1. Introduction

Such are the last sentences of Michael Friedman's Kant Lectures published as Part I of the fittingly titled monograph Dynamics of Reason. They make admirably clear what is at stake for Friedman: he sees it as philosophy's responsibility to speak up for reason and to uphold Enlightenment values in the face of a “widespread relativistic tide” (Friedman Reference Friedman2001, 57). In mounting his sophisticated defense of rationality, he tries to offer a way out of the incommensurability conundrum bequeathed to philosophy of science by the work of Thomas Kuhn, but he also explicitly denies that his specific way of defending the rationality of paradigm changes would commit him to scientific realism (117). This denial might be seen as creating an uneasy tension in Friedman's position, though (as argued, e.g., by Slowik [Reference Slowik2006]), since he stresses the convergence of succeeding paradigms as an essential part of the paradigm-transcending rationality. But belief in such convergence is something that is typically underwritten by realist commitments: why would one expect it, if not because one believes that our scientific methods allow us to learn something about the true structure of the world? And it is a striking feature of Friedman's presentation that he actually doesn’t seem to offer much of an argument for this convergence: he only introduces the historical fact that it is possible to notice a kind of convergence from the Aristotelian to the Newtonian to the Einsteinian space-time theories, which in itself obviously cannot suffice to believe that this will continue to hold true. So, is he implicitly endorsing scientific realism in stipulating the continuing convergence? And if he is not, what sense can we make of his stress on the necessary convergence? A careful consideration of these questions might help us to reach a better understanding of the precise nature of Friedman's defense of universal rationality. It will turn out that we should take his summary in the opening quote more seriously than most people might be tempted to do.

2. The Relative A Priori and the Dynamics of Reason

Friedman is probably best known for his forceful reintroduction of the notion of constitutive, a priori principles in contemporary philosophy of science. Taking his lead from Hans Reichenbach's early work, he argues that Kant's views on the a priori prerequisites for natural science still hold promise if we distinguish between the constitutive role of a priori principles, on the one hand, and their supposed unrevisability on the other. He stresses that a detailed analysis of the structure of Newtonian mechanics and special and general relativity shows that we should discern three different layers within these theories. On a first level we have purely mathematical structure (e.g., Euclidean geometry and calculus for Newton's theory); on a second level we have a special class of principles that allow the coordination of mathematical structures with empirical phenomena (e.g., Newton's three laws of motion, which allow one to identify the inertial frames of reference in which one can then first unambiguously relate mathematically characterized paths of motion to their empirical counterparts); and on a third level we have the theory's properly speaking empirical laws (e.g., the law of universal gravitation). The most important point about this layered structure is that the different levels have an epistemologically asymmetrical function: the first two levels allow us to ascribe truth values to the empirical claims of the theories, and thus are constitutive with respect to the third level. This asymmetry also implies that it makes no sense to think that the constitutive principles themselves could be tested empirically, since all meaningful empirical tests must already presuppose them. But this does not imply that for that reason they would be unrevisable, as the historical progression from the Newtonian to the Einsteinian space-time frameworks forcefully makes clear.

This does imply, however, that we are once again confronted with the specter of radical incommensurability, as it is no longer clear how we can rationally decide to switch constitutive principles: it seems that these principles in the first place make possible rational agreement because they first constitute what Friedman calls “the empirical space of reasons” (Reference Friedman2001, 85). But this problem, which arises exactly because we can no longer hold on to Kant's belief in the absolute unrevisability of constitutive principles, can be overcome by paying attention to another distinctively Kantian insight, Friedman claims. Kant's keen analysis of the crucial role of constitutive principles was only possible because of his specific understanding of the task of philosophical analysis vis-à-vis scientific practice. He saw that the most fruitful position for philosophy was that of a “transcendental” metadiscipline that critically investigates the presuppositions of science. Putting metaphysics on the sure path of science implied giving up the false idea that philosophical analysis could proceed independently from science to give the latter its true foundations, but also resisting the idea that it had to be conceived as an extension of scientific insights to a further, supposedly metaphysical domain. And it is exactly philosophy considered along these lines that historically provided the resources to make paradigm change intelligible and made it possible to switch constitutive frameworks through a process of consensus-creating reasoned argumentation.

