The Relative/Absolute Nature of Knowledge
and Economic Analysis
The Economic Journal
Volume 97, Issue 388
Dec. 1987, 951-970.
Central to economic analysis are assumptions and hypotheses concerning the extent and nature of the knowledge held by individuals. In this respect it is a noticeable feature of realist 1 accounts that they tend to focus upon the pervasiveness of uncertainty as an undeniable fact of economic life. (Non-realist, or instrumentalist accounts, in contrast, tend to assume that agents have perfect information, correct foresight, rational expectations, and so on.) It is also apparent that this realist focus upon uncertainty tends to give rise to a concentration of attention upon those situations in which (possibly numerically measurable) probabilistic knowledge is thought to be available. Clearly such accounts are important. Nevertheless, I wish to suggest that such a restricted choice of focus for the realist is not entirely satisfactory, since, in particular, it tends to lead to a neglect of the widespread observation that feelings of certainty and claims to truth are also pervasive in human society. I want to suggest, moreover, that if we are to account for the latter observation (and economists concerned with realism of analysis surely need to) then various arguments and conclusions have to be taken on board which, in turn, appear to bear consequentially upon the nature and conduct of economic analysis in general.
Section I below discusses such possible limitations of existing realist accounts of agent knowledge in some detail. The suggested route for responding to, and resolving, apparent problems - which involves the argument that aspects of knowledge can have a relative as well as an absolute character - is described in sections II and III; and the various analytical consequences of the account there set out are discussed in section IV. Section V contains concluding comments and touches upon ways in which the proposed account might be appraised and made operational.
It is not unusual for economists concerned with realist analysis, especially those concerned with matters of economic policy formulation, to defend their own beliefs as truths held with certainty. Indeed they are sometimes explicitly
1. Realism involves a commitment to a view of the actual (or at least possible) existence of some disputed entity. In the context of economic analysis I take realism to be the doctrine that the objects of research analysis exist independently (at least in part) of the enquiry of which they are the objects. As such realism incorporates the belief that there is an objective world which exists independently (in part) of consciousness but which is knowable by consciousness. This view contrasts with instrumentalism (among other doctrines) in which the objects of enquiry tend to be treated as mere ‘convenient fictions’ or some such. Realist explanations or ‘models’ postulate entities on grounds additional to analytical convenience and involve a commitment to their further/continuous scrutiny (see Lawson, 1987b).
criticised for doing so. Thus Hahn (1985, p. 20) asserts that the ‘display of certainty by many economists in discussing matters of economic policy is not only lacking in honesty but extremely harmful to the subject’. No doubt there are situations in which this ‘display of certainty’ is partly ‘bluff’ designed, perhaps, to intimidate opponents into agreement. Nevertheless it is difficult to attribute this pervasive ‘display’ entirely to dishonesty. Moreover, such certainty of belief is hardly confined to economists. Indeed it is found in all walks of life, as a moment’s reflection should confirm. In general, people holding theories or views about the world and society in which they live, about matters which affect them, do seem to hold on to them as if they represent the truth. Certainty of belief is not a peculiarity of economists.
It is noticeable, however, that economists who are concerned with realism of analysis, and who focus explicitly on the knowledge that individuals possess, tend to emphasise almost exclusively the fact of uncertainty surrounding economic activity. Now it is the case, of course, that uncertainty is a pervasive fact of life. Nevertheless, an over-preoccupation with emphasising and describing it can mask the fact that certainty is also a pervasive feature of human agency. This emphasis on uncertainty and consequential neglect of certainty and true belief, in fact, may be a reaction to the widespread supposition in economic analysis that situations of certainty and truth must somehow involve complete or perfect information. Certainly an assumption that is often made in orthodox economic analysis is that agents do indeed have perfect information or, if prediction is the aim, correct foresight. It is not clear, in fact, that such analyses ever take realism as their aim; but, in any case, the typical response by realist minded economists is to reject the possibility of perfect information, to emphasise instead the fact of uncertainty and, in so doing, usually to concentrate upon theories of probabilistic knowledge that do not appear to presume so much by way of agent information.
