Hume’s Determinism Refuted
by David R. Schneider
Abstract: Hume’s argument for assuming cause-and-effect is undermined by way of three empirical challenges. Further, the argument itself is shown to contain two critical flaws in logic. A new theory is presented to supercede determinism, in which causality and chance are relative. This theory is intended to help minimize obstructions to theory development, which can be traced to a historical bias towards seeing more causation and order in events than is objectively justified.
1.1 Philosopher David Hume developed an ingenious and historically important argument regarding cause-and-effect. In his book, An Enquiry Concerning Human Understanding , Hume showed that our belief in the existence of cause-and-effect is a direct result of observation, experience and inference. Our belief rests on empirical evidence; yet this belief can never be substantiated by a formal proof. We may be able to discern and express natural laws, but we cannot be sure they will be valid in the future (for example). Since we will never be able to deduce the existence of cause-and-effect (i.e. that the laws of nature are causal), Hume argued that we must assume its existence as a postulate of any scientific theory.
Determinism is the theory that causes lead to effects by mediation (in time) of laws of nature. Theories which assume that they are describing causal laws of nature are called deterministic theories. Since the publication of his book, Hume’s determinism has had a profound and significant influence on the progress of Western science. Determinism has been very useful. No fully effective challenge has ever been mounted against Hume’s determinism. Where is the repudiation of the empirical evidence in favor of determinism, or a counter-example to the theory itself?
Some philosophers have already taken stands against determinism. But as surveyed by Jeans , these arguments have not been persuasive, often because of complications in defining indeterminism. As an example, indeterminism is sometimes considered to be synonymous with uncertainty, although this seems to be a weak definition. Uncertainty is lack of knowledge of determinate conditions, and so it still assumes that the laws of nature are determinate.
1.2 The objective of this paper is to show that a bias exists in favor of determinism. Such bias seems to have historical roots. At any rate, I believe that determinism obstructs theory development in many situations. For this to be the case, determinism cannot always be useful as an assumption.
Hume’s determinism will be fully refuted in this paper. Three empirical challenges are presented in section 2. In section 3, the argument for determinism is shown to be based on circular reasoning; further, it leads to a logical absurdity. In section 4, it is shown that determinism can be replaced by a theory of natural laws in which both causality and chance are present, and are relative between observers.
2. Empirical Challenges to Determinism
2.1 When quantum theory was introduced in the late 1920’s, a surprising discovery was made: the success of the theory is dependent on the existence of statistical, non-causal forces. The behavior of individual atomic particles is completely indeterministic. As a group, particles act as part of a “deterministic ensemble.” The bigger the group, the more determinism which is seen. But detail analysis has demonstrated that all fundamental particles display completely random behavior at the atomic level. They appear, in some sense, to have “free will” when observed individually; their behavior is not “caused” by anything.
The so-called “Copenhagen interpretation” described above has been the subject of much controversy, scrutiny, debate and attack. Einstein believed until his dying day that an underlying cause for the observed indeterministic behavior of quantum particles would one day be discovered. In a famous quote , he said “God does not play with dice.” However, the following situation exists today: 1) empirical evidence supports the concept that nature is fundamentally comprised of microscopic particles which obey indeterministic, statistical laws; 2) a theory has been presented which ties such statistical laws to large-scale (macroscopic) patterns which otherwise appear causal; and 3) modern quantum theory is the most accurate of all scientific theories.
Based on the above, it is concluded that an empirical counter-example has been presented to the theory of cause-and-effect. A single such unanswered counter-example is sufficient to refute any theory.
2.2 The second challenge attacks the empirical evidence in favor of cause-and-effect itself. This is the backbone of Hume’s argument for determinism. Define a cause to be X; define an effect to be Y; and let L be the hypothesized causal law which relates X and Y. X must also precede Y in time. If we observe that X always leads to Y, then we normally infer L. X causes Y if and only if L is a causal law.
However, the following technical exception must be acknowledged. We have never actually observed two instances of X which led to two instances of Y. Every X - that is, every set of initial conditions - is different in some respect from every other X. Similarly, every Y is somewhat different from every other Y. Therefore, there is actually no empirical evidence for the existence of some causal law L which acts the same way (in all particulars) at all times.
It may be objected that L may be inferred when each X is approximately similar, since each Y will likewise be approximately similar. However, this situation would always be true by definition regardless of whether or not a causal law L actually existed. To the extent X’s and Y’s are defined by the experimenter as similar, they will automatically tend to support the experimenter’s causal “law” L. By the line of reasoning of this objection, empirical evidence is unnecessary to support the theory of cause-and-effect. Thus stated, we have a new definition of determinism: the similarity of X’s and Y’s is called cause-and-effect.
The quandary may be restated as follows: all empirical evidence can be considered as either supporting determinism, or as a counter-example to determinism, at the complete discretion of each individual observer. Unless you and I agree as to the boundary or definition of what we have observed, our conclusion as to the objective existence of empirical evidence supporting determinism is merely relative.
