1There are various different ways of envisaging how philosophy and economy cut across one another.  “Political economy / social philosophy,” but also “normative economy / moral philosophy” and “economic science / philosophy of sciences” constitute the principal axes of the crossover of the two disciplines. Within the specifically epistemological axis, which poses the question of the status of economic science in so far as it belongs to the philosophical question of scientificity in general,  we can distinguish two broad types of problems: on one hand, the canonical problem of the conformity of a hypothesis to the real—or, in other words, of the adequation of laws to a set of phenomena to be explained. Economics shares this problem of adequation with physics and, as a general rule, with every science that, in one way or another, has to pose the question of its relation to experience (such as physics, chemistry, or sociology—but not, or at least not immediately, mathematics or logic).
2Added to this problem of the empirical validity of theoretical laws is, on the other hand, the question of how to define the rationality of agents. For, as Guesnerie emphasizes, in making a distinction between physics as a study of atoms and economics as a study of “social atoms,” “social atoms differ from the atoms of the physical world not only in terms of the representations they have of that environment; they can also change it—that is to say, modify the rules of play, economic institutions, and organization themselves” (2001, 804). What type of rationality is mobilized by economic agents (whether individuals or companies)? Is the discipline of economics capable of characterizing this rationality, of understanding its principles, of demonstrating its laws, or, if not, of enumerating the uses made of them? With this second type of epistemological problem, economics detaches itself from the sole reference to physics (the question of reality) and meets with problematics inherent to human sciences (the question of rationality).
3Within these two broad questions, which delimit the properly epistemological axis of economics, philosophical references are frequent. The first question—that of reality—brings into play authors such as Kuhn, and even more so, Popper, thereby joining up with the questions that drove the debates of the Vienna Circle, such as those of reference, realism, experiment, and so forth. The second question—that of the rationality of agents—is more likely to convoke the second Wittgenstein (via the question of rule following),  but also Peirce (via the question of beliefs)  and certain naturalist philosophers (via the question of the biological basis of agents’ behavior, or computational models of it).  Without at all questioning the pertinence and fruitfulness of the utilization of these particular philosophers, the present article will seek to experiment with a more unexpected (or even untimely) apparatus, by proposing to apply the question of the transcendental argument to the field of economics. If, as Michel Bitbol says, transcendental arguments harbor “potentially considerable lessons for the philosophy of modern physics” (2000, 81), and if “the utilization of transcendental arguments [might] considerably change our understanding of what physical theories are” (2000, 82), then can we not posit, hypothetically, the possible fruitfulness of this approach within economics itself? What if the transcendental argument were to change our understanding of certain debates in economics? To try to flesh out what might, at first glance, seem a mere wager, we must, in a first stage, identify the most marked pitfalls to which the epistemological debate in economics is prone, so that, posing an apparently abstract and off-center question on the conditions of possibility of the economist’s discourse, we can show how a response to it allows us to resolve certain aporias encountered in the epistemology of economics, and to open up new perspectives concerning hypotheses on the rationality of agents.
1. The Epistemological Debate in Economics: The Twofold Question of the Status of Reality and of the Rationality of Agents
The Status of Reality
4It is without doubt Friedman’s (1953) work that gives us the truly contemporary formulation of the problem of the relation between theory and reality in economics.  For, breaking with the classical realism of the nineteenth-century economists,  Friedman seeks to show that a hypothesis does not have to be immediately realist—that is to say, it does not have to refer to a directly observable situation. He illustrates this assertion with the example of the law of falling bodies falling in a vacuum. This law is fruitful because “under a wide range of circumstances, bodies that fall in the actual atmosphere behave as if they were falling in a vacuum” (1953, 13).  Nevertheless, this does not at all mean that in fact “the formula assumes a vacuum” (1953, 13). In the same way, Friedman tells us, the economic hypothesis according to which it is “as if [companies] pushed each line of action to the point at which the relevant marginal cost and marginal revenue were equal” (1953, 16) is not to be judged on the basis of the way in which businessmen really and concretely make their decisions, in so far as “of course, businessmen do not actually and literally solve the system of simultaneous equations in terms of which the mathematical economist finds it convenient to express this hypothesis” (1953, 16). What is important is that everything happens as if people calculate in this way, just as everything happens as if the body were falling in a vacuum. Consequently, a hypothesis does not have to be directly realist, and its validity will be demonstrated if the predictions realized on its basis are most often corroborated by experience: “The hypothesis is rejected if its predictions are contradicted (‘frequently’ or more often than predictions from an alternative hypothesis); it is accepted if its predictions are not contradicted; great confidence is attached to it if it has survived many opportunities for contradiction” (Friedman 1953, 7).
5As we see, the irrealism of hypotheses is advocated here in order to align the approach of economics with the approach of physics, not so as to adopt a more regulative, or even relativist or skeptical, position.  In fact, Friedman’s recourse to the “as if” aims to permit a greater modelization, and thus mathematicization, of the discipline. The irrealism of hypotheses is a part of Friedman’s ambition to make economics a science on the model of physics and to constitute the discipline on the basis of the criteria of logical positivism: mathematicization and prediction. But whereas an ever-greater modelization of the economy, initially, had as its corollary the idea that these models may predict real phenomena,  it turned out that, in the last instance, the relation to the real was no longer defended except in hypothetical form. As Mäki emphasizes, we have arrived at a paradoxical situation where to do science means to postulate that “Entity X might exist, Theory T might be true” (Mäki 2002, 92) without ever being able to prove it. From the proposition “it is as if agents maximized their profit,” we have arrived, over time, at “it is as if this proposition were scientific.” This failing of “blackboard economics”  obviously generated various responses which are not satisfied with the simple affirmation according to which scientific theories X “might relate to entities that themselves might exist,” but which do indeed try to show that “scientific theories are, and must be, constructed and employed with the aim of describing the causal processes that constitute the way in which the world works” (Mäki 2001, 371). Our aim here is neither to critique mathematical models nor to give an inventory of the multiplicity of approaches in economics, which have a widening distance from “the way the world works,” but to make visible the skeptical or relativist trend of a theory (Friedman’s) which at the start, obviously, was not skeptical or relativist at all, since it inscribed itself explicitly in the wake of logical positivism. For, if to think economics as a science means merely to postulate that theory T “might” be true and that what it speaks of “might exist,” then it is clear that an initial positivism has undergone a metamorphosis, to the point of agreeing with certain forms of relativism and skepticism. This becoming-skeptical of economic science is sometimes clearly propounded by certain epistemologists of economics. Thus, McCloskey (1994) thinks that economics involves an analysis in terms of rhetoric, whereas Weintraub (2002) shows the extent to which economists’ language is linked more to the evolution of mathematical language than to the consideration of reality. Ochangco (1999), in an interpretation that uses the notion of language games in a skeptical framework, shows, on the basis of an analysis of the notion of value, that different definitions exist but that they cannot be evaluated against each other because they belong to incommensurable language games. 
