causation, and so on an assumption of the uniformity of nature. Whenever we explain, predict or attempt to understand events, it is on the basis that the universe is and will continue to operate as it has, though our knowlede of it might be limited, and that this operation is best understood in terms of causal laws. This seems undeniable; Hume can point to any number of instances where we explain events in causal terms, in terms of our expectation that one type of thing will give rise to another unfailingly.

A causal law is actually far more complicated a concept than it seems. At the fundamental level, causation comes down to �if event C, then event E� (and also, �if E, then C must have been�). We might then develop this simple principle into a more general law complete with caveats and exceptions. For instance, we know that a free-falling object will not slow or change direction unless contacted by another object. Given enough observations, we can flesh this knowledge out further into a mathematical model, ultimately without any exceptions, and we can see science as humanity�s most rigorous attempt at systematising these regularities.

But what, then, allows us to make the initial �if C then E� causal inference? Hume points to three criteria by which we can compare the causal with the non-causal: priority, contiguity and constant conjunction. By �priority�, he means that the cause precedes the effect, rather than being co-temporal or afterwards. This would seem to be a fair assumption, though past scientific theories of the ether, action at a distance and gravity, and now quantum mechanics, have given philosophers reason to pause over this. Contiguity simply refers to nearness in time and space, which usually holds for causal events. Hume allows for events which are temporally or spatially distant, by assuming that there is a chain of unobserved intermediary events which are contiguous to themselves. Indeed, Mackie is exceedinly cautious in his discussion of both these points, and praises Hume for hesitating to grant contiguity and priority the same importance in our notion of causation as he gives necessary connection, since he thinks that their inseparability from causation is debatable.

This brings us to his final aspect of causation, constant conjunction. By this, he means that every time the cause is observed, the effect has been observed to follow, or that every time the effect has been observed, the cause had also been present, with no exceptions (otherwise, we employ the principle of induction to look for a larger-scale regularity that would incorporate all exceptions, such as eclipses within Newtonian mechanics).

Hume is moving towards an account of causation which derives its force from the psychological, rather than the objective or physical. It is �custom�, or �habit�, that leads us to expect that B will follow an observation of A. This custom, this inductive faculty, is clearly present in many animals and babies, so it cannot be a rational faculty. Moreover, we cannot say a priori that there is any necessity that B should follow A � our certainty grows with every occurrence, but it never reaches absolute certainty. These truths are not demonstrative, since their denial does not lead us to contradiction. If the proposition that the sun will not rise tomorrow were to prove true, it would be a great surprise, but it would not be contradictory in the same way that saying �2 + 2 = 5� must be a contradiction.

Yet constant conjunction alone seems insufficient to us as a description of causation. This is where Hume�s theory of the mind plays a role. He describes ideas as faint, remembered or imagined versions of impressions, the original sense data. Ideas are given force by custom, so that our beliefs are simply stronger versions of propositions that we do not believe. The sense of causation that we see in and ascribe to the world around us derives from this expectation engendered by custom, based on constant conjunction.

If Hume rejects constant conjunction as well as insufficient to justify our idea of causation a priori then what, in his view, would constitute such a justification of causation? Causation, as distinct from regularity, depends on necessary conjunction. By this, he means that event A and event B have always occurred together when observed, and moreover, that there is something characteristic and defining about A such that it must give rise to B. The link between them is necessary, rather than merely contingently constant.

This statement is clear enough in its intent, but we need to try and be clear about what it might really mean. Talking in terms of �necessary connection� is certainly making an ontological statement, about causation as it exists in the objects � there is something about the object or event that its effect is determined. But what determines that the world should work in a causal way, i.e. what is it about the world that means that the future should necessarily follow a path based on the properties on the objects? And more particularly, what determines that the world should work in the particular causal way that it does? Hume is unhelpful in response to these questions, understandably, because he doesn�t believe that we have any knowledge of necessary connection and so we have no means of conceiving what it would be like.

