206 ALTERNATIVE VIEWS OF HERMENEUTICS

coherence approach as Timo Airaksinen’s fairly typical definition of the coherence theory demonstrates:

First S must collect a set of data which includes all relevant ‘information’ concerning his target-propositions. Then S applies some special reasoning methods to his data: S takes a consistent or inconsistent D [set of data], forms its maximal consistent subsets and derives the logical consequences from one or more suitably selected subsets.

(Airaksinen 1978, pp. 265–6)

Robert Lucas, a leading RE theorist, explains how a coherence-type approach could be applied. He suggests that we think of rational expectations as one stage of an ongoing process of adaptive behaviour that constitutes a ‘trial-and-error process through which our modes of behavior are determined’ (Lucas 1987, p. 217). According to Lucas, adaptive and RE theories are different but not mutually exclusive.

Lucas notes that no currently available economic theory, including RE, provides a useful basis for modelling adaptive processes. Instead, economic models start with ‘decision rules that are steady states of some adaptive process, decision rules that are found to work over a range of situations and hence are no longer revised appreciably as more experience accumulates’ (Lucas 1987, p. 218). Economists and agents in RE models could view the coherence standard as an adaptive context within which rational expectations are made. Lucas’ reference to this context as a ‘steady state’ condition seems to come directly from the type of systems theories that coherentists such as Rescher refer to. The standard of a ‘steady state’ together with the related criterion of ‘ordered wholeness’ are often used to describe systems which have reached a high degree of coherence.24

In addition to his implicit references to a coherence-type standard Lucas also notes that adaptive processes produce pragmatic rules which ‘work over a range of situations’. He views adaptive expectations as the context within which expectations are judged to be rational. As I indicate below, Rescher and other advocates of the coherence theory adopt a similar approach.

Problems with coherence theories

The coherence method for terminating the regress of justifications is in many ways an improvement over foundationalism, but numerous criticisms have been raised against it. These objections include: (1) the lack of empirical inputs; (2) inter-system indeterminacy; (3) ambiguous coherence criteria; (4) vicious circularity; and (5) a problematic view of truth. If substantiated, these objections would undermine the coherence defence of rationality. The following sections contain a brief review of these problems.

THE ECONOMICS OF RATIONALITY 207

The lack of empirical inputs

The ‘lack of empirical inputs’ objection is based on the claim that the coherence standard is detached from reality. A standard of justification based exclusively on coherence seems to ignore the fact that at least some empirical beliefs must be justified—not because of their relation to other beliefs but because of their source, an ‘outside world’. For example, without empirical inputs it may be possible for agents in an RE model to form a highly coherent set of fictional beliefs that is totally unrelated to the real economy. Several philosophers consider this type of objection to be the primary attack on coherence theories (Dancy 1985, p. 120; BonJour 1985, p. 111).25 As Rorty has observed: ‘Holistic theories seem to license everyone to construct his own little whole—his own little paradigm, his own little practice, his own litte language-game—and then crawl into it’ (Rorty 1979, p. 317).

In answer to this objection several advocates of coherence theories have agreed that coherence theories must include some empirical inputs.26 This requirement appears similar to the foundationalist rule of privileged justification, but it is also consistent with the coherence approach. ‘It is important to stress that the need for inputs is inevitable in the realm of factual knowledge; but this unavoidable recourse is not a concession to foundationalism. The coherentist’s inputs are raw materials and not themselves finished products’ (Rescher 1979, p. 69). Like Rescher, BonJour admits that empirical inputs are required: ‘it is possible to identify certain beliefs as cognitively spontaneous; and also to determine that certain classes of such spontaneous beliefs are, as judged from within the system, reliable, that is, likely to be true’ (BonJour 1985, p. 138). He notes that coherence theories, in fact, require such inputs (BonJour 1985, p. 141).

Although beliefs based on empirical inputs are thus ‘privileged’—in the sense that they cannot be completely ignored—coherentists point out that the process of justifying them is no different from that used to justify other beliefs. These inputs, which BonJour calls cognitively spontaneous, are not afforded the special role of foundationalist basic beliefs, but are instead tested like other beliefs. The general test is whether they increase system coherence. Instead of being ‘immediately justified’ they are assumed to be plausible candidates for justified beliefs, where the degree of plausibility is determined in part by considering whether beliefs of a similar ‘type’ have historically added to the system’s coherence. Ultimately, empirical beliefs as a class are required because cognitive systems exist largely to help us to interact successfully with an empirical world. The important point is that, even though in principle the cognitive system must include some empirical inputs, individual empirical beliefs are not justified because of their source but because they enhance the system’s coherence.