One such probabilistic account which seems to be increasingly, if belatedly, attracting the attention of economists (see, e.g. contributions and references in Lawson and Pesaran, 1985) is that set out in a realist vein by Keynes in his A Treatise on Probability, and this is considered in some detail below. Keynes’ theory is concerned with the degree of belief that it is rational for a person to have in a particular proposition, given the evidence available; and, as such, there is an explicitly objectivist or absolute aspect to this account. Perhaps the more widely adopted view of probabilistic knowledge in economics, however, is that provided within ‘subjectivist’ or Bayesian accounts. Although adopting this ‘subjectivist’ label some such accounts often appear to adopt some form of realism as an aim, and even criticise the lack of realism of the more extreme subjectivist analysis (see, e.g. Hey, 1979, p. 232; 1983b, p. 138). In fact, in situations where, in Keynes’ account, a prior probability is considered to be determinate (and possibly in other situations also) the accounts of Keynes and some prominent Bayesians would seem to possess significant similarities. There may be disagreement over whether initial or prior probabilities can be numerically determined in certain situations but (and contra many, more extreme, subjectivists, e.g. Savage, 1954, p. 3), some Bayesians contend, along
with Hey (1983a), that we should ‘expect increasing agreement between people as the amount of shared information increases’ (p.5), a conclusion that appears to be consistent with Keynes’, and indeed with any objectivist analysis. However, for present purposes a point to note is that, for these more realist inclined accounts of probabilistic knowledge, certainty and truth concerning the proposition under consideration only seem to arise when information is somehow complete; that is, the process of revising probabilistic knowledge converges eventually upon certainty and truth only as an ‘end product’ (Hey, 1983a, p.i) or with a ‘completeness of information’ (Keynes, 1973, p. 345). 2
Now, important though such probabilistic accounts certainly are, the fact is that claims to certainty and truth appear to be pervasive in society while, usually, information is incomplete or imperfect. 3 Depending upon what ‘completeness of information’ means here, it seems that, at the very least, something wider or more general than an account of probabilistic knowledge per se is required. A slightly different route seems necessary if both claims to truth and the fact of imperfect information are to be features of realist analysis. Ultimately, in fact, it seems necessary to acknowledge that claims to certainty and truth generally correspond to situations of imperfect information and that, as such, claims to certainty and truth must be fallible and susceptible to possible revision.
The path of questioning explicitly the meaning of terms such as certainty, etc., is one previously embarked upon by Alan Coddington (1982) and, as Coddington makes clear, it is a path which is not without its difficulties. Coddington 4 first focuses upon the ‘ambiguous’ nature of the term certainty. He recognises that certainty must be considered as a ‘state of complete confidence in a belief’ but rejects, as ‘evidently not much use’ to economic theory, the view that this is ‘irrespective of whether that belief is correct’ (p. 483),
2. In A Treatise on Probability, and later in the General Theory, Keynes essentially distinguishes three types of probability-relation: the first where a probability-relation is numerically indeterminate and possibly not even comparable to (in terms of less than, equal to, or greater than) other probability relations; the second where probabilities are numerically determinate but less than unity (and greater than zero); and the third where probabilities take the value of unity (or zero). The first type of probability, for Keynes, corresponds to a situation of uncertainty (see Lawson, 1985) and the third to a situation of certainty. Moving from the first type through the second towards the third, Keynes talks both of the argument in question being less ‘uncertain’ and also of the weight of the argument increasing. For Keynes the weight of an argument appears to be the ‘degree of completeness of the information upon which a probability is based’ (Keynes, 1973, p. 345) or, equivalently, the ‘balance… between the absolute amounts of relevant knowledge and of relevant ignorance respectively’ (Keynes, 1973, p. 77). Thus as relevant evidence increases so too does the weight. Keynes seems to suggest that certainty can only be reached when weight is at its highest, and this in turn appears to correspond to a situation where the relevant evidence is complete.
3. Moreover the assumption of perfect foresight and complete information can be made even in probabilistic accounts. Traditionally this assumption has been employed in the context of a presumed deterministic world. Recently it has also been carried over to a stochastic setting via the rational expectations hypothesis. As numerous commentators have pointed out the most significant presumption of this hypothesis is the claim made about the substantial knowledge supposedly held by individuals. Indeed, while in the traditional perfect-information models it is assumed that individuals know everything in a determinate world, under the rational expectations hypothesis it is assumed that individuals know everything that there is to know in a stochastic world. As such, the latter hypothesis appears to be merely a stochastic euphemism for ‘perfect information’ and just as unrealistic.
4. All undated references to Coddington refer to Coddington (1982).
concluding that certainty can only be a ‘state of complete confidence in a belief, together with the correctness of this belief’ (p. 483). Yet certainty remains a pervasive feature of human agency. Consequently, in order to avoid presuming infallible, omniscient agents, Coddington emphasises that much therefore hinges upon what is meant by the term ‘correct’ in the assumption of ‘correct foresight’. But in rejecting the implication that beliefs and expectations must be correct ‘in every particular, at every instant’ he finds himself in something of a quandary. In fact, having allowed for notions such as correct beliefs or foresight to contain error, Coddington is left with what he takes to be the undesirable conclusion that all beliefs must therefore be correct under some suitably lax definition:
But as soon as it is admitted that any reasonable conception of correct foresight must allow for some (reasonable) margins of error, we are on the beginning of a slippery slope. For we then have to admit that there is no clear dichotomy between certainty and uncertainty (or between knowledge and ignorance, for that matter). Just to emphasize the point, we could go to the other extreme and claim that all foresight is ‘correct’ (to some sufficiently lax standards of approximation), just as all foresight is ‘incorrect’ (by the absurd standards of comprehensive exactness) (p.483).