What if we were to have an omni-sensory observer, somewhat like a human being but with eyes and ears everywhere, existing for the duration of the universe? If we can imagine that this being could exist from the start to the finish of time, then we can largely ignore the problem of induction, since there are no unobserved events. One might argue that there are events that would have happened if things had been a little different at some earlier point but did not (�counterfactuals�), and these unobservable unoccurring events might have differed from our expectations. We will examine later whether this concept of events that haven�t or don�t happen can influence the problem.

In order to accept a being that lasts for the duration of the universe, we have to make certain basic assumptions about �time� being single, linear and finite, in the way that it seems to us. A Kantian might take issue with this, since it implies an objective temporal property in what they would regard as the world �in itself�, but it should be possible to re-cast our thought experiment to please them. If such a being equipped with completely wide-ranging versions of human senses were to exist, would it be in a better position to know of causation than a human? At the very least, it would have a complete empirical record from which it could certainly deduce mathematical models about the world in the way that science does. In fact, in a non-counterfactual, i.e. deterministic, world, then any such mathematical models that fit all the data would be scientific laws. Such a being would be able to codify this regular world into laws that have operated without exception on events. They would be able to reduce the informational content in the world by describing it, in the same way that the formulae for a fractal or the Fibonacci sequence are reductions of the original data. Eventually, the world (especially as it seemed to Hume�s 18th century Newtonian understanding) might be describable in just a few mathematical equations and rules. Such a being would have completely described constant conjunction and achieved perfect scientific knowledge. Surely this is an understanding of causation?

Hume would reiterate that this has in fact failed to show evidence of causation, i.e. that the events that occurred must necessarily have been so. We may be able to describe how they are, but without knowing why they must be so, we are no closer to seeing them as necessarily so conjoined. We still have no reason to believe on the basis of the regularity we see in our perceptions of the world that the world really operates on causal laws. What then would it take to elevate these laws of regularity to the status of causal laws?

The existence of God might be able to satisfy us. In his capacity as First Mover and Groundsman of the Universe, he sets out the lines, limits and laws within which all of the universe operates, and thereby defines necessity himself. God would be enough then, but only if we accept that his ability to make causation intrinsically necessary to the world goes beyond any understanding of the world we have been able to frame so far. But if our only chance of escape from scepticism lies in our faith in an omnipotent God not to deceive us pointlessly with his veridical-seeming causal laws, then we are left in the same tenuous position as Descartes � such a God would solve all our problems if he did exist, but proving his existence without requiring him to sanction our proofs seems impossible, and even if he did exist he would need to explain why there is suffering, vagueness and illusion in our world.

But can we be convinced of the causality of the universe without such a God? Physics and a rich materialist metaphysics together attempt to provide the same sanction of necessity to the laws of nature that God would, for example on the basis of the weak anthropic principle. This principle effectively states that the universe is the way it is because if it wasn�t, we wouldn�t be here to observe it. If the fundamental constants of the universe had been any different, life and intelligence as we know it would never have arisen. Had the weight of the neutron not been marginally greater than the proton, there would be no consciousness. This principle is intended to explain how it is that the universe is, coincidentally-seeming, exactly as it needs to be for us to have come into being. (More recently, even stronger claims that the universe must necessarily be have been attempted using quantum mechanics.) But putting aside such ambitious attempts to use mathematics and theoretical science to explain the universe as a causa sui, it seems inescapable that any scientific (i.e. empirical) conception of the world is wholly dependent on the assumption of causation, and so can only provide circular justifications for it.

For even the weak anthropic principle seems to be relying on certain very basic laws about universes coming to exist at all, in certain ways, with predefined fundamental properties with varying constants. Hegel provides one of the few examples of philosophers who have believed that the entire world can be built up from a single foundational point or principle, even down to the seemingly contingent minutiae of our scientific world view. Unfortunately, Hegel notoriously proved the motion of the celestial bodies, and from it that there could be no planet between Mars and Jupiter � ignorant of the discovery of just such a planet a few months previously. We have already discussed and dismissed Descartes� theistic attempt along similar lines above.