6Now, what matters to us here is that the becoming-skeptical of what amounts to the great majority of works in economics today, is parallel to the becoming-skeptical of philosophies of science which take as their starting point a reflection on physics in its relation to experience. To develop this point, let us recall that Popper, against the realism of the Vienna Circle, which he judged to be too immediate, established that a scientific theory cannot be validated, but only falsified, by experience. Although falsificationism renounces “any claim implying that theories can be established as true” (Chalmers 1976, 34), science is nevertheless distinguished from pure ideology in so far as it, and it alone, can be submitted to empirical testing and can develop protocols for evaluations against the real. This exigency, which is no longer a positive affirmation of the veracity of physical reference (logical positivism) but a submission to the possibly negative verdict of experience, will be judged by Lakatos as still being too dogmatic. For him, Popperian falsification presupposes the distinction between theoretical or speculative propositions and factual propositions immediately referable to experience (Lakatos 1970, 7s). Now, the reality of this distinction—a necessary condition in order for scientific theory to be falsifiable—in fact turns out to be impossible to establish. Consequently, scientific discourse is no longer possible—that is to say, a discourse of propositions that can be qualified as scientific as opposed to ideological propositions.
If . . . accepting the demarcation criteria of dogmatic falsificationism, we deny that facts can prove propositions, then we certainly end up in complete skepticism: then all science is undoubtedly irrational metaphysics, and should be rejected. Scientific theories are not only equally unprovable, and equally improbable, but they are also equally disprovable” (Lakatos 1970, 19).
8To escape this impasse, Lakatos develops his methodological falsificationism,  for which a hypothesis’s pertinence (it is no longer a question of the reality of the reference intended by the statement, but of a degree of acceptability of such a proposition by the community) will depend upon its capacity to resist history for as long as possible. It is therefore, in the last instance, history that will tell whether a research program will “degenerate” or not.  This weakening of Popper’s falsificationism logically leads to the relativism frankly advocated by Feyerabend. If, Feyerabend explains, only history (that is to say, the necessary utility or needs of a civilization at a given moment) allows us to attribute a scientific status to a discipline at such a moment, then current science has no more scientific legitimacy than ancient myth. Further still, Lakatos, says Feyerabend, “arbitrarily selects science as a measure of method and knowledge without having examined the merits of other professional ideologies” (Feyerabend 1975, 237). In short, we must be willing to adapt to a world where there is as much “truth” in the novel as there is in scientific discourse, where the statement “the earth is blue like an orange” is just as valid as “the earth rotates.” Feyerabend presents this frankly relativist thesis as the consequence of Lakatos’s thesis, itself a consequence of the “dogmatism,” judged to be still too close to logical positivism, of Popper—who himself aimed to relax the logical-positivist hypotheses he thought too “dogmatic.” What can we conclude from this analysis of the development of each of the positions involved? Firstly, that the “dominant” economics, just like philosophy of science in its Popperian aspect, is directly under threat from skepticism and radical relativism. Echoing the absolute relativism of Feyerabend, certain historians of economic thought conclude as a consequence that their interrogation must bear not upon the truth value of economic propositions but upon the rhetoric employed by economists. This is, for example, the project of McCloskey, who writes: “We need a new philosophy, or anti-philosophy, to understand economic science as it actually works. The form of anti-philosophy recommended here is the oldest one: ‘rhetoric’” (1994, xiii). This skepticism, far from being the opposite of positivism, seems to be an expression of its natural evolution. The oscillation between positivism and skepticism is not a pendular movement carrying us from one position to its opposite, but seems instead to belong to a logical tendency, since positivism generates relativism internally. This becoming-skeptical of positivism seems to be the principal pitfall toward which current economic theories are converging. Therefore, is this nullifying aporia (in which one position seems to imply its opposite, logical positivism calling for Feyerabend) inherent to the question of reality, or does it also afflict debates concerning the rationality of agents?
The Rationality of Agents
9Obviously economics contains multiple ways in which to think the rationality of the agent. Nevertheless, this multiplicity, far from being irreducible (with as many definitions of rationality as there are economists) can be boiled down to three broad groups.
10The first way, that of neoclassical economics, consists in positing the hypothesis of unlimited rationality. Within the framework of this hypothesis, which gave rise to Homo economicus, the perfectly rational agent adopts the behavior that leads to general equilibrium (this behavior being the best possible response, once the scarcity of resources is taken into account). The infinite multiplicity of behaviors observed in reality is reduced to one single model of the so-called “representative agent.” This representative agent, supposed to incarnate all individual agents,  knows, by definition, how to maximize his interests.  This hypothesis implies that the agent has at his disposal an infinite capacity for calculation and perfect information about the world, allowing him to make the best possible choice. This unlimited rationality is similar to the rationality of the Leibnizian god, who is omniscient, and calculates the best of all possible worlds. In this hypothesis, the agent’s behavior is entirely determined by the calculations he carries out—just as, in Leibniz, God’s will is subservient to his understanding.
11This conception, still dominant in economics, of a human rationality equivalent to that of a Leibnizian God, obviously finds itself opposed by the hypothesis of limited rationality. Within this second broad category we might recognize two subgroups. In other words, we find, in economics, two main ways of moderating the hypothesis of the perfect rationality of agents: either by saying that agents do not have perfect information about the world at their disposal, or by saying that they do not have an infinite capacity for calculation.
12The first way (imperfect information) leads to a sophistication of the analysis of the behavior of agents, who must, in order to make decisions: (A) anticipate other agents’ possible actions; and (B), construct contractual arrangements to reveal information and to avoid behaviors that seek to profit from asymmetries in information. But being able to anticipate in a world where information is imperfect becomes an extremely complex matter. In such situations, if agents wish to coordinate effectively, they will have to put in place devices through which information can be acquired or revealed.  To the calculations of the maximization of interest, the agent will therefore have to add calculations of the costs of finding information and/or calculations of the costs of the devices necessary for the revelation of information.  To resolve these complex problems, individuals would have to possess a calculatory capacity as great as that necessary for the agent in the situation of unlimited rationality. The second way of relaxing the hypothesis of unlimited rationality consists in considering that even in a situation where agents have at their disposal all information necessary to take an “optimal” decision, nothing assures us that they will actually do so (Heiner 1983). We are thus confronted here with an effective limitation of the cognitive capacity of agents (imperfect information, limited capacity for calculation). The rules of the functioning of the economy (the institutions, to use North’s expression) serve to palliate this cognitive limitation of agents, and to facilitate their choice, inducing certain regularities in their behavior.  Thus reformulated, the question of rationality becomes a more directly general question, close to philosophical, anthropological, or sociological thematics. (Is reason an adequate response to a set of external stimuli?  A simple calculation on the basis of individual interest? Free reflection? The product of genetics? Adaptation to the environment? And so forth.) In this framework, the problem for the economist becomes that of knowing how to characterize the type of regularities that he is trying to uncover in the behavior studied. Is it a question of an entirely predictable regularity, in which case a certain type of determinism or necessitarianism of behavior can be posited; or does the agent’s behavior remain so unforeseeable and irreducible to any regularity that ultimately any hypothesis about this behavior will be in vain? It is between these two extreme limits that we find set out the different studies that accept the general hypothesis of this type of limited rationality. 
