To have experience, to know by experience, is preferable to lack of experience, both in everyday life and in scientific knowledge. The concept of experience is rich and appreciative. How its content is described determines a type of philosophy, so central is its meaning for knowledge. If the content of experience is considered exclusively matter, that which affects our sensory endings from the outside, then we place ourselves in the framework of materialistic thinking. It is stipulated that only the material exists and is knowable through experience and this to the extent that the material object changes because everything that is given to the senses is in motion. Immutability, like that of mathematical entities, would therefore be a mark of non-existence. Nevertheless, if one considers, on the contrary, that the content of experience can be immaterial, extrasensory or intellectual in the sense in which we refer, for example, to mathematical or aesthetic experience, then ontology — the list of things that exist — is not materialistic. Now, by not being so, it opens up the possibility for knowledge of processes in other domains such as mathematical intuition or aesthetic experience that would otherwise not be adequately explained.

This bifurcation within the concept of experience, sensory experience / intellectual experience, has far-reaching consequences. The first branch of the bifurcation is associated above all with the development of the natural sciences (physics, chemistry, biology, etc.), and the second branch is associated with the development of rational sciences such as mathematics and logic, as well as with the development of sciences dealing with certain typically human activities such as aesthetic, ethical or psychological experiences. These can be the content of a phenomenological procedure that aims to analyse any content of experience, whatever the nature of its substrate, material or otherwise.

The Ancients

We shall confine ourselves here to the analysis of sensible experience in the context of natural science philosophically underpinned by the empiricist doctrine. The cognitive function of sensible experience in Platonic thought is not easy to describe. Let us recall that according to Platonism, Ideas are the only truly real objects, endowed with substance and light. The sensible things which participate in varying degrees in the Ideas serve as a stimulus to our mental process of reminiscence of the Ideas.

Experience stands before us like a ladder whose top would allow us to see the Ideas, and yet experience itself in no way possesses the intelligibility of the Ideas. Now, knowledge as reminiscence only seems possible in a field in which ideas (in the usual sense of the word “idea”) necessarily follow one after another. Clearly, not even the best interlocutor in the world would be able to extract from a child who has not studied the history of the 20th century a knowledge of the characteristics of the Second World War, while the young slave interrogated by Socrates extracts from his own background, necessarily and by deduction, a certain geometrical knowledge.

It is well known that the role of experience is clearer and more positive in Aristotle than in Plato. Experience, says Aristotle, arises from the multiplicity of memories and is not exclusive to man. Every living being develops it. The persistence of the memory of individual lived experience is both the content of experience and the basis for the construction of universal notions. What is certain is that experience — alone — is not enough to construct science. It must be complemented by imagination and reasoning. Today we say no other thing when we recognise that perception, be it sensorial or intellectual, provides us with the content of a knowledge that becomes scientific only if it is expressed in a formalism that fixes ideas. In this way, ideas become examinable, verifiable, integrable and communicable thanks to theory.

On the one hand, observation or sensible experience gives us the individual, Aristotle points out. We meet a particular man of flesh and bone, not the rational animal. We get wet by a particular rainfall during the Parisian autumn, not by the laws of physics and meteorology. On the other hand, the explanatory scientific discourse is constituted by laws that describe mechanisms referring to the universal, to the phenomenon or process in general, and not to particular sensible data. Thus, medicine deals with the (general) systems of our body, and specific anomalies are considered as cases of a certain genus. There is therefore a difficult threefold problem which persists to this day. The different epochs have deepened it without solving it once and for all in a completely satisfactory way: it is the problem of induction.

The first aspect of the problem concerns our capacity for abstraction and the formation of concepts: how can we progress from the individual cases we encounter to the species and genus, to the universal that defines them? How did we come to develop the capacity to see the universal in the particular, a sine qua non condition for the formation of concepts? Indeed, without this capacity there would be no knowledge, neither animal nor human.

The second aspect is logical, it is about generalisation: how to progress from individual cases observed locally to the whole class of cases? Then, how to move from observed regularities to laws governing the behaviour of a class of events considered as a whole?