At this point, we need to observe Galen Strawson�s distinction between ontological and epistemological claims about the lack of causation. He asks whether Hume is saying that there really is no such thing as causation in the world (ontological claim), or whether he is rather saying that we can never know whether the regularity is causal (epistemological claim)? Strawson argues quite convincingly that Hume�s intelligent standpoint in every other respect is to highlight our ignorance about how things are or must be, rather than to make ontological claims, even negative ones. On Strawson�s epistemological interpretation of Hume, he would be happy to admit that there may be scientific laws that are somehow self-defining and necessarily must exist and be so, in which case we could live in a universe without a Christian God but with fundamental incontrovertible a priori laws of nature � but we cannot ever be sure either through reason or experience that there is causation in the world.

In order to understand how he rounds off this claim that the idea of a necessary connection in nature cannot derive from either the senses or reason, we need to further analyse where our notion of causation does come from. Hume claims that it derives wholly from Custom, that our idea of an expectation of an effect is strengthened in proportion to the number of observed instances without exception of the events being conjoined, and that this process accounts for our notion of causation. Custom has habitually attuned us to expect that A will follow B, and we have elevated constant conjunction to a certainty that should really be attainable only with necessity.

But are there any other more plausible accounts of where we get our idea of causation from? Collingwood in his �Essay on metaphysics� considers that our notion of causation might be derived from our own experience. I consider that there are two ways in which our status as agents acting in the first person could contribute to causation.

Firstly and most obviously, our interaction with the physical world has given rise to our conception of ourselves as agents with free will, i.e. causal power that we can exert on the physical world. This view is so deep-rooted that it underpins our systems of morality and legality, and critically undermines mind-body dualism and various kinds of determinism. Knowledge about how we interact with the physical world can clearly only be gained empirically, through this interaction. In this sense then, it cannot be a priori. However, on the other hand, our awareness of how it feels to exert our will towards an end, such as picking up a pencil, is not derived empirically. I have specifically chosen the will, because even our awareness of internal bodily senses, such as hunger or nausea, are received empirically in the same way that the five �external� senses are. In the case of the will, we might imagine a person devoid of all five senses, who could still will certain acts on the basis of desires. Even at this point, Hume would be able to defend himself, since the causal relationship that we derive from our will is based on the effects of our will on the world, which can only be apprehended through the senses. In the case of a man without external senses, he would never know that moving his arm resulted in a smashed plate, because he has no sense of touch, hearing or vision. However, if we allow him internal senses, he might feel nauseous, will himself to vomit, and feel release in his stomach, and he could derive a cause-effect relationship from this. Hume can just as easily extend his defence to include all senses, external and internal (i.e. bodily), since they all provide varieties of experience, and so cannot be demonstrative. So, Hume is still safe in denying that any forms of experience can be the origin of our notion of causation.

However, there is a second means by which we might have gained our notion of causation as a result of our first person position. As rational agents, we are wedded to the idea that our mental world is coherent to some extent. If we take the case of our man without any senses, internal or external[1], we might imagine that his own rational process might be sufficient to give rise to the notion of necessary connection. We can certainly see constant conjunction in the simple habit of counting the natural numbers, where we come to expect the next number in our head. But for necessary connection, we need to turn to what Hume terms the �relations of ideas�. Can we not see in mathematics and logic a kind of atemporal causation. For instance, when faced with the syllogism:

������ All men are mortal.

������ Socrates is a man.

\ Socrates is mortal.