13For example, Hodgson (2001, 2004), taking up the thesis of so-called universal Darwinism, posits that the concepts of Darwinism—heredity, variety, and selection—can be applied to all complex systems.  Each of these concepts in fact has an equivalent in economics: “Underneath the very real differences of character and mechanism, biological evolution and economic evolution might have types of process or structure in common, when considered at a sufficiently general level of abstraction. At this level, we are not addressing mere analogy. We are considering a degree of identity in reality” (Hodgson 2003b, 368). At this level, the habits of individuals may, on the model of genes in biology, be considered as possessing inherited characteristics (storage and transmission) in so far as they say in what manner it is possible to act (storage) and are transmissible by imitation.  So there exists a social equivalent to the biological gene, in the form of habits, encompassing habits of behavior and of thought alike. In sum, “the crucial point is that all action and deliberation depend on prior habits that we acquire during our individual development” (Hodgson, 2003a, 167). Habits are, therefore, essentially the result of social causes explained in biological terms. Thus, we have here an attempt at the reduction of the behavior of agents to regularities governed by necessity. As is the case for every naturalism,  of which Hodgson’s evolutionism is a foremost case, it is a matter of thinking rationality on the basis of laws of nature. 
14In opposition to such an enterprise, other economists recuse any application of determinism to human behavior, any submission of rationality to natural laws; thus Commons, for example, takes into account the factor of human will, and shows how the human sciences, unlike solely natural sciences, must accept what he calls “voluntarity”—that is to say, “the active and organizing character of thought” (Bazzoli 1999, 72).  “To suppose that the product of human action is predictable, is to inscribe it in a determinist conception of the world (the world follows a rational order) and to deny the complexity and incompleteness of evolutionary systems” (Bazzoli 1999, 76). But how far can we push this incompleteness without renouncing the discourse of science, which serves to establish regularities? Commons advocates Peirce’s pragmatism, but how far are we to go in the interpretation of this “pragmatism?” Must we, like Rorty (who sees himself as its most complete representative) go to the extent of “considering science as a genre of literature” (1982, 63)? Is not complete relativism the vanishing point of pragmatism? 
15However this may be, two extremes limit economists’ inquiries on the question of the rationality of agents: on one side, an explanation of the behavior of agents on the basis of laws, an explanation that always induces a certain form of determinism (the Leibnizian God or the submission to historico-biologico-social laws); on the other, the threat of relativism as incarnated by Rorty, when he thinks of scientific theories as many different forms of rhetoric. Most authors situate themselves between these two extremes, oscillating between the two positions, sometimes juxtaposing them. 
16Thus, at the end of this analysis, we can allow ourselves to characterize the principal aporias that the epistemology of economics encounters relative to the two questions that structure it (those of reality and rationality). In the first case, we have shown the paradoxical engendering of skepticism on the basis of positivism; in the second, the constant oscillation between these two extremes. Certainly, in both cases, most authors seek to occupy median positions—for example a Popperian position in the debate on the status of reality, or a Peircian position in the controversy over the nature of rationality. However, is not such a position condemned to the problematic juxtaposition that leads finally to relativism, the true vanishing point toward which they all converge, and on the basis of which they are all ultimately organized? In other words, do we not find the truth of Popper, in fact, in Feyerabend’s position, that of pragmatism in Rorty’s—just as, moreover, the latter both affirm? Faced with this situation, must we not find another angle from which we can pose these questions in another way? Is it possible to organize the space of discussion of economics otherwise than on the basis of this one sole vanishing point?
2. The Transcendental Argument
Definition of the Transcendental Argument
17The angle from which we propose reconsidering the preceding questions is a type of argument that, since Strawson, has been called the transcendental argument, and which, as S. Laugier reminds us, stimulated debates “that dominated analytic philosophy at the end of the twentieth century” (2000). Let us recall, then, that Strawson (1966, 72–89) identifies in Kant a type of argument that seeks to show that if one does not accept a certain concept, then one can neither think nor act as one in fact does.  Thus, Kant justifies the recourse to a concept of cause by indicating that this concept is the necessary condition of our experience of succession. Showing how certain concepts or series of concepts are necessarily implicated in cognitive operations which, in fact, we do carry out is, to Strawson’s eyes, the core structure of the “transcendental argument.” He will utilize this argumentative device to show, for example, that we could not differentiate the objects we perceive were we not able to think ourselves and the objects as two entities coexisting in space. Following him, G. Evans (1982), setting out from a general type of cognitive operation presently effectuated (such as, for example, perceiving objects, referring to something, or predication), tries to identify the presuppositions it necessarily implies. In the same way, J. Habermas and K. O. Apel use this type of reasoning when they undertake to bring to light the necessarily underlying conditions of our common communicational experience. The transcendental argument thus consists in drawing out the presuppositions that are necessary to an experience of thought or speech that, in fact, is familiar to us. It does not bear directly on the reality of the world, but on cognitive operations.  With it, it is possible to show that the everyday representations of human beings would not be possible without a certain number of conditions, conditions which it is the researcher’s aim to determine. This is why the transcendental argument is often stated in a negative manner through propositions of the type: “We cannot do otherwise than to posit q,” or “We cannot not posit such-and-such conditions, in order for such-and-such representations or operations to become intelligible.” It is clear, within such a framework, that the more difficult it is to question the initial premise or factum, the more compelling the transcendental argument will be.
18Given this general definition of transcendental argument, we must consider the first attempt to introduce it into economics—that of Lawson.
An Attempt to Introduce the Notion of Transcendental Argument into Economics: The Example of Lawson
19To our knowledge, Lawson is the first to utilize, expressis verbis, the notion of transcendental argument in the field of economics. His objective is to respond to the twofold question of the relation between economic discourse and reality on one hand, and that of the nature of the rationality of agents on the other, with, in both cases, the intention of overcoming the arguments of skeptics and radical relativists. 