The sort of reasoning displayed by Coddington in this statement has led others to claim that indeed there is no difference between knowledge and ignorance, etc. Thus, once imperfect information and fallibility are acknowledged, there arises the possibility of an inexorable slide towards what is considered to be the opposite polar assumption to perfect information/foresight (although in some ways it has much in common with it), the view that expectations (and possibly current knowledge also) are merely creative acts of the imagination. In fact some economists embrace this form of extreme subjectivism as the only position which is, in their view, consistent with human freedom. This, however, is not Coddington’s recourse, for Coddington, it seems, wishes to maintain the realist view that knowledge can retain some grip on external reality. In fact, following on from his reasoning as set out above Coddington suggests that we are ‘driven to the notion that correctness must consist in the foresights being within certain “reasonable” bounds of approximation’, but retreats thereby to the view that it is a probabilistic theory of knowledge that is the solution - that ‘the notion of correct foresight is crying out for statistical elaboration’ (p. 484). Once more, clearly, the realist response is to rely upon accounts of probabilistic knowledge.
The pervasive claims to certainty and true belief amongst individuals in society thus appear to present the realist economist with a dilemma. On the one hand, a coupling of this observation with the view that certainty and truth correspond to a situation of complete information raises the unrealistic spectacle of perfectly informed and ultimately infallible omniscient agents. On the other hand, by relaxing the assumption that certainty, etc., correspond to a state of omniscience, there arises the subjectivist alternative that knowledge
is nothing more than creative acts of the imagination. And, as we have seen, the problems inherent in each view seem to have led to a realist neglect of any account of claims to certainty in knowledge, and instead to a restricted emphasis upon situations of uncertainty and a concentration thereby upon accounts of probabilistic knowledge.
I wish to argue that the sort of path embarked upon by Coddington contains the potential for overcoming the problems raised above, but that the fall-back upon a probabilistic view of knowledge is not in itself a complete solution to such problems. In what follows the issues and problems raised will be addressed by following a slightly different route, one which involves the argument that many aspects of knowledge are, in an important sense to be made clear below, both absolute as well as, in some sense, relative. Such attributes, in fact, are often considered to be oppositional. The intention in the following two sections is to argue that they need not be so, and in so doing to suggest how the contention that there is a relative/absolute nature to aspects of knowledge can lead to an avoidance of the apparent analytical problems outlined above. It should be emphasised that the aim is not per se to replace accounts of probabilistic knowledge such as those discussed above (although, the arguments which follow do involve a critical reconstruction of aspects of Keynes’ account). Rather, the intention is to suggest how theories such as these might be supplemented with an account of knowledge revision which has a bearing even when the beliefs involved reveal a (genuine) ‘display of certainty’.
One of the more thorough-going and still relevant accounts of knowledge is provided by Keynes in A Treatise on Probability. (It is indeed much more than an account of probability per se.) Moreover in recent years there has been a significant revival of interest in the epistemological aspects of Keynes’ contribution. 5 As such Keynes’ account of belief and knowledge provides a convenient context in which to develop the arguments I wish to make. This choice of focus is reinforced by the fact that it is Keynes’ account of knowledge that Coddington seizes upon as representing the viewpoint which he is attempting to move away from; a viewpoint which Coddington labels ‘justificationist skepticism’.
The first thing to notice is that Keynes, 6 like Coddington, appreciates that the term ‘certainty’ tends to be used ambiguously (p. 15). Keynes reserves its use to describe the ‘highest degree of rational belief’, which, in fact, can occur only when the proposition in question is correct or true. It also corresponds to knowledge:
knowledge of a proposition always corresponds to certainty of rational belief in it and at the same time to actual truth in the proposition itself. We cannot know a proposition unless it is in fact true (p.11).
5. Thus Harrod’s (1972) comment concerning the Treatise on Probability that ‘Keynes’ work, which is likely in any case to be for long unmatched in its scope and erudition, will remain of living importance as a starting point for discussion, until a more satisfying logical solution of the central problem of human knowledge is found’, (p. 778), seems as relevant as ever.
6. All undated references to Keynes refer to Keynes (1973): A Treatise on Probability.
Keynes and Coddington thus seem to be in general agreement on this point. In fact this observation, in turn, leads us to question why Coddington should refer to Keynes as a sceptic when, apparently, he does not accept the label for himself. Coddington, in fact, criticises a view which conflates knowledge with certainty and ‘then argues, in effect, that since certainty is not attainable, neither is knowledge’ (p. 484). Such a position is indeed one of scepticism, but it is not Keynes’. As we shall see below, Keynes argues that knowledge is obtainable and so is not the sceptic that Coddington suggests.