The third aspect derives from the first two and concerns the temporal or historical dimension of induction: how to move from the information obtained at a given moment to knowledge of the past and the future? The common denominator of the problem of induction is, clearly, the passage from the known to the unknown. I think, in an Aristotelian spirit — I do not know whether the Stagirite would have agreed exactly with what follows — that these three procedures, the formation of the general notion, the formation of the law and the projection of knowledge into the past and the future are justified, science exists, because the intellect is part of nature. The intellect, being a natural entity, is capable, normally and without error, of abstracting and generalising, of filling in the gaps of information left by observation or experience. The history of ideas knows no better road to solve the problem of induction than the adoption of a realist naturalism, which recognises that the intellect is natural.

In fact, any problem of epistemology can be adequately dealt with only within an appropriate metaphysics. Metaphysics precedes science, underlies it and prolongs it. Thus, what is necessary is to recognise that the intellect is part of nature. This identity between the intellect and natural entities and processes enables the intellect to intuit reality. Without this intuition of reality it is not seen how the scientist could feel and understand what is significant in his necessarily individual and limited experience in order to form notions, laws and theories applicable to a whole class of phenomena.

Experience and induction in the modern age

The moderns forgot Aristotle's lesson: without a proper metaphysics, the problem of induction is insoluble. They went astray by claiming either (i) that generalisation procedures are a priori valid, or (ii) that inductive experimental sciences have no rational justification because generalisation procedures are not logically justifiable, or (iii) that experience is negatively useful in informing us of error when we are wrong. However, it is never positively useful because all verification is partial and local, whereas theories claim to describe and explain a universal truth. Let us look at these possibilities.

Hume: all true knowledge or reasoning can only be mathematical or empirical. Everything else, i.e. most of the discourse that forms our culture, can only contain lies or ill-formed ideas and propositions. The mathematical and the empirical are clearly separable, and it is not possible to obtain a priori mathematical knowledge of facts. These must be empirically discovered or deduced from experience, and concepts are derived from our sensory impressions. It follows that it is impossible to conceive of anything that is of a different kind from the things that are given to our experience. And even if such a conception were feasible, even if what we said about non-empirical things made sense, we would have no certainty about the truth of our judgements because they are empirically unverifiable.

Thus we find in Hume, as in all empiricists, this double sense of the concept of experience: it refers, on the one hand, to the process of experiencing something before any formalisation, demonstration or reflection, and, on the other hand, once the knowledge has been duly formalised, to the process of verification or corroboration. For empiricists, truth does not exist in a Platonic intelligible world independent of our experience or faculties. Once it comes down from its transcendental pedestal, truth is an epistemological process of verification, and propositions become true as we make discoveries.

It is true that scientific knowledge is expressed in universal laws that we cannot encounter as such in our experience, which is why there is a problem of induction. According to Hume, it is illusory to find a logical justification for generalisation. If we say that generalisation is possible by appealing to a metaphysics that presupposes natural uniformity and causality, i.e. the idea that the presence of similar causes will, in appropriate circumstances, generate similar effects, then, says Hume, we are reasoning in circles. This is so because uniformity and causality, as universal properties, are not given to our senses, and induction is already presupposed there. Nevertheless, not all empiricist philosophies are sceptical like Hume's. J.S. Mill and B. Russell, among others, believed that true inductive knowledge was possible — otherwise they would not have given themselves so much trouble to make a list of conditions to be respected in the passage from the known to the unknown.

Three characteristics, according to Hume, must be present in order to recognise that the causal relationship, the basis of induction, exists in nature: spatial contiguity -the cause 'touches' the effect-, temporal contiguity -the cause acts before the effect-, and necessity -information must necessarily pass from the cause to the effect-. The first two conditions are observable, but not necessity. What is observable, instead of necessity, is a constant conjunction. When the billiard ball A hits the ball B with sufficient force, B moves, and we can do and observe this as many times as we like. We see a fact, but we do not see that something necessarily flows from A to B. This observed constant conjunction produces in us a habit (psychological attitude) which will lead us, in the future, to expect the effect whenever the cause presents itself in favourable circumstances. This is why Hume's answer to the problem of induction is psychological, neither logical nor ontological.

Kant, having accepted, on the one hand, Hume's observation that experience is incapable of justifying causality, and believing, on the other hand, that science cannot do without it, placed causality among the a priori categories of the understanding: since experience does not give causality, thought imposes it, and epistemology comes to the rescue of ontology. Kant recognises that induction is an empirical reasoning that has a logical presumption, but the propositions thus obtained are only general and not universal. Kant's principle of generalisation states that what is appropriate to many things of one kind is also appropriate to other things of the same kind.