Friedman argues for the latter point along two lines, with the first taking clear predominance: he first presents a subtle historical narrative, and he then gestures toward a philosophical argument. The historical story explains in convincing detail how Einstein's revolutionary moves were only possible because of his engagement with the philosophical debates on space-time that crucially included Helmholtz and Poincaré. And these philosophical debates in turn were a direct outcome of a critical investigation of Kant's transcendental analysis of the conditions of possibility of (applied) mathematics. Side by side with the successful application of the Newtonian framework to ever more empirical phenomena (the Kuhnian process of normal science) was developing a lively dialogue on the quid juris of this framework, a dialogue that was initiated by Kant's own analyses but further informed by developments internal to these mathematical and empirical sciences (just as Kant's views were informed by the state of these sciences at his time). And it is these philosophical reflections that as “philosophical meta-paradigms or meta-frameworks” play “an indispensable role, by serving as source of suggestions and guidance—for orientations, as it were—in motivating and sustaining the transition from one paradigm or conceptual framework to another” (Friedman Reference Friedman2001, 46). Einstein could only have made his revolutionary moves because he operated against this level of philosophical metareflection; and exactly because his moves had this background, they had to be taken serious and considered as live options in the science of his day. In the place of an empirical testing of the constitutive principles comes a sustained philosophical inquiry of their right to claim special status, which in the right circumstances (Kuhnian crisis) can help open up the conceptual space for the crafting of a new “empirical space of reasons.” Kuhn's views on history of science only raise the specter of irrationality because they unduly neglect the parallel evolutions in history of philosophy.

In arguing for this point of view, Friedman introduces the notion of “communicative rationality,” which he takes over from Jürgen Habermas (Friedman Reference Friedman2001, 53ff.). This notion is intended as an articulation of the Enlightenment faith in human rationality, by offering the ideal of “non-coercively uniting, consensus creating power of argumentative speech” (Habermas, quoted in Friedman Reference Friedman2001, 54). And this is exactly the kind of rationality that not only characterizes normal science (due to a shared empirical space of reasons) but also paradigm change (due to the presence of philosophical metaparadigms sustaining well-articulated “argumentative speech”). Conclusion: we can still uphold our best mathematical sciences as exemplars of human rationality.

It is after having added historical evidence that we can indeed describe paradigm changes as communicatively rational that Friedman adds that “from a philosophical point of view,” “we can thus view the evolution of succeeding paradigms or frameworks as a convergent series, as it were, in which we successively refine our constitutive principles in the direction of ever greater generality and adequacy” (Reference Friedman2001, 63). This claim is backed up by historical evidence (the possibility of retrospectively approximating features of the old framework within the new one, and the possibility of “prospectively” seeing the new framework as arising out of the old through a continuous development—a continuity only made possible by the philosophical metaframework, of course), but as already indicated in the introduction, its scope is actually much wider than can be established by purely historical evidence, and the latter must thus be supplemented by philosophical argumentation—as Friedman himself indicates by introducing it as following “from a philosophical point of view.” This is further underscored by his switch to explicitly normative language in repeatedly stating that we should always be able to view paradigm change in this way. These strong statements are backed up by an appeal to Kant's conception of the regulative use of reason, which offers an ideally completed state of scientific progress as a focus imaginarius guiding all empirical research (see also Section 3.2 below), but which Friedman now extends to also make possible “trans-historical universality” within the domain of shifting constitutive principles. This idea is inspired by Cassirer's way of rethinking Kant's transcendental idealism in the light of revolutionary changes, but Friedman stresses that, more than Cassirer, he still wants to leave room for relativized constitutive principles within this ideal progression (Friedman Reference Friedman2001, 66, note 80)—providing temporary points of rest, as it were.