Our mind steps through the propositions in sequence, with the conclusion necessarily resulting from the premises. Can we not see the same sort of progression here as in �cause and effect�. In a simpler case, of �2 + 2 = 4�, don�t we find that �2 + 2� and �4� have always �occurred together when observed, and moreover, there is something characteristic about A such that it must give rise to B� (Here, even the Kantian would at least agree that the process is a priori). As for contiguity and priority, we might attempt to argue that these are in some way present, in �2� and �4� being near each other, and that our premises always precede our conclusion, but I think this might be stretching the argument too far. Rather, we can simply refer back to our earlier agreement that they are less critical.

Such an explanation would turn Hume�s discussion on its head, since it would allow an a priori origin for the notion of causation. However, certain rather glaring problems with this account should be noted. It is highly debatable whether such a wholly insensible human being could ever attain rationality in the way described. After all, without senses of any kind, we are effectively deprived of the notion of an external world, except as that from which our thoughts bubble up. More seriously still, we have to question whether our thought experiment is ill-posed in conceiving an insensible man who is able to reason about his own ratiocination[2], that is, perform second-order reasoning. This would almost certainly seem to require some sort of linguistic approach, which the man would have to have devised entirely himself inside himself. At this point Wittgenstein�s private language argument probably throws up more problems than we can possibly address, especially given the extremely high level of rationality being required of our insensible man in order for him to derive the notion of causation from his own ratiocination. Wittgenstein argues that, tempting though it may be to try and learn about the nature of (our) psychological phenomena by introspection (turning our attention inwards), it is fruitless. Introspection is not a means to a definition of sensations, without which such a private language cannot exist. Without any public behaviour to attach linguistic behaviour to, our insensible man would remain inchoate.

So if we are to deny our insensible man the language he needs for his high-order speculations, can we salvage anything from this idea that ratiocination is the basis of our notion of causation? Perhaps, if we assume that he is as intelligent as we are, but simply in a non-linguistic way. Could he not have non-consciously abstracted from the structured, sequential, necessary nature of his ratiocination, in the same way that Custom works on experience without us rationally bidding it. Is it any different to talk of Custom as applied internally to reason rather than to our experience? We cannot afford to doubt that our reason will continue to be structured and necessarily correct in the future, as we can doubt experience in the future, since to doubt reason is admit complete scepticism. If we allow the Evil Demon the power to subvert our reason at every level, the philosophical enterprise is utterly doomed. If we are prepared to rely on rigorous reason, as we must, then the problem of induction can be safely ignored when talking about Custom and reason. We trust that the nature of our rational process will still be structured, sequential and necessary in the future, else we can trust nothing. In this case, only a complete sceptic would doubt that our notion of causation could be derived from a priori necessary connections, viz ratiocination.

There is perhaps one last doubt that ought to be addressed. What about the so-called �anomalousness of the mental�, as Davidson describes what he sees as the lack of deterministic causal laws operating in the mind. Certainly we find it difficult to follow a single train of thought, and we really have no control over what thoughts will well up next. I think that this anomalousness can be re-characterised for our purposes in terms of volition � the difference between my seeing the room shaking during an earthquake and my flicking my eyes up and down rapidly is that the latter results from a willed action. Our thoughts are anomalous insofar as we don�t directly will them in advance to appear � the other way round, in fact. It is the same with experience: we don�t will the chair to be placed in the middle of the room � rather, the impression of the chair impinges upon us, yet we don�t describe the outside world as anomalous � quite the opposite, in fact. Though somewhat of a departure from the spirit of Hume�s theory of mind, the apparent anomalousness of the mental should be seen just like experience, as events in our minds that we can try and process. Rationality is the faculty of judging rationality in our thoughts, rather than producing rational thoughts themselves. If we allow ourselves �reason� in this sense, then we are allowing ourselves the faculty perceive indubitable order in the mental anomalousness, just as we can perceive order in the physical world.

