The possibility of universal causality

Abstract: The concept of ‘causality’ deciphered with the Universal Philosophical Method (UniPhiM). This root concept was swept away from ontology by Bertrand Russell, then revived by different models: counterfactuals, agentism, probabilism, transfer —with in particular Max Kistler’s solution in 2003, the transfer of a conserved quantity. I show how UniPhiM makes the ontological invisibility of causality coincide with the multitude of its teleological appearances. Science uses a pseudo-ontological language that does not access reality per se. The equations do not contain a principle of causality but answer only part of the ontological questions. I show how their acronyms can conceal the principle we are seeking.


To show the practical interest of the Universal Philosophical Method, let us submit to it the concept of ‘causality‘, very controversial today. It is a root concept, “everything has a cause”, but currently split between science, philosophy, and everyday pragmatism. Science is mainly concerned with ontological causality, with the “How?”. Philosophy is interested in epistemic causality, in the “Why?”. Finally, pragmatism mixes these two directions, when causality opportunely becomes a “useful” concept.

Can we unify these approaches to achieve a universal concept? This is what Max Kistler attempted in ‘Causality and the Laws of Nature’ (2003). A masterful investigation and a real reference. Book completed by a more concise and readable article for a layman in Intellectica in 2004. Causality, as a root concept, is ubiquitous and popular in both practical and intellectual life. If we try to seek its universality by a reductionist method, its development must then find the different facets and coordinate them.

The conservation of a quantity

Kistler’s solution is, at first, a reduction: He makes causality the transfer of a conserved quantity from cause to effect. The quantity conserved in physics is called ‘constant’ and corresponds to energy. Kistler adds a nomic1in connection with a law necessity: one fact is causally responsible for another fact if there is a law of nature by virtue of which the first determines the second. Kistler’s work is epistemic; he seeks to avoid contradictions between ontology and teleology. His solution appeals to physics, where energy is a well-identified constant, but how to apply it to the human sciences? How can we recover our causes and our daily responsibilities?

Kistler will confront his solution with criticism. The task seems insurmountable. Engaged in fundamentalism, how do we associate root concepts such as ‘constant’ and ‘law of nature’, foreign in essence? The former has a mathematical definition, the latter does not. Is the concept of ‘law’ really free from that of ‘causality’? The definition seems circular. But this is an almost inevitable pitfall for root concepts.

In a 1st part I summarize the problem facing Kistler and his proposals. In Part 2 I show how the Universal Philosophical Method (UniPhiM) simplifies the investigation and leads to a coherent general theory of causality. The main elements of the UniPhiM are summarized here. Recommended reading to understand the 2nd part. The stakes are up to the task. Once it is accepted that downward/epistemic and upward/ontological views are not reducible to each other, Kistler’s pruning of definitions of causality is no longer necessary. On the contrary, it is important to preserve their complementarity. We recover our “useful” causalities by keeping them compatible with the “fundamental”.


Part 1: Causality and the laws of nature

In 2003 Kistler exhumes the corpse of causality. This skin of grief is what remains of Russell’s passage at the beginning of the twentieth century. The replacement of causal terminology with the functionality of a model signals the maturity of a scientific discipline, Russell said. He invited philosophers to definitively dispense with the concept of causality, with three arguments.

Russell’s three Stabbing Strokes

1) Obsolete anthropomorphism : causality is believing processes with the same motivation as ours. It is also to see regularities where there are none. To believe that “the same macroscopic cause produces the same effects” (Hume) is to neglect to see its micromechanisms closely. In reality, a macroscopic event is so complex that it is unlikely that it can recur identically. Hume’s causality holds only because he substitutes the anthropogenic concept of ‘similar’ for ‘identical’. The idea of “regularity” makes no sense in ontology, similarity appearing only to an observer, not to the elements themselves.

2) The equations are symmetrical between cause and effect. As a good reductionist, Russell argues that causality is invisible in the laws of association of fundamental elements. One element in an equation does not deserve the title of cause any more than the others, which is embarrassing for a concept intrinsically endowed with a direction from one to the other.

3) The notion of succession has disappeared from physics. It is replaced by the coexistence of elements within equations. Time is not implicit in the notion of sequence but it is implicit in that of succession. But equations, as sequences, are intrinsically devoid of succession. They have no temporal character. Nor do they “scroll” one way or the other. Time is only a parameter that can be integrated into them.

The corpse is still moving

Useless then, causality? Scientific reductionism, in the wake of Russell, seems to do very well. However, one clue makes me doubt. If causality is an irrelevant concept, why debate retro-causality, or top-down causality, the bogeyman of reductionists, so vigorously? Would it overshadow a bottom-up causality that is still necessary in physics? The corpse moves in the closet.