20As far as the first question is concerned, the inability of today’s scientific activity to demarcate itself from other activities (such as that of fiction), according to Lawson, is the result of the runaway success of Popperian falsifiability. Resorting to the transcendental argument, for Lawson, consists in demonstrating the impossibility of not positing reality. In his argument we find a negative form—“one cannot not posit X”—whose inherence to the transcendental argument we have already noted. In this regard, Lawson writes: “Thus, those who would reduce science to a conversation presuppose objects of that conversation. Even if they claim that conversation is only about conversation they thereby introduce a distinction between the conversation as an (intransitive) object of knowledge, and conversation as a (transitive) form or medium of knowledge” (Lawson 1997, 48–49). Thus the “conversation” of economists bears upon an object—“economic reality”—and they cannot not posit this reality as their horizon.
21This realism—which overcomes the skeptical objection and which is doubtless more secure than that with which Bouveresse (1992), for example, will counter Rorty, under the name of a “motivational realism of science”—nevertheless does not yield a discriminating criteria able to tell us why the theory of general equilibrium, or universal Darwinism, might be better than other concurrent theories. Realism as advocated by Lawson is not incompatible with the affirmation of an infinite pursuit of the determination of reality: “I have allowed that things exist and act independently of their descriptions, this stance is complemented not by an absolutist or foundationalist position in knowledge but by an epistemological relativism, by the thesis that we can only know these things under particular (historically and socially relative) and potentially transformable descriptions” (Lawson 1997, 58).  In this precise case, the transcendental argument consists in showing that we cannot not affirm a certain form of reality. Thus, it allows us to avoid the slippery slope of fallibilism; but it does not at all seek to determine directly which theory best reveals this reality (that is to say, for Lawson: its structures, mechanisms, powers, tendencies).
22The author uses the same argument to confront the question of the determinism of the behavior of agents. For example, he undertakes to show how a certain freedom of agents is a necessary condition that cannot not be posited. If one “holds that, under certain conditions x, an agent chooses in fact to do y,” this means that one maintains that it “could really have not chosen y” (Lawson 1994, 269). Without such an argument, according to Lawson, the notion of choice obviously becomes meaningless. Here again, the type of argumentative strategy proper to the transcendental argument appears clearly: we begin here with a fact, which is a cognitive operation rather than a substantial entity, namely that we speak of a choice between x and y. Now, to speak of “choice” presupposes a certain number of conditions without which this usage would be meaningless. Among these implicitly presupposed conditions is the capacity of agents to not have an entirely predictable behavior reducible to an external causality.
23Nonetheless, as Viskovatoff (2002, 703) notes, such a transcendental argument rests upon a postulated premise—namely, “the intuition that human agents have the ability to make real choices” (Lawson 1994, 269). Consequently, if the transcendental argument has a certain validity, this validity is linked to the truth of the premise thus posited. This is admitted by Lawson, who writes: “It is necessary to recognize that the premises of the transcendental argument are always contestable, never indubitable” (1994, 271).
24Doubtless it is on this point that it seems possible to go further than Lawson, who, to our eyes, does not deploy the full potential of the transcendental argument. For Lawson considerably weakens the argument by reducing it to a chain of reasoning of the type: p; p would not be possible if we did not think q; we must think q; it is true or necessary that q. 
25Such a chain of reasoning is in fact entirely reducible to a classical logical argument (for example, one of the modus ponens type). For proposition two (“p would not be possible if we did not think q”) can be unpacked as follows: that proposition one (p) presupposes proposition two (p would not be possible if we did not think q) means that one is only true if two is true. The transcendental argument would thus be nothing but a deductive logical argument in which proposition two is just a premise with the status of a postulate. This is why A. Boyer (2002), challenging the specificity of the transcendental argument, will affirm that there is no place in science for arguments other than:
26(a) Deductive arguments, in which the conclusions cannot contain anything more than the premises, and in which, consequently, all ampliative inference is necessarily illegitimate. These deductive arguments cannot be said to be absolutely certain in so far as the premises can never be anything other than postulates admitted by pure convention (such as, for example, Euclid’s parallel postulate).
27(b) Inductive, generalizing inferences, which, certainly, help orient us in the real, but never have any definitive validity whatsoever, since they depend on experience.
28This is why, Boyer concludes (2000, 37), recourse to the transcendental argument is in vain in so far as “reasoning is either deductive or non-deductive: in which case, it is inductive. Tertium non datur.” Faced with this objection, to which Lawson’s utilization of the transcendental argument is assuredly vulnerable, we should like to propose a version of the transcendental argument that prevents it from removing its specificity.
Another Version of the Transcendental Argument: Performative Noncontradiction as Structure of Transcendental Argumentation
29The contribution of K. O. Apel, in relation to the accepted meaning (as Strawsonian as it is Habermasian) of the notion of the necessary condition of possibility of a fact, is to have thought this notion on the basis of the concept of performative noncontradiction. He writes, on this subject: “I am of the view that it is important to reflect, at the beginning of any philosophical work, and thus equally at the very beginning of our analysis of the lived world, about necessary transcendental presuppositions of argumentative discourse—that is to say, about what cannot be contested without performative contradiction” (1981, 924). In Apel, the transcendental presupposition becomes synonymous with pragmatic or performative noncontradiction. Let us recall briefly that performative contradictions are by definition always false, not because of their meaning, as in the analytic proposition: “every married man is a bachelor,” but because of the status of their utterance.  Because of this, since the contrary of always-false utterances is necessary, then we can term “pragmatically necessary” utterances that cannot be denied without self-contradiction. It is clear here that performative contradiction is not reducible to strictly analytic (A is A) or deductive (if A then B) formal reasoning, but is defined strictly as a statement that annuls itself from the very fact of its having been said. Within this category, two types of contradictions can be discerned: (1) “pragmatico-empirical” contradiction, which is to do with the empirical conditions of discourse. In this case, the content of discourse and the empirical conditions that must be in place in order for this discourse to take place are performatively contradictory, as, for example, in the statement: “Here I am in this shipwreck, of which there were no survivors;” (2) “pragmatic-transcendental” contradiction, which has to do with the conditions of fact of the argument itself, for example when I say: “Truth does not exist.” We find this second case in the classic self-refutation of radical skepticism, and the consequent necessary suicide of the sophist. As the inverse of this self-contradiction, performative noncontradiction postulates the necessary congruence between the statement and its utterance. So how does this reformulation of the transcendental presupposition in terms of performative noncontradiction allow us another employment of the argument in the specific field of economics?