At the very least then, we have convinced ourselves that there are cases of rationally necessary causation � and such cases, though mental, must be considered as �in the world�, since even thoughts must �exist� in some sense. We are making an ontological claim about necessary connection in the world, about events that are necessary in the only way we can understand necessity, that is, reason. Yet this does not really help us make the claim we want to make, about causation in the empirical world of objects and colours and tables. We still have no reason to believe that our claim about necessary connection between rational propositions extends to rationality in the world. One might say that it is a smaller claim to make, to say that on this basis it seems likely that the world will also be rational, but this would not be a leap of certainty.

Perhaps though, we should re-consider what we mean by our having a �causal role� in the world. Certainly, we can view ourselves and each other as objects, whose effects on the world can be regularised within the physical laws. But what we really mean when we talk about ourselves as causal �agents� is that although we chose to do one thing, we might just as well have done another. Our status as agents is bound up with free will. So, in the case of our all-seeing anthropomorphic observer, as long as there are humans in the world, there are counterfactuals, �what might have been�. In fact, even if there were no humans in the world, contemporary science would argue that the decay of a uranium atom, for example, is random, that is to say, stochastic, and cannot be predicted. In this way, we are very close to the bizarre notion of an �uncaused event�. In this case, the decay of a uranium atom might emit a particle at time t₁, or time t₂, with radically different consequences. And lastly, there is the firmly-established mathematics of chaos, where miniscule difference in the initial conditions snowball into large-scale alternatives.

So the notion of counter-factuals cannot be easily dispelled. This is important, since it forces us to reconsider how weak or strong the claim of regularity on its own is. The word �regularity� initially connotes a pattern that our minds project onto events, as though events just happen and we notice resemblances between them, and then learn to see resemblances and regularity all around us. And, in a wholly deterministic determined world, this might be all that happens. But in a world where where one possibility becomes reality from amongst a host of equally valid but discarded counterfactual alternatives, then there must be some means by which the universe decides what will happen in the future.

One of the worst case scenarios is set out in the New Riddle of Induction. It would not be inconsistent with Hume�s scepticism about induction and causation for all our scientific conclusions to be exposed as holding true until time t, at which point gravity will invert and the speed of light will halve. Less likely still, the universe might operate on a random or semi-random basis, and at any point the regularities we have so carefully codified over time will all be shown to be illusions, collapsed by some wholly inexplicable exception.

In a way, this is the fear that quantum mechanics engenders. Its low-level indeterminacy threatens to demote the scientific notion of causality to mere probability. We seem to be faced with a situation like the above, where atoms� states could be genuinely unpredictable and our macroscopic notion of causation would have to be expressed in terms of likelihood. If the universe were simply navigating its way simultaneously through a host of �many worlds�, each the reification of a quantum possibility, then the idea of necessary connection would be being undermined ontologically as well.

I think this fear will probably prove ungrounded. We have to consider where the indeterminacy lies in quantum mechanics: currently, there is a rift between our understanding of the classical and quantum levels of observation. Yet both are effectively deterministic, through relativity and Schr�ger�s equation respectively. The real indeterminacy lies in our measurement, and may prove to be only epistemological.

But given that we are so loth to abandon our faith in reason, that 2 + 2 = 4 will continue to be so in any counterfactual universe or under any future laws, we have the germ of an Archimedean point of certainty. On this basis, it seems much more plausible that there is a rhyme and reason by which the universe determines what state will result from its current state. Perhaps it operates according to some sort of algorithm, to which our physical laws are increasingly accurate approximations. Certainly, this optimism drives physicists. It is also possible, of course, that the universe obeys rules, but that we can never truly know them.

Hume might rejoin that we are missing the point of his scepticism, with our talk of causal algorithms and the like. Even if the universe were to obey some sort of algorithm for determining outcomes from its present state, that would simply be a regularity-generating algorithm. Without there being something about the state of the universe such that it must give rise to one and only one future, we cannot talk of necessary connection and causation. According to Hume, an algorithm of the universe that contingently specifies connections, even universally regular connections, is insufficient for our notion of causation.