Philosophers have heard it and causality has not been eliminated from their thinking materials. It received several successive lifebuoys:

1) Deductive-nomological (DN) model (Hempel, Oppenheim): the explanandum, what is explained (the effect), is deduced from the explanans, which explains (the cause). Besides the fact that this model cannot defend itself against Russell’s objections, it suffers from many counter-examples that have made it prefer the following:

2) Counterfactuals (Lewis, Stalnaker): event e depends causally on event c if the following two counterfactuals are true: if c had occurred, e would have occurred; if c had not taken place, e would not have taken place. Counterfactuals also have their counter-examples and for Kistler this is because they rest on the same foundations as the DN model. I had dealt with counterfactuals here before. My conclusion is that they are not adapted to the complex dimension. Contextual and not universal, they must be kept within a well-defined level of reality.

3) Agentism / interventionism (Gasking, von Wright, Cartwright, Price, Menzies, Keil): bases the asymmetry of causality —which poses insurmountable problems to previous theories— on the asymmetry between the past that we cannot influence, and the future that presents itself to us, agents, as an open space of possibilities. Agentism fully assumes its anthropocentrism, with the disadvantage of reducing the field of causality to human interventions, which discredits it as a general theory.

4) Probabilism (Suppes, Papineau, Mellor, Eells): weakened version of deductive-nomological theory, which substitute probabilistic or statistical laws for the deterministic laws of explanans. But it is also possible to arrive at probabilism by agentism: it is rational to do A in order to obtain B if one believes that A causes B because the probability B with A is much greater than B with non-A.

5) Transmission (Reichenbach, Salmon): the characteristic of causal processes is to have the ability to transmit a character, where a character is “the result of intervention through an irreversible process”. A murderous objection falls immediately: the 2nd part of the sentence, which defines a character, is the definition of causality. While the 1st part defines causality from character. Obvious circularity. However, transmission can be saved by specifying that the ‘character’ is the transmission of a conserved quantity, present in cause and effect, by example energy. Character then frees itself from cause as well as effect to define itself.

Knight Kistler enters the scene

Kistler’s solution is in gestation here: he makes causality the transfer of a constant quantity between events (and not objects), and then convincingly explains how this simple definition resists previous criticisms.

But others are popping up. The solution does not account for mimicry: the simple observation of an event can trigger its reproduction in the observer, or an action without interaction and therefore without transfer of any physical quantity. Everything happens at the observer’s house by delegation. Yet the event corresponds to a causal trigger.

Causal intermediates

Kistler gets around the problem with “causal intermediaries.” A mental representation is a causal intermediary between sensory stimuli and actions. This new category, intermediaries, seems to me to be a bin of crumpled questions. What are the “constant quantities” transferred, when they are no longer quantitative/energetic but qualitative? The sight of a falling apple triggers the development of the equations of gravitation. Proven causality but transfer very difficult to follow. Kistler uses a stack of emergent properties, from physical neural exchanges, to recover his conserved quantities. But if causality is to be seen as a pure transfer of these fundamental magnitudes, it obliges us to flatten reality and its qualities. It is difficult to get away with not reflecting more on the complex dimension of causality.

Kistler: “It emerges from this analysis that the application of the concept of causality to the explanation of complex phenomena, for example biological, is conditioned both by the presence of a transfer and by the existence of a nomic dependence between certain properties of the events to be explained.

In other words, to save causality as a transfer of a constant quantity, for complex elements, Kistler introduces a nomic dependence (a law) specific to interacting elements. It gives back to these elements the responsibility for their interactive causality. Nomic dependence based on the properties of the elements, it is inscribed in their local level of reality. These are not universal natural laws. Causality becomes local, multiple, contextual again.

What about intentional causation (top-down)?

Kistler tries to save the universality of his solution by hanging up the nomic dependence on constant magnitudes. He proceeds in this way: properties are the appearances of conserved magnitudes, he says. It is based on reductionist correlations between complex properties and underlying physical constitution. This forces him to be biased on emergence. For the quantities to be properly conserved, it is necessary to assume a completely bottom-up causality in complexity. Unfortunately without the possibility of top-down causality, the problem of intention as causality is not solved. Kistler does not directly address this difficulty, which requires a very long development and has no consensual answer.

This is where Kistler’s work ends, as he has remarkably unraveled the subject within the limits he has set for himself.