30We can show this first of all on the basis of Lawson’s first question as to the status of reality. In fact we can show not only that a minimum of realism must be posited, but what is more, that it is contradictory not to hold that, in this matter, one hypothesis is better than the others. To do so, let us take Feyerabend’s example, which posits that “the only principle that does not inhibit progress is: anything goes” (1975, 20). In saying this, he falls into the following aporia: If “anything goes,” then this means that the proposition “anything goes” cannot be accepted. Here we can see the contradiction between the content of the proposition (“anything goes”) and one of its presuppositions (that the proposition “anything goes” must be accepted). This type of contradiction allows us to assert that the thesis according to which all hypotheses are valid cannot be accepted, since, at a certain level—that of the justification of this hypothesis itself—it contradicts itself. The proposition “there is a hypothesis that is better than the others” must therefore be accepted, since its contrary is always false. This obviously does not determine the content of the hypothesis likely to speak more correctly of this or that reality; but it does lead, against relativism, to a position that goes beyond Lawson’s, in so far as it lets us say that some hypotheses are indeed better than others, and that science cannot not presuppose this point. It is possible, then, to see here a difference with the “motivational realism” formulated by Bouveresse, according to which scientists have a propensity to presuppose a reality, and if they did not do so they would have to stop working. The transcendental argument is more constraining, since it says that it is inherent to scientific discourse to make this presupposition because, were it to posit the inverse, it would negate itself, self-destruct, contradict itself. Generally speaking, it is a question of accepting a detour via metatheory, qua interrogation of the “claims” of a discourse, in order to be able to determine to what extent certain stances are inconsistent. Feyerabend, asserting that “anything goes,” claims the validity of his discourse, and cannot not make this claim in relation to the regime of discourse he is invested in (a nonliterary regime with a claim to validity). With regard to this claim inscribed within his sentence, the latter becomes necessarily false—falsified not through recourse to the empirical but through the consideration of its pragmatic consistency (performative contradiction).
31This transcendental argument reformulated on the basis of performative contradiction allows us, furthermore, to sketch out responses to the second question that economics poses, in relation to the rationality of agents. For with the aid of this argument, we can show how certain theories about “human rationality” self-destruct. For example, to reduce human rationality to a pure calculation of strictly individual interest (something frequently considered in economics) comes down to claiming, at the same time, that the scientific discourse that affirms this proposition is itself also the product of pure individual interest; in which case, it has no general validity, and there is no need for us to discuss it or to argue against someone who implicitly tells us that, in saying what he says, he does nothing but express the preferences of an individual interest.
32The necessary compatibility between the rationality imputed to agents and the rationality exercised by the economist implies the impossibility of certain hypotheses.  The metatheoretical aspect thus turns out to be decisive, including at the level of the most specific theses. Now, it has to be said that few economists take it into consideration, and this disinterest cannot boast that it is anchored in the concrete, since we have seen the extent to which taking this abstract level into consideration allows the evacuation of a number of self-contradictory theories. Simon was not wrong in wanting to consider this aspect of the question as inherent to any reflection on economics. It is through a confrontation with his theory that we wish to show the real fecundity of the transcendental argument in the field of economics.
3. Fecundity of the Transcendental Argument in Economics
33To better appreciate the contribution made by the transcendental argument, we must, in a first step, recall the conception of rationality defended by Simon.  He writes: “By a weak definition of rationality, virtually all human behavior is rational. People usually have reasons for what they do, and, if asked, can opine on what those reasons are” (1991, 1). Nevertheless, this rationality is limited, in so far as man finds himself precisely in an environment whose every element he does not have mastery over, and all of whose potentialities he cannot calculate; he can moreover harbor illusions as to what he does or what he is (Simon 1991, 1). Thus, individuals decide as a function of limited “aims or values,” on the basis of a series of finite inferences and observations, which may be false. But even if the inferences are false, the behavior can still be said to be rational. This limited rationality can be described with the aid of procedures deployed by the agents to make their choices. Thus, the rational agent seeks different solutions (so-called “alternative solutions”—a search procedure), and adopts one of them, not because it is the best in itself, but because it seems satisficing  in relation to the aims to which the agent aspires. In a word, it is a matter of solutions that are subjectively meaningful even if they are not objectively or in themselves (from the point of view of a calculating, omniscient God) the best. These decisional procedures can be described and analyzed scientifically using artificial intelligence. The algorithms of computers give us a good approximation of human thought and are capable of describing the way in which individuals make their decisions. 
34Having recalled this conception of rationality, we can now analyze the structure of Simon’s argument. In his discourse, three parameters have to be taken into account, namely: (1) the discourse of the economist concerning the rationality of agents; (2) the content given to the rationality of the agents; and (3) reality. These three parameters multiply the economist’s tasks, in comparison to those of the physicist. On one hand, the economist must try to determine what reality is; here we find the classic question of realism and reference. On the other hand, he must define the rationality of the individual—the economist must determine the representation that the economic agent makes of his reality.  And finally, he must try to show there is no contradiction between the rationality he puts to work qua scientist, and the rationality he imputes to the individual or economic agent. One might represent these tasks with the aid of the following schema, in which the agent makes himself a representation of the world with the aid of the rationality that is his, and the economist, for his part, constructs a conception of the world that includes a representation of the way in which agents represent reality to themselves (see Schema I).
35Faced with these tasks which are those of the economist, what justifies use of the transcendental argument? Human rationality and the rationality put to work by the economist must be congruent. The rationality of the economist, through its very exercise, cannot enter into contradiction with what it says about human rationality. In a word, scientific discourse must beware not to invalidate itself by imputing to people a rationality that the very fact of its own discourse would contradict. Here we do indeed rediscover the transcendental argument, which might here be defined as a reflexive a priori, since we can say in advance of all experience of the theory to be analyzed that the enunciative status and the content of the statement must not enter into conflict and that, if this does happen, the theory must be rejected.
36If we apply this framework to Simon, his position implies that the scientist’s rationality also belongs to procedural rationality (Langley et al. 1987). Now, Simon tells us that this human rationality is determined, that is to say, is predictable, reducible to mechanisms that the computer can effectuate. This is highlighted by Boyer who, discussing Simon’s conception, says of it: “The agent ‘satisfying himself’ appears not to be able to be anything other than an automaton following rules and routines, incapable, at the limit, of any critical or reflexive view on the routines in question” (1994, 165). This determinist conception is also discussed by Searle:
Herbert Simon . . . says that we already have machines that can literally think. There is no question of waiting for some future machine, because existing digital computers already have thoughts in exactly the same sense that you and I do. Well, fancy that! Philosophers have been worried for centuries about whether or not a machine could think, and now we discover that they already have such machines at Carnegie-Mellon (1985, 39). 