In his assertion that necessity is unobserved, Hume only considers information from natural perception, whereas physics teaches that there is a transformable energy that is necessarily transmitted e.g. from ball A to ball B. Even if B does not move, it will be warmer. Energy has a special status: it is conceived as something whose magnitude remains invariant despite transformations. It should be noted that this statement is not derived from experience, it is a principle, and the Principle of Conservation of Energy is one of the postulates of physics. Without the rational principles of conservation, science is impossible: what could a scientist do in a world where from nothingness something could emerge, where something could literally annihilate itself? Nothing. His hands would be as tied as if his world were evolving without law, without any stability. The weakness of the empiricist view of knowledge is its inability to account for the great principles that rationally structure the world without which the observed would have no meaning.

The empiricist criterion of existence asserts that nature is equivalent to perceived nature, that only the observable exists. Ontology (what there is) is not separated from epistemology (our knowledge of things). Empiricists support a causal theory of perception: only the object that makes an impression on our senses is knowable, and we cannot assert anything, not even that an object exists, if it is not given to perception, or if it has no verifiable relationship with an empirical object. In this context, abstract or mathematical beings do not exist. Indeed, the objects of geometry are empirical, says J.S. Mill, for whom this science is, like the others, knowledge by observation. The words that are supposed to refer to abstract beings ("point", "curve", "number", etc.) are only conceptual or symbolic instruments, fictions that when added to sensible beings do not produce contradiction. They are useful procedures for stating the laws that relate observed facts to each other. According to the empiricist tradition, there is no reason to accept or believe anything unless it is essential to the formation of our beliefs about observable beings. However, is this assumption well founded?

The empiricist criterion of existence is too anthropocentric, and it would be more in keeping with man's place in the universe — with the limited capacities of our organism and mind to grasp or reflect other natural systems — not to take our perception or conception as the sole criteria of existence. Nevertheless, it must be recognised that a theory based on our experience must be welcomed insofar as it gives us the personal right to claim that we know something. The world of experience is indeed the starting point of our knowledge. Our familiarity with the objects and events within our reach defines a kind of evidence, the empirical, the unmediated relationship that gives us things on our scale. This is why objects and notions that are too far removed from natural perception — such as discourses on the infinitely small or the infinitely large — are quickly lost in the unintelligible. Here is an inescapable question posed by empiricists: how clear and true is a discourse that cannot be brought back, by successive and duly justified steps, to the world of our perception?

During the 17th and 18th centuries, for the optimists who trusted reason (conceived either as the ultimate support of the intelligibility of the world and of truth, or as the instrument that will enable man to fight against the darkness that surrounds him), science is the work of mathematical reason. Descartes, Kepler, Galileo, Newton, and even Copernicus, although he belongs to the 16th century, come to mind. Those who saw God as the supreme guarantor of truth were convinced that the creator had constructed the world according to the perfection, beauty and simplicity of mathematics. In this context, the meaning and interest of knowledge is given by theory, and if experience is called upon, it is to answer well-posed theoretical questions and to fill in the information gaps left by theory. The attitude towards experience of much of 19th and especially 20th century science is quite different.

Today's conception of experience

For our contemporaries, patience in observation, the extreme care with which the scientist prepares an experimental set-up, are as valuable as boldness in the imagination of hypotheses. Some scientists go further: it is not uncommon to come to consider a well-done and well-controlled experiment as a criterion of existence, truth and communication. What is observed in an experiment exists, a statement is true or false depending on the outcome of the experiment, and a truth can be transmitted under the right conditions when other scientists, duly prepared, manage to reproduce the experiment and its results. Thus experimental science seems to recover, in its own way, the empiricism and idealism of Berkeley (1685-1753) for whom esse est percipi. Many experts in quantum mechanics are happy to agree with Niels Bohr that what we see is what we get — nothing else is real.