But isn�t what Hume is actually asking for incoherent? He is requiring that there is some �power� inherent in objects such that their effect is an �infallible consequence�. But this only begs the question, �But how could a power be a necessary cause of its effect by virtue of itself alone?�, that is, how could some property be a cause, except within a system? Properties alone are meaningless. To talk of �mass� as a fundamental property of the universe says nothing until we talk about the rules which operate on mass. Hume cannot legitimately talk about some �power� inherent in an object without there being laws by which that power is governed. As discussed above, even God the First Mover would need to specify laws for there to be necessary connection � it is simply inconceivable to human reason that there could be properties in objects that give rise to effects without laws specifying why that should be so.

Reason remains the only case where we can see that something should necessarily be so, without being able to specify why. If we try and talk in terms of analyticity, say, it is clear that our rational capacity far exceeds semantic reduction alone.

What Hume is really requiring for necessary connection is far more difficult than first appears. In order for there to be this power that gives rise to necessary connection, there must be a set of rules that are themselves necessarily so, which determine that the power has this effect. But then we are simply moving the burden of necessity to the rules, since any rules that are not themselves necessarily so are simply rules of regularity. But what might generate and rule these causal rules if not the universe�s causal rules themselves? At this point, we seem to be trapped in an infinite regress, which either demonstrates that Hume�s requirement for necessary connection is incoherent, or so unfulfillable as to lead us with certainty to an ontological claim about a-causality in the world.

This entire essay, I have tried to show that faith in reason would seem to undermine Hume�s conclusions. But unfortunately, I think that where Hume does err, it is by placing too much faith in reason. If we are prepared to doubt the bedrock rational premise, that shared human reasoning is objectively verifiable, even amongst ourselves, then any attack on Hume crumbles. Indeed, philosophy crumbles. Because this would be admitting that the only standards for judging reasonableness are reason itself.

We can easily then doubt that reason is necessarily connected � after all, how else could we argue indefinitely over points which seem so clearly and distinctly so to one of us, yet incontrovertibly misconceived to the other? There appears to be a systematic 2-way relationship between our mental states and our brain states, and if, for instance, we are caused by a knock on the head to conceive a prejudice for a particular notion, then our reason may be powerlessly and invisibly subverted. But this is an extreme case � I think reason can be attacked in its every little operation. It has proved extremely powerful within a closed, highly specified system, like mathematics or formal logic. But natural language cannot provide such a system, and we cannot specify experience in any other, more formal way. There is no algorithmic means for us to conclude whether or not something is right, no final recourse � in logic, we can use truth tables or the laws of logic that we have chosen. In science, we can prove a theory by subjecting it to experimental results. But in philosophy, we can only argue with reason by appealing to reason. When our thoughts dance in every direction, there is no public way to agree whether or not they are rational beyond consensus.

This may seem a rather nihilistic view of things. But it needn�t be. Rather, it supports the Humean standpoint. Obviously, Hume�s own conclusions were set out rationally themselves, but if we cease to discuss the problem, it goes away. When we stop demanding deductive justification for induction, we have no reason to doubt induction. Similarly, when we stop assuming that reason is our supreme mode of comprehension, then a deductive understanding of causality is revealed as both hopeless and worthless.

---

[1] This thought experiment, though rather extreme, has some basis in physiology. Loss or lack of one or more of any of the external senses is common and well-documented, (blindness, deafness etc.). Loss of them all could conceivably be achievable with surgery or through some horrendous congenital condition or accident. As for the internal senses, a rare affliction known as �anasognosia� deprives the sufferer from any bodily feedback whatsoever. This leads to some extremely bizarre cases of anasognosics who refuse to recognise extreme bodily dysfunction, for instance. See Damasio, Descartes� error, 1994.

[2] I am using this term in the sense of �the action or process of reasoning� (New Shorter Oxford English Dictionary, 1997)