38We can now confront this definition of rationality by Simon, and show that such a determinism cannot be upheld in this precise form. If we envisage rationality in terms of routines or automatisms, of storage and symbols that are recombined with each other according to precise algorithms that we can reconstruct; if, therefore, we can predict the combinations to come—then scientific discourse can also be understood in this manner. Now, if Simon, when he speaks of science, is claiming that it is a matter of a succession of routines or a rule-governed combination of stock structures, he renders his own activity and discourse null and void. As pure succession of routines, science cannot claim to be anything other than a mechanical repetition, repeatedly carried out. So it is clear that the act of the construction of rationality realized by the scientist supposes the mobilization of a rationality that is an exception to his theory of rationality; in which case, his theory of rationality can no longer claim validity since it cannot be applied to that which sets it in motion. To say it again in the same terms we used above, Simon’s theory of rationality is falsified not through recourse to the empirical, but through the consideration of its own pragmatic consistency (performative contradiction).
39It is significant that Simon himself recognized that he was not able to explain the rationality of the sciences, and particularly the process of discovery. He even speaks of a missing link in reference to the process of the creation of new representations: “The process of discovering new representations is a major missing link in our theories of thinking and is currently a major area of research in cognitive psychology and artificial intelligence” (Simon 1996, 109).
40It will be said that what Simon is arguing is that, in the future, we will succeed in accounting for this process. Nevertheless, not only can we not accept a proof by “induction on the basis of the future,”  but, what is more, it turns out that the model used—that of computation as a succession of routines or a rule-governed combination of stock structures—generally makes the discovery of anything new impossible! With his model, Simon prohibits innovation and, at the same time and under the same relations, says that later an innovative discovery will be made. Paradoxically, Simon does not ask a science to come and confirm his hypothesis, but instead to prove its falsity.
41In regard to our analytical schema, the question of rationality is not the only problem that crops up in Simon’s theory.  For the relation to reality also remains questionable. The concern for a broader realism of hypotheses leads Simon to convoke reality as an argument for the greatest validity of a hypothesis. Thus, in the debate opposing him to Friedman on the hypothesis of the “as if,” he will defend the idea of a more realist vision of the behavior of agents, the most realist vision that the scientist can bring about, even if it is imperfect (Simon 1963).
42Taking up the above schema once more, we can, therefore, apply it to Simon, bringing to light two gaps: the incompatibility between defined rationality and defining rationality, which leads to the impossibility of taking account of changes in representation, and to the convocation of reality as justification for the more realist character of hypotheses, when this reality is described as not being accessible outside of the representation one may have of it (see Schema 2).
43This bringing to light of the utility of our analytical framework in the refutation of certain of Simon’s propositions allows us in return to sketch out some more positive propositions as to what rationality is, and thus to display some eminently concrete aspects of the utilization of the transcendental argument.
44Two propositions that emerge from the transcendental argument should be retained: (1) Theses that posit an entirely determined rationality (submitted to strict natural or social causalities) are self-refuting. Consequently, the freedom (or nondetermination by a natural causal series) of the rationality of agents must be presupposed in order for scientific discourse itself to be protected from performative contradiction. This necessity of positing the total nondetermination of the rationality of agents shows us that every vision of economics that claims to provide natural laws or to require rules, to the exclusion of every other hypothesis, may be recused on the sole grounds that it denies freedom, and that this denial makes the economist’s own discourse impossible (self-refutation). Economic science cannot index human behavior solely through natural or social laws, without renouncing the freedom of its agents and, in doing so, destroying itself qua science. Naturalism, so dominant today, is a position condemned, sooner or later, to self-refutation.  It is important to emphasize here that freedom is not a moral postulate (the domain of moral and political philosophy) but rather a condition of possibility of the consistency of the economist’s discourse (the domain of the philosophy of knowledge and epistemology). To save economic discourse from self-refutation demands that we hold certain hypotheses on the rationality of agents; these hypotheses are made on the typical mode of transcendental argumentation: “It is impossible to give such and such a content (x, y) to rationality without falsifying this content through the very fact of deploying it.” (2) One can no longer, in relation to the same argument, posit that individuals are incapable of creation—that is to say, incapable of imagining or anticipating new forms of organization of the economy or of society that do not exist today. Creation, innovation, and anticipation, necessary for the construction of a new scientific discourse, cannot, without contradiction, be denied to individuals when they think a form of economic organization. It is doubtless here that the contributions to an economic science can be the most concrete: two perspectives for reflection can, in fact, be opened up.
45Firstly, the necessary position of the nondetermination of the rationality of agents leads us to rethink the question of the relation between individuals and existing rules. For, in relation to the debate between a conception of the rule as a necessarily imposed constraint (naturalism, social determinism) on one hand, and the conception of a dissemination of every rule through an incessant interpretation (hermeneutic skepticism) on the other, we can permit ourselves to contribute the following points, as consequences of our argument: (1) Faced with the same situation, an agent may have many simultaneous visions of the same phenomenon; (2) he may modify this representation independently of all necessitating external stimuli, by free reflection on his existing representations (innovation). Hence, if existing rules can be considered as shared beliefs about the functioning of the world (see on this point Searle (1995) in philosophy, and Aoki (2001) in economics), it nevertheless remains that the application of these rules does not in the slightest signify approval of the rule. 
46Next, the question of agents’ capacity for innovation can also be rethought on the basis of the results of the transcendental argument that demonstrates the pragmatic impossibility of an entirely causal explanation of agents’ behavior. This nondetermination allows us to understand the possibility of innovations that break with existing routines (something Simon cannot account for without coming up against the question of the “process of discovering new representations”). Thus we have here the possibility of rethinking the Schumpeterian innovator, who, breaking with the existing equilibrium, initiates the innovations that explain the dynamism of economies.
47In consequence, the world that the economist must try to apprehend is a world where the hypothesis of a freedom of rationality of agents is an impassible condition of possibility. This freedom, while of course not excluding the possibility of economic agents’ actions being constrained or governed by rules, does allow us to understand how the interpretation and the modification of rules is possible.
48At the end of this journey, four results can be highlighted.
49(1) Firstly, the transcendental argument allows us to show that neither strict determinism nor radical relativism is acceptable. Determinism (whether of a biological or social order) is self-refuting for the same reasons that relativism is. But, beyond this refutative aspect, our analytical framework has a positive aspect, since, defining the limits or breaking points beyond which a theory must be refused, it allows us, in return, to propose other definitions or hypotheses. This is the second outcome of our analysis.