At this stage, then, it is important to question what it means, within Keynes’ framework, to claim that a proposition is correct or true; what it means to assert that certainty of belief is attainable. Just as Coddington sets out to qualify what is meant by terms such as ‘correct’ so shall I attempt to do the same, but in a way that I believe makes sense of numerous epistemological insights provided by Keynes (as well as Coddington), and which does not, of necessity, lead to Coddington’s ‘slippery slope’. In this, however, it is also necessary to consider Coddington’s claim that there is a justificationist aspect to Keynes’ account. In fact Keynes’ account does not presuppose, as Coddington suggests, that ‘knowledge, to count as such, must be demonstrable, provable, indisputable’ (p. 484). It does remain, however, a form of justificationism. In order to understand the sense in which this is so it is necessary first to follow a slight ‘detour’, and provide a brief sketch of Keynes’ theory of probability and an overview of his account of direct and indirect knowledge.
In his A Treatise on Probability Keynes emphasises ‘the existence of a logical relation between two sets of propositions in cases where it is not possible to argue demonstratively from one to another’ (p. 9). This is the probability relation or argument. Any conclusion a is related to a given premiss h via a probability relation, written a | h. In fact Keynes’ notion of logical probability provides an immediate example of the sort of relative/absolute interpretation of knowledge that will be developed below, an interpretation, I believe, that applies in a much wider context than Keynes seems to acknowledge. For Keynes a probability is in some sense relative. Just as no place is intrinsically distant - it is so many miles away from a given position - so no proposition is intrinsically probable; a proposition has a certain probability relative to given evidence or background knowledge. At the same time, however, the probability in a given proposition is not arbitrary, subject to human caprice, merely relative. Rather, once the background knowledge or evidence is given, the probability in the given proposition is fixed, and in this sense it is objective or absolute; it is concerned with the degree of belief that it is rational for a person to entertain in the proposition, given this evidence. Thus the probability in a given proposition has both an absolute as well as a relative character. Acquisition of new relevant evidence (k say) for a particular proposition may change the probability (to a | hk, say) but this does not entail that the original probability-relation (a | h) was incorrectly held; acquisition of the new ‘evidence’ merely
gives rise to the new probability relation. Thus at a particular moment in time, given the background knowledge available to an individual, the probability in the relevant proposition is fixed, or absolute. However as evidence ‘accumulates’, so too the probability attached to the proposition may change; and in this way probability is relative to the background knowledge to hand.
This kind of logical relation between sets of propositions, or from premisses to conclusions, is an important component in Keynes’ account of how certain types of knowledge are ‘justified’. Keynes, in fact, distinguishes two types of knowledge: knowledge that is obtained directly and that which is obtained indirectly - between ‘that part of our rational belief which we know directly and that part which we know by argument’ (p.12). Direct knowledge, it seems, is not knowledge which can be justified in any sense. It arises as a ‘result of contemplating the objects of acquaintance’ (p. 12 - a matter explored further below). Indirect knowledge, however, is justified knowledge. That is, relative to such premisses as are known directly, we seek to determine degrees of rational belief in other propositions which then form part of our indirect knowledge. In this process, in fact, it is not only the propositions that make up the premisses of the argument that may be known directly, for the argument or probability relation is itself a form of direct knowledge. Thus in all knowledge there is some direct element (p.15), and it is on the basis of this direct element that we seek to justify numerous further propositions. In Keynes’ words:
That part of our knowledge which we obtain directly, supplies the premisses of that part which we obtain by argument. From these premisses we seek to justify some degree of rational belief about all sorts of conclusions (p. 121).
Clearly, a fundamental feature of Keynes’ account, then, is his notion of direct knowledge which appears to provide a foundational aspect for all knowledge. In order, therefore, to address the question raised earlier as to the way in which correct beliefs or certain knowledge are to be interpreted in Keynes’ account, it is to an examination of the particular nature of direct knowledge that I now turn.
Attempts within traditional epistemology to provide foundations for knowledge have usually been made from the viewpoint of one or other of two doctrines: empiricism which stresses the primary role of sensory experience in knowledge and rationalism which claims a primary role for a priori reasoning in knowledge. As we shall see, Keynes’ account of direct knowledge has been interpreted, on occasion, as supporting each of these positions.
It has already been noted that Keynes argues that direct knowledge arises through contemplating objects of acquaintance. Such reasoning immediately
appears to place Keynes within the empiricist camp. This particular aspect of Keynes’ thinking is most fully captured by the following statement:
We start from things, of various classes, with which we have, what I choose to call without reference to other uses of the term, direct acquaintance. Acquaintance with such things does not in itself constitute knowledge, although knowledge arises out of acquaintance with them. The most important classes of things with which we have direct acquaintance are our own sensations, which we may be said to experience, the ideas or meaning, about which we have thoughts and which we may be said to understand, the facts or characteristics or relations of sense-data or meanings, which we may be said to perceive; - experience, understanding, and perception being three forms of direct acquaintance (Keynes, 1973, p.12).
The empiricist aspect of this statement has been noted by, for example, Hodgson (1985) who argues that the passage ‘supports the allegation of empiricism’ especially in its ‘assumption that sense-data carry meaning for the subject, without any mention of the language, the symbolic order, or the conceptual framework though which they are perceived’ (p. 22).