208 ALTERNATIVE VIEWS OF HERMENEUTICS

Inter-system indeterminacy

A second objection to the coherence theory is that it does not provide an adequate standard for choosing between competing coherence systems.27 Critics claim that the coherence standard does not guarantee that a single system will be optimally coherent (Sosa 1980, p. 557). The coherence theory must therefore not only provide a standard for deciding which beliefs should be included in cognitive systems, it must also address the possibility of having to choose between a variety of equally coherent but mutually inconsistent systems.

The justification of individual beliefs is ambiguous when the choice of cognitive systems is indeterminate. If economists and RE agents cannot rationally choose between different sets of expectation-formation approaches that contain equally coherent sets of equations, then they cannot justify their expectations. Stating this argument somewhat differently, since it is conceivable that any belief could be a member of some coherent system, the inability to decide between competing systems appears to make all beliefs equally justified and thereby forces the coherentist to accept relativism and scepticism.

In addition to the more obvious problems of system indeterminacy, the challenge also contains an implicit presupposition which, when questioned, exposes further difficulties. The indeterminacy objection assumes that the choice between competing holistic systems occurs outside the context of the competing systems. This assumption is subject to the same arguments made against the foundationalist ‘privileged justification rule’. Foundationalist theories have been unable to justify basic beliefs without relying on other beliefs that must be defended. This problem creates a dilemma for coherentists who must choose between two or more competing systems.

On the one hand, if the context for choosing between competing theories is assumed to be the correct foundation for making such choices, then it is subject to the criticisms that have already been raised against basic beliefs. On the other hand, if the choice of evaluative contexts is justified by appealing to another coherence standard, then the infinite regress problem cannot be solved. Any ‘meta-system’ used to justify the choice between cognitive systems would face the problem of indeterminacy. The attempt to justify a meta-system by appealing to yet another context-based standard would initiate a regress of justifications.

Because foundational and coherence theories face a similar difficulty, the problem of system indeterminacy cannot be used as an argument favouring the foundationalist view over the coherence approach. This is a weak defence, leaving both theories open to the charge of relativism and scepticism.

Ambiguous coherence criteria

The coherence standard is generally recognized as a measure of how well a system ‘hangs together’ but there is no consensus on the exact meaning of the term ‘coherence’. In addition to the obvious standard of logical consistency,

THE ECONOMICS OF RATIONALITY 209

coherence theories typically include criteria such as comprehensiveness, cohesiveness, uniformity, simplicity, and regularity.28 Rather than specifying a single coherence rule, coherence theories usually include a combination of rules that act as the standard of system coherence.29 This can lead to difficulties when the importance of two or more measures of coherence must be compared. Rescher, for example, notes that individual measures or ‘desiderata’ may conflict with each other in concrete cases. Rather than rejecting one of these rules he suggests a pluralistic approach. ‘The pluralism of desiderata—the fact that each must be taken in context of others within the overall picture of systematicity— means that in the pursuit of these factors we must moderate them to one another’ (Rescher 1979, pp. 16–17). For example, in the tension generated between coherence rules, some degree of simplicity may have to be sacrificed for a gain in uniformity. RE agents might have to weigh the loss of completeness associated with one set of equations against the greater accuracy of another set that is less comprehensive. Rescher even admits that, while no one wants an inconsistent belief system for its own sake, it may be advisable to accept some measure of inconsistency in exchange for an increase in the degree of completeness (Rescher 1979, pp. 176–7).30

The solution to this problem requires the introduction of coherence ‘metacriteria’ which regulate the use of individual rules. The meta-criteria can be thought of as a special class of beliefs—namely ‘epistemic beliefs’ which define what constitutes a properly constructed cognitive system.31 Epistemic beliefs could determine the weights assigned to various coherence criteria—such as completeness, consistency, etc. The correct assignment of weights is important because a change in the relative significance of one of the criteria could affect the system’s coherence. As with other beliefs, these must pass the coherence test of epistemic justification.

Rescher argues that epistemic beliefs can ‘break the regress of justifying theses by theses’ because they are ‘justified by reference to certain practical criteria (preeminently, success in prediction and efficacy in control)’ (Rescher 1979, p. 93). He suggests that the coherence standard could be justified by using a pragmatic cost-benefit analysis to determine the ‘tradeoff’ between employing various measures of coherence (Rescher 1979, pp. 11–12).32 This approach does not solve the regress problem. Rescher’s cost-benefit standard cannot be justified on the basis of its ability to increase system coherence because it also used to determine the meaning of coherence.33 In his article on the economic theory of cost-benefit analysis Israel Kirzner notes: ‘No matter how calculative a man’s behavior may be, it seems impossible to avoid having accepted, without calculation, some framework within which to self-consciously engage in costbenefit comparisons’ (Kirzner 1982a, pp. 144–5).34

By relying on a methodology such as the cost-benefit analysis, coherentists are forced to adopt an approach that is at odds with the coherence strategy. As Cornman has noted, some form of foundationalist theory is necessary to show that probabilities or something else should be used to establish the maximally