But why exactly do we need this regulative use of reason in our defense of rationality? At this point Friedman is disappointingly vague. It would seem that the appeal to philosophical metaparadigms already does the job of securing communicative rationality, in offering the possibility of coming to reasoned intersubjective agreement. Why does he want to add an appeal to an ideal of reason as giving extra, “transhistorical” direction to this process of argumentative deliberation? I think that a glimpse of the largely hidden motivation can be seen in the way Friedman introduces a summary of his views with the question, “how … can the proposal of a radically new conceptual framework be … both rational and responsible?” (Reference Friedman2001, 66; my emphasis). But by thus indicating that reasoned intersubjective agreement might not be enough for fully universal rationality, this extra appeal to “responsibility” could lead us back to the suspicion that this hidden motivation is actually one of a kind with scientific realism—as realists typically portray it as the responsibility of scientists to uncover the world's true structure, which true structure would then ground the convergence of scientific theories. As already mentioned, it is clear that Friedman himself does not intend his argument in this way, as he explicitly warns us that his views on the necessary convergence of successive frameworks must not be understood as “ever better approximations to a radically external world existing independently of the scientific enterprise itself,” since the “original Kantian conception of objectivity … was explicitly intended to undermine such a naively realistic interpretation of scientific knowledge” (Reference Friedman2001, 67). If we thus want to come to a better understanding of Friedman's possible agenda in stressing the need of “responsible” paradigm change—and especially its relation to the series of converging frameworks—we must start by taking a closer look at this original Kantian conception of objectivity and its possible relation to present-day discussions on scientific realism. This will prove to be especially interesting because we will see that Kant's own appeal to the regulative use of reason is closely connected to some of the issues that are at stake in these discussions.

3. Transcendental Idealism and Scientific Realism

In the following I will propose a reading of Kant's transcendental idealism that has to meet two constraints: it should be formulated at a high enough level of abstraction that it also can be used in helping to characterize Friedman's position (this implies remaining as uncommitted as possible on the vexed issue of the status of a priori forms of sensibility—Friedman definitely owes us further clarifications about the position such forms occupy in his modified Kantianism), and it should be a reading that fits Friedman's claim that being Kantian implies not being a scientific realist.

4. Reason's Responsibility

In sketching this background to Friedman's denial of scientific realism, we have actually gathered the elements that will now allow us to diagnose what seems to be a fatal ambiguity in his characterization of interparadigmatic rationality through the idea of a convergent series of constitutive frameworks, guided by a regulative ideal of reason. It is not that such an idea would commit him to realism, as his reference to the Kantian pedigree of his position is indeed enough to ward off such suspicion. It is rather the idea itself of having a series of constitutive principles being guided by a regulative ideal that does not make sense when seen from within this Kantian perspective.

We have seen that in Kant's system, the regulative ideal of reason is indeed genuinely normative—and thus a candidate to ground communicative rationality—but we have seen that it is also dependent on a constitutive framework within which the ideal can further guide the understanding. In Cassirer's modified version of Kantianism, which Friedman cites as the important source of inspiration for his own views, this dependency is exactly inversed. The regulative ideal of systematicity takes absolute precedence, and a constitutive framework can only be defined as the set of principles which turn out to be preserved in the progress from theory to theory toward the projected complete system, and which as such are the conditions of every possible theory. But what about Friedman's proposal? At any stage of development in a mature physical theory we do have a set of constitutive principles that determine the space of reasons and allow us to form truth-valued statements about an objective reality. Within any such framework it makes sense to project the ideal of nature as a fully unified system, but what kind of unified world could be “transhistorically” projected to span all these frameworks? Empirical truth makes sense only within a constitutive framework (it is this possibility that grounds the latter's special status), so speaking about the truth of the changing frameworks is precisely what cannot make sense (also according to Friedman [Reference Friedman2001, 118])—but as my exposition in Section 3.2 should have made clear, this also takes away the ground under the transcendental justification of the normative status of the Kantian regulative ideal. The basic problem is that both Kant's and Cassirer's regulative ideal plays a role in (further) articulating what is ultimately one empirical space of reasons, and that it is not clear what sense it makes to speak about “convergence” of synthetic structures in the absence of such a unitary empirical world—whether this would be interpreted realistically or transcendentally. I already noted in Section 2 that Friedman gestures toward an argument rather than really providing one when introducing his normative remarks on the convergence of frameworks. It now seems that he actually gestures in the wrong direction.