It must also be noted, however, that if Keynes’ account is an empiricist one then it is not of the extreme variety that reduces all knowledge to replicas of sense impressions. In fact the following passage appears to undermine the empiricist interpretation of Keynes’ account altogether:
In the case of non-logical or empirical entities, it seems sometimes to be assumed that our direct knowledge must be confined to what may be regarded as an expression or description of the meaning or sensation apprehended by us. If this view is correct the inductive hypothesis is not the kind of thing about which we can have direct knowledge as a result of our acquaintance with objects.
I suggest, however, that this view is incorrect, and that we are capable of direct knowledge about empirical entities which goes beyond a mere expression of our understanding or sensation of them (p. 292).
Keynes provides two examples of such direct knowledge about empirical entities which goes beyond a mere ‘understanding’ and ‘sensation’. The first is the ‘causal irrelevance of mere position in time and space’ (p. 292) and the second is ‘the law of causation’ - the belief that every object in time ‘has a “necessary” connection’ with some set of objects at a previous time. He adds that it ‘is to be noticed that neither of these beliefs clearly arises, in spite of the directness that may be claimed for them, out of any one single experience’ (p.293).
It would seem, therefore, that Keynes’ account is open to the accusation of inconsistency. The above passage suggests that direct knowledge itself involves interpretation and reason, while the former passage may be construed as denying this. Such a denial, moreover, seems even more explicit in the following statement: ‘... as to when we are knowing propositions about sense-
data directly and when we are interpreting them - it is not possible to give a clear answer’ (p. 14).
The last passage certainly can be construed as implying that Keynes believed that some direct knowledge might be possible that is devoid of interpretation. If this is indeed Keynes’ position then it would be seen to be untenable from the standpoint of most schools of modern philosophy which ‘accept’ that there is no possibility of any knowledge (direct or otherwise), even of sense-data, that does not involve interpretation. As such the account seems to stand in need of critical reconstruction.
At this stage, however, it should also be noted that some economists have interpreted Keynes’ account of direct knowledge as being effectively the ‘opposite’ of empiricism, as having more affinity with rationalism: the view that a priori reasoning takes a primary role in knowledge.
In a recent contribution which interprets Keynes’ epistemological account as a form of rationalism O’Donnell (1982, p. 78) writes:
Keynes’s philosophy is a form of foundationalism in which intuitions (or direct knowledge) serve as base and eliminate infinite regresses by virtue of the fact that they profess to provide true, certain and self-evident knowledge. This contrasts with empiricism which locates its foundations solely in sense-experience....
Intuition is seen as a synonym for direct knowledge and seems to be something other than reason or experience, as the following passages from A Treatise on Probability suggest:
the fact that we ultimately depend upon an intuition need not lead us to suppose that our conclusions have, therefore, no basis in reason, or that they are as subjective in validity as they are in origin (p. 76).
… where our experience is incomplete, we cannot hope to derive from it judgements of probability without the aid either of intuition or of some further a priori principle. Experience, as opposed to intuition, cannot possibly afford us a criterion by which to judge whether on given evidence the probabilities are or are not equal (p.94).
But, while some commentators interpret an approach resting upon intuition as being rationalist (and the latter statement by Keynes does suggest that, for him, intuition is a form of a priori principle), others such as Braithwaite (1975) see the emphasis upon intuition as yet a further indication that Keynes’ account is a form of empiricism:
The welcome given to Keynes’s book was largely due to the fact that his doctrine of probability filled an obvious gap in the empiricist theory of knowledge. Empiricists had divided knowledge into that which is ‘intuitive’ and that which is ‘derivative’ (to use Russell’s terms), and had regarded the latter as being based upon the former by virtue of there being a logical relationship between them. Keynes extended the notion of logical relation to include probability relations, which enabled a similar account
to be given of how intuitive knowledge could form the basis for rational belief which fell short of knowledge (p. 237).
It 7 is not, however, my particular concern here to argue that Keynes’ account is rationalist or empiricist. It is clear that his account is not without significant ambiguity and that, consequently, different interpretations have arisen. 8 The one factor, however, that does appear to be common to all interpretations of Keynes’ account is that his philosophy is considered to be a form of foundationalism. It is this aspect, ultimately, that gives rise to apparent problems or inconsistencies. In fact, the arguments above seem to suggest that Keynes does not particularly want to lodge exclusively in either the empiricist or the rationalist camps; yet he does want his notion of direct knowledge to form the premisses, the foundations, of all other knowledge. Keynes is then left with deciding where direct knowledge comes from: unmediated sensory perception? a priori thought? or is it something else?