Part 2: Application of UniPhiM

Now let’s apply the double look of our UniPhiM. For Russell no theory of causality is ontological. Literally he is right, since the authentic ontological look —that of reality per se— is inaccessible. However, we use in its place the upward, pseudo-ontological look, lending to reality per se the intentions of our theories, whether scientific or not. Deductive-nomological theory clearly belongs to this upward look. Hume’s causality and agentism are the property of the downward, teleological look. As for probabilism, it belongs to both looks. How does it achieve this miracle? On the one hand, current bottom-up models are rooted in quantum probabilism. On the other hand, the mind from which the downward look starts is Bayesian: it constantly updates its predictions by updating probabilities.

Well, this merit of probabilism is a bit forced. In fact, both the quantum model and the Bayesian model of the brain belong to the upward look. The experience of looking downward is to “know” that an effect will follow a cause. It is an intuition —an impression— and not a calculation of probability, except in those who have trained their brains to do this calculation, to model their own downward look by the upward.


So we have causal theories that complement each other very well, as tools of downward and upward looks. The difficulties that each encounters as an explanation? They only appear if the inadequate look takes hold of them. The upward laments the imprecision of agentism, which mocks micromechanisms; while the downward praises its pragmatic side, which manages to account for causes of incredible complexity. Long live approximation! Long live the probability that allows us to move forward on a path of uncertainty.

When the approximation is in failure, the downward look will seek help from the upward, anchored in micromechanisms. The upward asks in return for the help of the downward to categorize the multitude of possible outcomes. The controversy died down. Every look supports each other.

… in the difference

The essential difference between these two looks is that the upward proceeds a single causal process, while the downward can invent a multitude of them, according to the regularities it perceives and the categories it applies to them. At least things are supposed to be that way ideally. In practice as upward causality is pseudo-ontological, several versions follow one another to better stick to the real per se. This is the principle of scientific revolutions. An upward look replaces another. Nevertheless, science strives to remain consensual on this subject, helped by the objectivity of the experiments. Bottom-up causality can be likened to as unique.

But in this case how to make a single upward causality coincide with multiple downward(s)? This is possible with the complex dimension. The UniPhiM explains that the upward look sees a unique and continuous path through the complexity, that of the unfolding of the processes. While the downward look sees the discontinuity of the organizations of the systems and their specific properties. It sees well-individualized layers of complexity, each with its own local causal relationships. A large number of local causalities resulting from mental representations.

Does this denigrate the possibility of a universal principle of causality? Never, actually. I am content to say how we define it with our double look, if it exists. We’ll talk about that in a moment.

Do not confuse the principle/causality and its results/causes

No proven universal causality, but local causes. Let us not confuse causes, the types of which are innumerable, and causality as a principle. The principle includes a temporal succession, Russell says, judging it to be discredited, because time is only one parameter of the equations. A direction is added to this succession, from cause to effect. Causality thus seems inseparable from time, and the causal arrow from the arrow of time. A crippling defect for Russell and his upward look, but a capital advantage for our downward looks, which brilliantly multiply the criteria for separating and identifying real entities.

We come directly, in fact, to the most popular definition of causality: it is the name given to the temporal succession of events, from the past to the future. The effect is the present succeeding the past cause. Very universal definition, too vague for analytic philosophers, but nevertheless very effective. It discards its vagueness on that which surrounds the characterization of real entities and their relations. It is not the principle of causality itself that is vague —we have just defined it very precisely—, it is the subjects to which it refers. One can imagine the causality of the downward look as in front of interlocking Russian dolls. Each doll is made of all the apparent interactions between elements of a system. The dolls can be more or less solid, depending on the quality of the interactive model. But this in no way undermines the principle of causality itself, which simply says: there is a causal model explaining the fate of the system.

In this view, the interest of causality is purely epistemic, and so is its nomic fragility. The usefulness of a causal model is correlated with the effectiveness of the model in describing the ontology of the system. This utility collapses when the model is poor. But this does not affect the usefulness of the principle, which perfectly integrates these gaps in success.



Let’s give the floor to Kistler: “The philosophical debate on the concept of causality is far from over. As I have tried to show, all attempts to find a relatively simple analysis that reduces causality to counterfactual dependence, increased probability, or transfer face significant difficulties. Some will answer that, on balance, Hume, Russell and Carnap were right to conclude that the concept of causality has no place in ontology, that is, in the theory of the types of entities that exist objectively. Rather than persisting in conceiving causality as an objectively real relationship, in a way that accords with both our intuitions and scientific results, we would do better to fall back on the study of causal appearance: there is no sense in giving to the idea that reality is causally structured; all that exists are representations of causality; However, the study of such representations belongs to the field of psychology, not ontology. Another conclusion that one might be tempted to draw from the (provisional) failure of these attempts at analysis would be to say that there is no single relationship that applies in all contexts: there are only heterogeneous causal relationships that share no common principle. According to this view, there would be a causality specific to physics, another specific to biology, and yet another that applies to psychology.”