Today's empiricists (B. Russell, A.J. Ayer, C.I. Lewis, C. Hempel, and many others) know that if the criterion of existence and knowability is direct verification by someone's own sensible experience, then much, and arguably the most interesting part of natural science, is devoid of cognitive meaning. The theories of the natural sciences, especially those of the more developed sciences, include explanatory statements that appeal to unobservable entities, e.g. the forces or fields of physics; quarks, the constituents of all particles sensitive to the strong interaction, currently considered elementary. These unobservable entities are accepted because of their explanatory value and because their action is subject to strict constraints imposed by theory and accumulated information. Thus, today's empiricists no longer demand that an entity be directly perceived, but that it be at least indirectly perceptible in principle, that it have the capacity to impress our senses, even if in practice this is not feasible.

Empiricists want to do justice to science. They subordinate their criteria and demands to the development of science as it is practised, and so it is science itself that has forced these epistemologists to modify their philosophy. Since there is a scientific consensus, experience cannot be reduced to that of a particular person: collective experience must be admitted. Etant donné qu'il y a des sciences qui vont au-delà du présent (l'histoire, la théorie de l'évolution, la cosmologie, etc.), il faut admettre la mémoire de l'humanité. Les empiristes partagent avec les scientifiques l'autocritique, et cela a permis aux uns et aux autres d'enrichir le concept d'expérience. Scientists believe that experiments are used to corroborate, validate or verify a theory.

From the time Einstein stated the theory of general relativity in 1916 until the years of the space shuttle, many observations, more and more refined and sophisticated, prove him right. We can say that it is true, that it is corroborated, at least until further notice. Scientists generally know what kind of test makes it possible to say that a hypothesis or a theory is true, and they regard the remark of the fallibilists — for whom a theory is never definitively true because there is always the possibility of finding facts that contradict it — as a misplaced observation: there is no need to demand accuracy or logical or mathematical validity where there can be none.

The following remark, popularised by a recent fallibilist, Karl Popper, is logical: a large number of observations or experiments in favour of an idea does not logically justify it because the idea has a universal claim, it is supposed to cover a whole domain of phenomena, whereas any observation is necessarily local and partial. On the other hand, a single counter-example should be enough to defeat the universal validity of an idea. This logical asymmetry means that a theory can be definitively disproved, but not definitively verified. Science would progress by moving away from error, without knowing whether it is approaching the truth. Thus, any theory would be either proven false or potentially false, but never true. What is curious and implausible about this thinking is that experience is given the negative power to show us that we are wrong, but is denied any positive power to show us the way to the truth. I think, on the contrary, that it is reasonable to consider that the role of experience is also and above all positive: it is thanks to it that there is knowledge, learning and adaptation.

It is obvious to us today that observation and sense experience are part of the basis of scientific knowledge. This was not always the case (and still is not the case for some people or societies) because there were times when people turned to sacred mythical texts instead of consulting nature for knowledge. However, one may wonder whether our century has not placed too much emphasis on experience, at the expense of theory, whereas it is theory that gives meaning to experimental activity. From a scientific point of view, the experimental fact is an answer to a well-posed question. For example, optics was developed because man was intrigued early on by the properties of light phenomena such as the reflection of our face in water. Experiments accumulated without theory (as often happens in the social sciences and sometimes in natural science) remain anecdotal, and a theory without experimental basis (as often happens in philosophy and sometimes in science), without necessarily being a dream, deprives itself of an important means of touching reality and of a common space which allows us to communicate and agree.

To explain is to climb the ladder of necessity

Theories are the result of a skilful arrangement that interweaves propositions of different nature. We can make a list of these propositions, classifying them according to the distance that separates them from sense perception. We can find in a theory (i) sentences that can be reduced to propositions in the first person singular, such as: "I see coloured spots on the paper"; (ii) impersonal propositions that begin with the expression "there is ..."; (iii) propositions that describe regularities; (iv) laws themselves; in the more developed sciences these are often functional laws (they use the mathematical notion of function) that describe quantities that vary together, such as the law of falling bodies; (v) even more abstract propositions that appeal to physical entities, albeit unobservable, or to mathematical entities (e.g. geometrical structures); (vi) finally, theories may include metaphysical propositions that describe certain very general regularities of nature such as tendencies towards simplicity, optimisation (“nature does nothing in vain”, Aristotle), analogy, determinism.

The purpose of these different levels of abstraction of the propositions is explanatory: except for the first one, each level is there to give an account of what happens at the level immediately below, i.e. to show the necessity of the facts. This is why, as we move up the scale, we progress towards greater universality and necessity. It can be seen that scientific theory, when well developed, is not clearly separable from metaphysics.