50(2) We can propose a redefinition of the notion of rationality, through the demonstration of the impossibility of imputing to agents either an exclusively instrumental rationality (whether it takes the form of a calculative rationality or a rationality adaptive to its environment), or a rationality exclusively founded on social determinisms. The rationality of agents is a rationality capable of escaping routines, a free rationality, capable of innovation. This definition has not been attained empirically—by showing, for example, that purely calculative, adaptive, and determined forms of rationality do not allow us to account for all the phenomena (altruism against the quest for personal interest, innovation against adaptation and social determinism).  This definition has been attained on the basis of the demonstration that every contrary hypothesis is self-refuting. And this general argument implies, more precisely, our third outcome.
51(3) The forms of rationality can be multiple, and agents can have a multitude of representations of the same phenomenon. This necessity of thinking the multiplicity of possible representations allows us to open up new perspectives—for example, that of another approach to the notion of shared beliefs. Beyond its general and metatheoretical aspect, our argument allows us to resolve the question of how individuals, even when they are surrounded by institutions that constitute so many rules (including the rules of language and so forth) can construct new representations independently of any external stimulus; and in so doing, it allows us to bridge the gap between holistic approaches and approaches in terms of methodological individualism.
52(4) Finally, this specific angle on the problem also makes it possible to place in question the traditional divisions between economic theories. For example, the manifest and real opposition between Friedman and Simon cannot hide the fact that, in the last instance, these two authors share the same vision of science as predictive, the same will to test through experimentation, the same concern to identify necessary laws of the behavior of agents. At the same time, they are prone to the same types of critiques, once our framework is applied to them—for example, the impossibility of according to agents a form of rationality that does not come into contradiction with the rationality that they themselves exercise. The argument thus furnishes, in relation to the history of economic thought, another hermeneutic framework that makes it possible to understand and to analyze economic theories otherwise, and on the basis of divisions other than those traditionally accepted.
53A demonstration of the conjoint impossibility of determinism and relativism, the necessity of the freedom of agents, the comprehension of their capacity for innovation, a new framework for reading to understand the authors of the economic tradition—these are the four outcomes of our analysis of economic science on the basis of the transcendental argument.
We mean “cutting across” [recoupements] in the sense that, as Alain Leroux and Alain Marciano write, “economists and philosophers practicing their own disciplines discover common problems” (1998, 5)—not in the sense of deductive links, where one of the disciplines would imply the other.
This formula of “scientificity in general” may be stigmatized—by disciples of Quine, for instance—as harking back to the old dream of an external, underlying epistemology that defines science in general in abstracto. In fact, by “scientificity” we refer to the straightforward usage in which we distinguish, in a set of statements, those with a scientific claim and those with a claim that is other than scientific, for example literary (Apollinaire does not have the same claims as Newton, or indeed Wittgenstein). We thus define science or literature on the basis of the notion of “claims to . . .” inscribed within the statements that they produce. Consequently, to speak of “scientificity in general” here means only to observe that there are discourses that make a “claim to” science rather than to literature, and are received and discussed in light of this claim.
Wittgenstein, omnipresent in the philosophical field, is also used by certain economists, for example in the economics of conventions.
We might think here of Commons, to whom we shall return later.
For example, G. Hodgson, to whom we shall return; and, as a general rule, evolutionists (for biology), and H. Simon (for artificial intelligence models).
See Mongin (2000) for a presentation of different epistemological interpretations of Friedman’s positions.
See, for example, Cournot: “We are here to observe and not to critique the irresistible laws of nature” (Cournot 1838, 9).
In Kant’s sense of the use of the “as if [als ob]” in the Critique of Judgment, which aims to define a judgment that is other than purely determinant, that is to say other than solely physical or mathematical.
For example, Sugden writes: “I believe that each of these models tells us something important and true about the real world. My object is to discover just what these models do tell us about the world, and how they do it” (2002, 108).
We allude here to the expression used by Coase, who ceaselessly critiqued the absence of any taking into account of reality on the part of certain economists: “The dominant economics . . . is a view disdainful of what happens in the real world” (Coase 2000, 4).
Three definitions of the notion of value are studied by Ochangco: Smith’s thesis, which turns entirely on a description of reality (first language game); Ricardo’s, which seeks to construct a theory (second language game); and Marx’s, which seeks to construct a critical analysis (third language game).
For Lakatos, the scientific approach consists in positing a research program whose fundamental hypotheses (“the hard kernel”) will be momentarily posited as unfalsifiable. The auxiliary hypotheses (“the protective belt”) will become refutable and a consequent research program will have to render these hypotheses refutable (1970, 62–70).
We pass over this well-known argument hastily. What interests us here is the continuous passage from logical positivism to the most radical relativism. As I. Stengers emphasizes: “Just before his death, Lakatos himself wound up recognizing that the judgment of the ‘man of science’ could only take place retroactively. It is we who did not know that this or that program was degenerating. But in this case, it is history itself that gives philosophers the power to judge, to determine ‘at what point’ it was rational to abandon this program for another. And this power, conferred by history, is in fact redundant: philosophers confirm the ‘losers’ that they have indeed lost, but they have no proper resource for evaluating and judging the reason for which these losers hung on to their program: they can only say that history has not retained these reasons” (1993, 44).
Thus, for Friedman, companies are represented by a firm whose function of production consists in maximizing results, taking account of the constraints of production. Enterprises that do not seek to maximize their profits will tend to disappear: “Unless the behavior of businessmen in some way or another approximated behavior consistent with the maximization of returns, it seems unlikely that they would remain in business for long” (1953, 16).
For a critique of this representative agent, see, among others, Kirman (1992).
One of the most cited articles in economics on this type of problem is Akerlof (1970), in which the author shows the difficulty of using price as a signal of information on the secondhand car market, where the essential problem is the difficulty for the buyer of knowing the quality of the car.
For example, recourse to a third party for the observation of behaviors, to experts, etc.
North (1990, 2005) distinguishes, for analytical simplicity, on one hand the rules of the game (the institutions) and, on the other, the players (organizations (enterprises, syndicates, political parties, etc.) and individual agents). On this point, see D. Chabaud, C. Parthenay, and Y. Perez (2005).
Thus, Simon, indubitably the most cited author in economics when one is considering the hypothesis of limited rationality, affirms that his central interest is human reason, an approach that gives rise to a philosophical approach: “The nature of human reason—its mechanisms, its effects, and its consequences for the human condition—has been my central preoccupation for nearly fifty years” (Simon 1983, vii).
Here it is a question of bringing to light tendencies (determinism or skepticism) toward which an economic theory tends: implications that are not necessarily intended or affirmed as such, but which are possibly contained as consequences of a position.