A specific example may help demonstrate this point. If I measure the force of the earth’s gravity on an object, I see the relation F=gm. However, magnets and gases exhibit a different relation with respect to the earth. Is gravity deterministic? F=gm contains no term regarding the make-up of the object, so the only way to explain the exception is to say that gravity obeys the relation F=gm unless something else causes it not to. We have bounded our causal law so that our hypothesis stands, but we now must postulate the existence of new phenomena to describe the other “causal” forces which are at work; and so on. How can this program be falsified? If it cannot be falsified (since we can always postulate additional “unknown” causes), isn’t the conclusion questionable?
2.3 The third challenge to determinism is not dependent on either of the prior two challenges. Strictly speaking, determinism postulates that there is, at least in principle, the possibility that an independent cause can be shown to lead to an independent effect. That is, there may be some attribute X(1) which always leads to some attribute Y(1) within a particular cause-and-effect sequence. In technical terms, this is known as the problem of identifying the independent variable. This has been shown to be an important criterion in developing psychological and sociological theories, for example. Specifically, the independent variable must be identified to insure that the hypothesized correlation is not “spurious.”
The challenge: where is the independent variable? No truly independent variable has ever, in the entire history of all science, been identified. If determinism is all it is said to be, why is identification of an independent variable so difficult - shouldn’t it be easy?
A review of the problem shows that we cannot, in principle, ever observe an independent variable. For it to be identified unambiguously as being independent, such variable can have no causal connection to other observables. (If there is any causal connection to another variable, then the cause cannot be narrowed to the hypothetical independent variable.) If it has no causal connection to other observables, then it cannot be observed! For all intents and purposes, it would not be part of the observable universe. If nothing else, the independent variable must be connected to time (so as to identify X as being antecedent to Y). But if it is connected to time, then it is also causally connected to space, and mass, and energy, etc. Therefore the attribute so measured would no longer be truly independent. Q.E.D.
2.4 In summary: 1) Evidence exists which contradicts determinism. 2) There are no pure empirical examples of cause-and-effect. 3) Determinism calls for the possible existence of at least one independent variable, yet none has ever been found. Based on these three points, it appears that the empirical argument for determinism is untenable. Of course, new information could be discovered in the future which would allow us to reconsider one or more of the above three challenges. However, such information must first be discovered.
3. Errors in Logic Exposed
3.1 Hume’s argument contains two errors in logic: 1) a tautology (circular reasoning; you assume that which you sought to deduce); and 2) a logical absurdity (an argument which gives rise to mutually contradictory deductions is a logical absurdity, and is automatically considered to be in error).
Hume’s determinism rests on the existence of empirical evidence in favor of cause-and-effect. Such evidence consists of those situations which tend to support theories of cause-and-effect. Situations which do not support the concept of cause-and-effect are deemed, by definition, to be different sets of circumstances (i.e. due to the fact that the causes are not similar). If the evidence for cause-and-effect is bounded in such a way as to include only supporting examples, and cannot by definition include non-confirming examples, then we can see that the evidence is no evidence at all. It is merely circular logic; definitional only. (This is discussed in section 2.2 also.)
3.2 Hume’s argument can be turned around and applied equally well to the theory that the laws of nature are indeterministic. We look out on our world, and note that chance events are commonplace. Numerous events transpire which are outside of our causal control, and are completely indeterministic by any objective standard. (If you question their status as indeterministic, then simply present a causal law which predicts such events. If you have assumed they could be explained by “undiscovered” causal laws, then you are not being objective.)
If we are injured while out in public (drunk driver, robber, etc.), then we cannot be held responsible since we did not “cause” the event. It was only a statistical possibility - a risk. (Note: it cannot be said that we were responsible by being out in the first place. Accidents can occur anywhere, and we have to be somewhere.) Reductionism won’t help restore determinism either, since we can decompose chance events into smaller chance events as easily as into deterministic ones (by our choice of definitions and subjective standards).
In addition, numerous laws of nature exist which appear to be tied to probabilities rather than cause-and-effect. The weather appears to be associated with temperature and humidity. Yet these are only measures of macroscopic phenomena, and do not “cause” rain. Human behavior is correlated to numerous phenomena which cannot possibly be reduced to anything remotely resembling causes or effects. What caused you to arrive at work this morning at a specific time? Nothing did; yet your arrival time may be statistically correlated to “office hours” or your first appointment. Doesn’t chance explain and describe at least as well as determinism?
Thus, there is certainly ample evidence for indeterminism. Events occur which are outside our causal control, and events occur which can best be described in terms of probabilities. Yet indeterminism can never be proved, since the statistical laws of nature could become causal at some point in the future. Thus, by Hume’s line of reasoning, we must assume indeterminism.
But Hume also showed that the laws of nature must be assumed to be deterministic (see section 1.1). Therefore, unless the laws of nature are both absolutely causal and absolutely statistical at the same time, a logical absurdity is demonstrated. Determinism describes our universe, and so does indeterminism. The order you see corresponds to the order you seek.