It is probably for this reason that, notwithstanding his stress on the convergence of frameworks, Friedman at one point also speaks about “approximation to a final, ideal community of inquiry” (Friedman Reference Friedman2001, 64; my emphasis). Such an ideal community would have “achieved a universal, trans-historical communicative rationality on the basis of … fully general and adequate constitutive principles” (64)—but securing this kind of rationality is of course exactly what Kant himself believed he had achieved!

I suggest that we thus had better understand Friedman's view on what can secure the ultimate rationality of paradigm change as a very serious call to uphold the full promises of the Kantian philosophical project itself—but as exactly that: a project. The only reason we have for positing convergence comes from the fact that it always is (and arguably must remain) a projected ideal: it is not a point we need to believe will ever be reached by any human community of inquiry (and even less must we believe it to be only possible provided we would have grounds to believe that we can be successful in latching onto reality)—it is just that which we as rational inquirers must see ourselves as striving toward. So even if we are doubtful that we will ever reach this ideal, this need not imply that we should not hold this up as our investigative norm.

In a sense one could say that Friedman very seriously wants us to try to become Kant again: an impossible but maybe lofty dream!Footnote ² And this makes good sense of his appeal to the necessary role of philosophy in making possible revolutionary but rational change, as this role is actually a modification of Kant's original transcendental deduction into a necessarily inconclusive transcendental reflection. And as such reflection cannot but start from the (cognitive) situation we are in (as was already the case for Kant), it also makes good sense of the requirement that a new constitutive framework should necessarily evolve continuously from the older one. And, finally, there is still a way in which we could see this as “progress”: we do actually learn something—not because we gain insight into the empirical structure of reality, and even less because we would gain insight into the transcendent structure of reality, but because we gain insight into how we do, and can, cognitively approach empirical reality.

In this sense paradigm change is both less and more than communicatively rational. It is less, because it is not objectivity-constituting (which it seems to be for Habermas, at least to be judged by the quote in Friedman Reference Friedman2001, 54, and which it certainly would have to be if it were linked to Kant's or Cassirer's regulative ideal); but it is more, because it does exemplify another deep feature of human rationality. We saw in Section 3.1 how a transcendental perspective unavoidably brings with it a distinction between things as they appear and things as they are in themselves. This was of the utmost importance to the Kantian project in that it curbs the pretensions of reason, and thus actually confronts it with its own responsibility—which is to grasp its (cognitive) autonomy. And this finally brings us back to Friedman's curious invocation of responsibility besides (communicative) rationality. What the possibility of paradigm change and the concomitant historicization of reason shows is that the unavoidable contingency that characterizes our approach to reality runs even deeper than Kant could see: it not only depends on our being the kind of human beings we are, but also on the historical situation we always necessarily find ourselves in. But this only further highlights that we have no choice but to take full responsibility—which now also implies that we must be ready to question even our most cherished basic framework when a situation demands it. And it is in this sense that we should see ourselves as approaching, and as necessarily only approaching, the ideal Kantian situation: because only thus can we fully uphold the ethical-cognitive ideal that is part and parcel of the transcendental idealist stance. Only thus can “reason grow increasingly self-conscious and thereby take responsibility for itself.”