These systems can be natural (an ecosystem, the structure of the brain, etc.) or social (an economical system, a social organization, etc.).
For Hodgson, habits are “a propensity to behave in a certain way when faced with a particular class of situations” (2001, 290).
Beginning with Quine, “philosophical naturalism comes down to a position that seeks to explain thought in terms of natural science, such as psychology, biology, or even physics” (D. Fisette and P. Poirier 2000, 100); on the different versions of naturalism (ontological, epistemological, anthropological), see R. Ogien (2003).
Hodgson does not deny that individuals have a deliberative capacity, but he aims to make this, in the last instance, the expression of a natural causality: “The explanations of human motivations must pass by way of biology, psychology, anthropology, or other social sciences” (Hodgson 2004, 57); the whole set of these sciences referring to universal Darwinism. Hodgson here comes close to a Spinozist configuration that postulates that it is through the absence of the knowledge of complex causalities that we believe in a freedom of choice. Like all phenomena, deliberative capacity is explained by necessitating laws.
Here we follow the remarkable interpretation of Commons’s oeuvre by Bazzoli (1999).
For a more detailed analysis of the possible degradations of pragmatism, up to and including Rorty’s position, see Thomas-Fogiel (2005, chap. 1). We do not say, obviously, that Commons would accept Rorty’s position—we sum up here in the form of a question an analysis made at length in the book cited above on the “consequences of pragmatism.”
This problematic juxtaposition of two types of economy (one detailing rules, the other accepting the perturbing variable of the freedom of agents) appears very early on in economics—for example it is Cournot (1838, 12) who divides economics into scientific economics, which does not concern itself with the real behavior of agents, and social economics, which takes into account the fact that the laws of the pure economy can be contradicted by the behavior of agents.
On the debate around the question of whether this argument stricto sensu can be found in Kant, see D. Henrich (1979) and Hintikka (1972).
Which justifies in part the appellation “transcendental,” since, according to the celebrated formula of the Critique of Pure Reason, “I call transcendental any knowledge that deals in general not with objects, but with our mode of knowledge of objects”—except that Strawson depsychologizes the Kantian inquiry. In this acceptation of the term, one no longer thinks in terms of faculties, but in strict terms of conditions of the meaning of a cognitive operation.
We have given a precise illustration of what should be understood by these denominations by evoking the radical positions of Feyerabend and Rorty.
Emphasis in the original.
On this structure, see also Virvidakis (2001).
On this point, among a very abundant literature, see Récanati (1981).
“Necessary compatibility” does not mean that they must be rigorously identical. One can obviously envisage quantitative differences of information, of time, or of tools for reflection, etc. But this means that the rationality imputed to agents must not annul the possibility of a scientific discourse. Now, if we take again the example of individual interest, to reduce everything to the latter is by definition contradictory since, in science, I claim the validity of my discourse, and, through the content of what I say, I say that what I say is nothing other than the reflection of my pure particularity.
Recall that, for Simon, it is a question of thinking human reason as a whole. He therefore does not aim only to determine the possible behavior of the agent in an economic context alone—that is to say, to think only one aspect among others of human behavior. Instead, as is attested by the phrase cited above about reason, he seeks to discover the “mechanisms” of human rationality in general. For, if the economist shuts himself up in the sole domain of economic rationality, affirming that this rationality is not at work in other (moral, political, scientific) domains, then the transcendental argument cannot be used in this way to analyze his propositions. Moreover, the tendency of certain current economists, far from Smith’s prudence, is to extend to all of rationality one aspect of it, for example the calculation of interest (reason=calculation of interest). If economists do not wish to fall victim to the present argument, they must reduce their domain—but then they will need to confront other epistemological questions, such as that of the compartmentalization of behaviors (an economic behavior in context z but a moral behavior in context y)—a problem that has to be faced up to but which is not the subject of the present article.
The word satisficing is a neologism introduced by Simon: “Since there did not seem to be any word in English for decision methods that look for good or satisfactory solutions instead of optimal ones, some years ago I introduced the term ‘satisficing’ to refer to such procedures” (1996, 119). On this point in Simon, see Parthenay (2005).
On artificial intelligence programs, Simon writes: “As primitive as these programs of understanding may be, they truly provide a set of basic mechanisms, a theory, to explain how human beings are capable of understanding problems, at once in the new domains in which they know nothing and in domains in which they have a greater or lesser amount of anterior semantic knowledge” (Simon 1969/1996, 179).
We use the term “representation” in its most current sense. “Representation” here is synonymous with the set of attitudes by way of which an individual believes something, thinks something else, knows something else, etc. In short, the conceptions, readings, or usages he has of the world at a given moment. This “world” must itself be understood in a very broad sense, as the natural and technological environment, but also as the set of social conventions or rules.
Carnegie-Mellon is the university where Simon pursued his research, especially around artificial intelligence.
Induction is a procedure that is already questionable, as everyone knows—but it is all the more so when it becomes a wager that consists in saying that an experiment, which has not yet been done and which we do not know whether it ever will be, proves a current hypothesis. It is to highlight this paradoxical character the too-frequent utilization of recourse to the future, that we use the expression, “induction on the basis of the future.”
We do not claim that these are the only missing links of the chain of his theory that Simon may recognize. The question of the determination of aims and values also remains suspended within his apparatus. If rationality is a procedure to attain an aim, who decides the content of that aim—for example, Nazism (see on this question, Simon 1983), or perpetual peace among all peoples? It is not the shortcoming of the theory that matters to us here, but the contradiction that appears at the level of the discourse of justification.
See the definition of naturalism given above. On its importance today, see Ogien (2003, 128) who notes with irony: “There is probably not a week that goes by without a scientist or a philosopher proposing to naturalize something: ontology, intentionality, meaning, epistemology, ethics, or the normative in general (for his colleagues). . . . It is only naturalization itself that seems to have escaped, up until now, the ardor of the naturalists!” If we take up these terms, our argument consists in saying that, precisely, naturalization cannot itself be naturalized; consequently, it cannot be complete. Naturalism in its strict version falls under the transcendental argument.
In terms of game theory, if the rules are equilibria (in the sense of a regularity of behavior), they can only be considered as Nash equilibria if we give the most minimal definition of this equilibrium—namely, the fact that an agent has no interest in just modifying his behavior. It is thus not a matter of considering or of thinking these equilibria as pareto-optimal (on this point, see Aoki 2001).
It goes without saying that the transcendental argument does not prohibit this type of argumentative strategy. It is situated at another level, one that is not in itself, certainly, incompatible with the appeal to experience.