Kistler’s conclusion is truly astonishingly accurate. On the one hand, he refrains from boasting about the solution he has proposed, the merits of which are nevertheless indisputable: it succeeds in circumventing the main pitfalls of previous theories. But Kistler is too honest not to acknowledge his weakness in the humanities. Faced with this relative failure, he encourages finding solutions on a case-by-case basis, thus abandoning the idea of a universal principle for heterogeneous causalities.

The loan of our intentions to reality… who accepts them

The story is not over, as Kistler argues. But while he is waiting for an outcome of the hybrid analyses, I think it is already there, right under our noses. Let us contrast the accessible multiplicity of the concept ‘causality’ seen by the downward/epistemic look with its inaccessibility by the upward/pseudo-ontological look. Let me explain:

The usefulness of epistemic causalities is not questionable. They allow the development of our models. To anchor our intention in the physical world, we must lend it to it. That, at its core, is causality: the lending of our intentional nature to material reality, no matter if it is not endowed with it. Everything happens as if Nature wants to behave in a certain way. Even ontologists do not dispute the ownership of its laws. So it works pretty well.

Causal pragmatism

But calling oneself ontologist requires modesty. Our fundamental theories are a pseudo-ontology, let us repeat. Reality per se is not accessible. To assert that it knows no principle of causality is an untenable fatuity. At best we can say that we do not see any trace in the equations of the intentions that we attribute to things. But how could they show them? Mathematical language is not equipped for intention. It is purely descriptive. The equations unfold. Our instruments speak the same language. If quantons have any impressions, no scientist has the means to account for them.

Here we are in the end with:
1) a large sample of epistemic causalities, a rich tool of the downward look, which it is possible to organize thanks to the complex dimension,
2) an absence of single causality for the upward, pseudo-ontological look, which does not eliminate a fundamental principle in itself in this direction.

But to tell the truth the upward look hardly needs the principle, science is doing very well without. Let us therefore keep causality where it is useful, without getting the wrong address: it is not always easy to position it correctly in the complex dimension. Its reliability is based entirely on that of our model, at the crossroads of upward and downward looks.


What’s next?

One last question is important to settle: what are the relationships between the concepts of ‘causality’ and ‘time’? Their statutes have some amazing things in common. Teleological to be sure, it is not excluded that these concepts are also ontological. Impossible to question reality per se. That physical equations do not require these principles is not sufficient to assert that they are not ontological. To understand it we must examine what a physical equation is.

An equation is the model of a portion of reality per se. It is certainly a language, before any other hope. A language that reality per se understands. But it is an anthropoid fatuity to claim that it is the real per se, to say for example that the universe is made only of information. We have no way of validating this opinion, no matter how successful the fundamental theories may be.

Since mathematics is a language, it is impossible to say what its words actually cover. What do the signs “+” or “=” correspond to in reality itself? What are these additions and equalities in essence? Our instruments are unable to tell us. Certainly they are not mere mathematical speakers like our minds; they enter directly into relationship with the real per se. But what they transcribe to our minds remains linguistic in nature. In the best case, it is the simple confirmation that our words are correct, or rather an acceptable approximation. The measurement never achieves the ideal of accuracy that is the hallmark of the equation.

So we don’t really know what a word as essential as “=” means. Or rather we guess that it covers several types of equivalence. Combining field intensities into a new intensity, or combining the possible number of states of a molecular system into heat, do not belong to the same type of equivalence. The emergences are there, hidden in the mathematical language. Those who always give the same meaning to the same word are not aware of it. Russell himself was tricked into postulating that physical equations contain equivalences that are always identical, equating language with reality per se. This is not the case, and we must suspect in these differences the origin of temporal and causal asymmetries.

Physical equations incorporate hidden dimensions in the very acronyms of their language. This is a certainty for the complex dimension. Since this defines discontinuities in the structure of reality, it is possible to reassign individually the notions of time and causality to each level of reality. This is what contemporary physics admits by returning to the elements of a system the ownership of their relational framework. In this framework are included the effective time and the causality of the system.


Causation and Laws of Nature, Max Kistler 2003
La causalité dans la philosophie contemporaine, Max Kistler 2004, Intellectica, 2004/1, 38, pp. 139-185

Leave a Comment