In fact, foundationalism, at least of the form that appears to be presupposed by or attributed to Keynes, is not an obviously tenable position. Sensory experience, on the one hand, is never a matter of direct or straightforward contact with observable reality; it never gives us pure and uncontaminated contact with something. All experience involves an element of interpretation and judgement. The ‘sensory experience’ that we have goes beyond what our sense organs entitle us to. The argument here is not that there is no immediate sensory aspect, but that it is inextricably bound up with its mediation by education, training or thought in general. Thus sensory experience, in part, is also the result as well as the stimulation of thought and interpretation. Interpretation, judgement and inference are always involved. On the other hand, thought, reasoning and inference (which typically may be considered to be developments (i.e. in the light of experience, etc.) of simpler and more primitive biological and physical responses) 9 are not a priori in the sense of being totally prior to, and independent of, experience. Thus, knowledge can be understood, not as the building of a superstructure upon an unchanging foundation, but as proceeding in stages where the foundation at each new stage is the previous one. In the course of acquiring and developing knowledge, provisional starting points come to be questioned and criticised and existing views are rethought and reinterpreted. Thus, despite strong temptations to draw distinctions between what is immediately given to the senses and what is
7. In fact Braithwaite’s reference to Russell’s contributions is not without significance. Russell wrote a series of papers on ‘knowledge by acquaintance’ and ‘knowledge by description’ throughout the period in which Keynes was working on A Treatise on Probability (see for example Russell, 1905, 1912, 1914, 1917); the two of them were in frequent contact with each other, and much of the terminology they employ is very similar. Similarities exist even down to the level where, for example, they both cite ‘the other side of the moon’ as a case in point of something which we do not experience or have direct knowledge of - see Russell (1914, p. 130) and Keynes (1973, p. 14). Russell is also variously interpreted as being empiricist, rationalist and ultimately inconsistent.
8. It should, of course, be noted that different contributors do use notions such as empiricism and especially rationalism in slightly different ways. Such differences, however, do not seem sufficient to explain the very different interpretations that commentators on Keynes have arrived at.
9. Even Popper (1974, p. 46) talks of ‘inborn reactions and responses’ which are not conscious but which we may still speak of as ‘inborn knowledge’.
contributed by interpretation, these two aspects - the immediate and the mediated, the given and the constructed - cannot ultimately be separated. Knowledge development involves thought and experience where neither aspect can be isolated as foundational in any absolute and permanent way.
It would, therefore, seem to be the case that to the extent that Keynes’ notion of direct knowledge is supposed to provide absolute and immutable foundations, it is no longer tenable. How, then, should claims to direct knowledge be treated? Perhaps the notion of direct knowledge should be dispensed with altogether? The problem with this suggestion is that to do so may put us back on the ‘slippery slope’ to complete subjectivism; it may take us once again to the conclusion that knowledge is nothing more than the result of creative acts of the imagination, with no objective or absolute aspects whatsoever. The dilemma that was facing Coddington appears to arise once more.
A possible and indeed tempting reaction to the assertion that there is no such thing as direct, unmediated knowledge is to maintain that, despite the arguments set out above, there simply is such a thing, in the sense that we do know certain claims or propositions directly we hear, think or ‘perceive’ them, while other claims for knowledge are not so immediate. The concept of’ truth’ may be potentially mysterious with numerous epistemological, and perhaps definitional, problems associated with it, but is it not clear and straightforward that some knowledge - of everyday things, ourselves, our surroundings, etc. - is immediate? The answer, I suggest, is that we do indeed have direct knowledge, but, just as with the notion of probability, direct knowledge is both a relative and an absolute notion. Just as no place is intrinsically distant, and no proposition is intrinsically probable, so no proposition is intrinsically directly knowable; the directness of knowledge is relative to the background knowledge of the knower. What a trained physicist experiences, and sees when he or she is presented with reactions in an X-ray tube, is not so immediately accessible to, or knowable by, someone without the same training and education. Closer to ‘home’, perhaps, an econometrician glancing at a computer print-out may perceive, more or less immediately, that certain claims to statistical consistency ‘hold up’ in some sense or do not. Others possessing the same print-out may need a. course on introductory econometrics and familiarity with the relevant computer program before such knowledge can be acquired. Similarly a story or idea written in English may be more or less directly accessible to someone who is proficient in the English language, but not to someone who is illiterate, or perhaps literate only in a different language. Direct knowledge, and thus all knowledge, then, is a social phenomenon. Different people in different situations, societies, historical periods, etc., with different educations, roles and experiences, are liable to see the world differently. Knowledge has a relative aspect and as we shall see this is a conclusion of consequence for economic analysis. However, in accepting this conclusion there is no compulsion, thereby, to mount the ‘slippery slope’; it in
no way necessarily follows from this conclusion that knowledge must be merely relative. Relative to training, etc., there can still be an absolute aspect to what a physicist experiences in the test-tube or to what an econometrician perceives in reading the ‘printout’. Relative to training and experience, etc., we can consider such knowledge to be direct, absolutely.