3.3 It may be objected that the human mind derives its success from deterministic inferences; therefore determinism occupies some sort of preferred status. However, recently discovered evidence by Nisbett and Ross  provides conclusive proof that this is not so. The conclusions of the human mind are frequently based on statistical, rather than causal, connections. They call this the “availability heuristic.” If we see any correlation between similar X’s and Y’s (from our own definitions of similarity, of course), then a causal pattern is inferred. This statistical pattern is believed just as strongly as any causal pattern, even when it is demonstrated that the pattern cannot possibly be causal. We tend to see causal connections where there are none.
Consider the relationship between the height of a father and the height of his son. Most people believe that there is a causal connection between the two - “like father, like son” - if the father and son are both tall, for example. The father’s height is considered to “cause” the height of the son. But any biologist can confirm that this cannot possibly be so - it is physically impossible. The son’s height can, at best, be causally connected to the grandfather’s height. But the father’s height can, at best, be causally connected to his grandfather’s height (the son’s great-grandfather). There is absolutely no causal connection of a genetic nature between the height of a father and his son. But there is a very good statistical correlation.
It is the availability heuristic which leads us to see a connection, and we mistakenly identify it (by means of “rational” thought) as a causal relationship. There is a relationship, but it is one of chance. This same heuristic is responsible for many people’s strong belief in paranormal phenomena, even when analysis shows that their belief is based on erroneous identification of causal mechanisms.
A further review leads to an interesting speculation regarding the source of the deterministic bias mentioned in section 1.2 above. It seems that the availability heuristic is a subconscious tool. Rationality seems to be a conscious tool. Therefore, it is natural that we would be more aware of rational, deterministic relations and less aware of statistical, non-causal relations. Each could be present and accounted for in human behavior, while it is the causal influences which we see most clearly (because we are conscious of them).
4. The Relativity of Causality and Chance
4.1 From section 3 above, it becomes clear that different observers can always come to different conclusions regarding the existence of causal laws of nature. Such conclusions rest on the individual’s own experience, biases, etc. Individuals will not necessarily identify the same patterns, nor will they necessarily postulate the same causal laws to explain observed patterns. Further, where is the line to be drawn as to similarity of causes, and similarity of effects? These conclusions must necessarily be subjective.
Are natural laws deterministic or indeterministic? If they depend on the observer’s reference frame, then they must be relative - they cannot, by any method other than simple agreement, be said to be absolutely one or the other. This leads us to the theory of the relativity of causality and chance: causal relativity.
Specifically, it is postulated that we can never know - completely, finally and unambiguously - whether or not the laws of nature are causal, statistical or some unknown mixture of the two. We are stuck with simultaneously dealing with both determinism and indeterminism. Yet it is still possible to draw conclusions about the patterns and relationships we observe, as long as an absolute causal reference frame is not postulated . We do this every day.
It is important to note that within the theory of the relativity of causality and chance, no assumption is made regarding whether or the laws of nature are actually deterministic, indeterministic or some unknown mixture. All that is said is that they cannot be observed to be one or the other.
A determinate observer will see some indeterminism in an object system under study. An indeterminate observer will see determinism in the object system. Thus, when the laws of nature are considered for both the observer and the object system together, causal relativity will be seen. This is held to be true whether the system is physical or social.
4.2 In conclusion, we are now shown that our universe is neither deterministic nor indeterministic in any sense which we can hope to prove by way of empirical argument. This may disturb some because chance is elevated to a position not possible in purely deterministic theories. But the facts draw us to this conclusion, despite our personal philosophies.
I choose to view the situation in a positive light. After all, this certainly gives us a clever way out of the problem of free will, which is inherent in deterministic theories. How can we have true (rather than apparent) free will if our actions are predetermined? If we can never know if they are predetermined (due to causal relativity), then the problem is solved.
Hume’s argument for determinism cannot be maintained unless entire new classes of events are discovered. Despite its contributions in the past, determinism may be unwarranted in many disciplines. We must prepare for a re-evaluation of our direction of study with respect to our understanding of the laws of nature.
Deterministic bias, leftover from the time of Hume, may lead us to see more causation and order than is really present in our world. Perrow  sees this as a major problem within the social sciences. Such bias may be negatively affecting theory development by tending to place unnecessary restrictions on theories (by requiring them to be deterministic). If it is impossible to observe absolutely causal laws of nature, then we could never hope to substantiate deterministic theories by way of experiment.
Determinism may have been a useful assumption in the past; it may be a useful tool in simple situations in the future. But at the present, it should not be used as a barrier to theory creation in realms in which deterministic theories appear to have reached a stopping point.
1. D. Hume, An Enquiry Concerning Human Understanding, (1748).
2. J. Jeans, Physics and Philosophy, (1942).
3. A.P. French, Einstein - A Centenary Volume, (1979).
4. R. Nisbett and L. Ross, Human Inference: Strategies and Shortcomings in Social Judgment, (1980).
5. D.R. Schneider, “The Relativity Of Causality and Chance”, (1987). Unpublished; causal relativity is applied to physical theory.
6. C. Perrow, “Disintegrating Social Sciences” (1982). Phi Deltan Kappan, Vol. 63, 684-688.