Thus, on this view there is such a thing as direct knowledge which corresponds to our common everyday understanding of the term. But it is not a matter of absolute and pure immediacy. Rather it is a relative immediacy, depending upon our level of biological and social development. The directness of knowledge is relative in this sense but, relative to such background knowledge, etc., it is absolute.
The suggested relative/absolute nature of direct knowledge and so of knowledge in general bears, in turn, on the nature of truth and certainty, and thus returns us to the issues raised at the beginning of the discussion. According to accounts such as Keynes’ ‘only true propositions can be known’ (p. 18); knowledge corresponds to certainty of belief which in turn corresponds to correct belief or truth. In terms of the viewpoint that is now being put forward, therefore, it seems clear that, if such a notion of truth is to be retained, then it must be understood as having both a relative and an absolute character. It is a relative judgement that is being made when it is asserted that current beliefs are true or rational. Our beliefs are justified relatively, not absolutely; they are justified relative to currently available empirical evidence and our understanding and interpretation of it. In this sense such beliefs are justified and in this sense they are true. Knowledge and truth are never purely absolute, but it does not thereby follow that they must be merely relative. On the view being put forward something that is ‘known’ or is ‘true’ is so relative to existing interpretations and experience; but, as such, this knowledge is something that is justified.
We may even talk, in a given context, of absolute truth,
but this supposes that we take the limits and relativity of our viewpoint for
granted. We may equally find,
however, that what we hold as true changes.
In short, beliefs held as true, as knowledge, are not pure truth from an absolute standpoint, never to be replaced by a closer or better ‘approximation’
to reality; but neither are they sheer error. Rather they represent the best account available in terms of both logical coherence and correspondence with evidence and experience. They represent a step on the path to a fuller truth from a wider viewpoint. Truths are partial and relative, to be replaced by better ways of understanding things, by fuller truths.
Having set out this account of the relative/absolute nature of knowledge it is now possible to address explicitly the issues which motivated this discussion and, further, to indicate various implications that the discussion bears for economic analysis in general.
The foregoing discussion, then, appears to provide an optimistic response to the problems and issues raised earlier on. If we accept that knowledge can have a relative/absolute character, it is possible, even in a non-probabilistic context, to avoid the presumption of infallible omniscient agents without giving way to extreme forms of subjectivism whereby beliefs are treated merely as creative acts of the imagination. In the account set out, knowledge can involve a ‘display of certainty’, and although currently held knowledge or beliefs held as certain or true may be found to be deficient from the perspective of knowledge, experiences or interpretations yet to be acquired, they are, provisionally at least, fixed and in that sense absolute.
These general conclusions, clearly, do not necessitate an account of probabilistic knowledge. Nevertheless, probabilistic knowledge may be pervasive even if it sometimes amounts to little more than the view that a given event is more probable to occur than not. For example, it seems that rules and rule-systems are central to most human knowledge (Lawson, 1987 a). Individuals ‘go on’ in their ‘day to day’ activities using their knowledge of matters such as the highway code, rules of social conduct, the structure of language, and so on. The conformity of individuals to such rule-systems tends to make their actions predictable, and enables each individual to get by in a purposeful manner that is coherent to others. However, there will always be some individuals who ignore part of the highway code, or who are not fully socialised into the norms of society (or who rebel against them) or who are relatively illiterate in the relevant language, and so on. In such cases, clearly, a belief about the ability/desire of other people to conform with given rules/ conventions may have a probabilistic interpretation even if it amounts to little more than the view that an unknown individual is more probable than not to ‘conform’ in a given situation. In any case, whatever the significance of probabilistic knowledge in such situations, the point remains that all knowledge is fallible, and probabilistic knowledge is merely an example of such fallible knowledge.
The significance of the above outline can be more fully brought out by focusing upon an essential element: a recognition that context matters, and does so in a significant way. Human action always takes place in some context and human agency is conditioned by context-related knowledge. Whether the object of economic analysis is to understand and describe human activity or whether it is to formulate and implement policy, context cannot be neglected.
This sort of conclusion, or course, ought to inform all
areas of social analysis, but it seems to be adhered to in a significant way
only outside economics. 10 Economists, in fact, seem quite prepared to assert,
without further discussion, that modes of behaviour that exist, for example, in
10. Not surprisingly, perhaps, such conclusions seem most widely appreciated in subject areas where communities under study are typically widely separated from the researcher’s own - such as history and anthropology (see, e.g. Winch, 1970).
11. As Kahn et al. (1983) observe in their
12. For example, it is often supposed by researchers working within the Labour Market Segmentation Theory framework that there necessarily exists a one-to-one correspondence between segmentation in product markets and segmentation in labour markets (see Lawson, 1981). The reasoning, typically, is that because a primary sector firm will be able to support primary employment conditions it will therefore choose to do so - on the assumption that this is the only way a large-scale firm can secure relatively high productivity from its workforce and at the same time minimise turnover costs. Secondary workers, instead, are assumed to reveal unstable work habits, high absenteeism and poor work-rates. In a case study (Lawson, 1981), however, I found that, on the contrary, primary sector firms are able to secure both secondary employment conditions and a reliable, highly committed and loyal workforce. The important factors behind this observation were the relatively sheltered experiences and interpretations of a largely rural local workforce, giving rise to a particular world view which was clearly recognised and encouraged by local businesses, enabling paternalistic forms of industrial relations, or employer authority, to be maintained.
The main point is a general one and bears repeating. All action occurs in context, and thus all capable, intentional action is influenced by context-related experience and knowledge, including modes of rationality, inference and thought in general. In short, knowledge has a relative aspect. As argued at length above, however, this observation does not mean that knowledge need be interpreted as merely relative. It does mean, however, that the relative dimension cannot be neglected. It follows, of course, that if economic analyses are to be at all relevant, if economists want to understand and describe existing forms of rationality and purposeful action and to influence human activity through policy formation and so forth, then to some extent economists must give up attempts to theorise purely a priori and must go and ‘look’ as well. At the very least there seems to be a case for allocating greater resources to examining the results of case-study and other primary sources. If context or the relativity of knowledge is completely omitted from economic analysis (and to a significant extent it does seem to be) it is difficult to understand the point.
The last comment can be put a bit more precisely, as follows. From the realist perspective of the relative/absolute interpretation of knowledge it follows that economic doctrines which presume immutable foundations must either make claims which, interpreted realistically, are untenable, or, alternatively, must be concerned with something other than realism of analysis. Now attempts, within economics, to provide such foundations appear, in fact (and perhaps not surprisingly), to correspond to one or other of the philosophical foundationalist doctrines discussed above: empiricism or rationalism. To illustrate the point being made, therefore, examples of each can be considered in turn.
In economics, empiricist inclined accounts tend to assert the primacy and objectivity of collected data as though these represent uninterpreted facts, uncontaminated by judgement and theory, and as though knowledge arises from them via analysis. An unusually explicit 13 example of this sort of position is provided by Wolfe (1924, p. 450):
If we try to imagine how a mind of superhuman power and reach, but normal so far as the logical processes of thought are concerned,… would use the scientific method, the process would be as follows: First, all facts would be observed and recorded, without selection or a priori guess as to their relative importance. Secondly, the observed and recorded facts would be analysed, compared, and classified, without hypothesis or postulates other than those necessarily involved in the logic of thought. Third, from this analysis of the facts, generalizations would be inductively drawn as to the relations, classificatory or casual, between them. Fourth, further research would be deductive as well as inductive, employing inference from previously established generalisations.
Perhaps not all accounts in economics which emphasise a primary role for data and data analysis are so explicit or extreme; nevertheless, some empirical
13. And consequently one that is often discussed - see, e.g. Hempel (1966, p.ii) or Chalmers (1980, p.10)
work does appear to labour under the misapprehension that collected data represent relatively uncomplicated, purely objective observations upon which all analysis must somehow be founded. From the perspective of the above discussion such claims are clearly untenable.
More commonplace, perhaps, than the empiricist orientated foundationalist claims are those corresponding in some way to rationalism: those stressing a primary role for purely a priori reasoning. Prominent amongst these are axiomatic approaches to economics as exemplified by Debreu’s (1959) renowned contribution concerning economic equilibrium. Clearly such a priori foundational approaches do not meet the realist view of assessment through continual interaction between theory and experiment/practice/observation and, consequently, they seem unlikely convincingly to claim realism as a goal. Indeed, for Debreu it seems that an axiomatic structure is necessarily constructed without any prior claims to realism and, indeed, without any reference whatsoever to any interpretation that might eventually be attached to it:
Allegiance to rigour dictates the axiomatic form of the analysis where the theory, in the strict sense, is logically entirely disconnected from its interpretations...
… Such a dichotomy… makes possible immediate extensions of that analysis without modification of the theory by simple reinterpretation of the concepts (p. x).
In fact, for Debreu, the axiomatic structure represent neither model nor abstraction, and there is no requirement that it be confronted with empirical observation. The axiomatic approach gives rise to a formal structure and it is legitimate to analyse it merely as a formal structure, independently of any interpretation that eventually may be tagged on. The intended aim is clearly not realism and there seems to be no reason to suppose that such a structure would be susceptible of realistic interpretation. (On all this see Ingrao and Israel, 1985.)
In truth, however, a claim that economic axioms are chosen completely independently of interpretation is not particularly coherent, and in any case it does not appear to be widely subscribed to. Hahn, in particular, seems prepared to suggest otherwise:
Axioms are not plucked out of thin air and far from distancing the theorists from what somewhat mysteriously is called the ‘real’ world they constitute claims about this world so widely agreed as to make further argument unnecessary (p. 5).
A relevant problem 14 with this criterion is that the axioms that are widely
14. An interesting feature of this criterion is its historical relativity. A problem with it, suggested by the above account, is that argument is rarely, if ever, unnecessary. Many theories once held as knowledge, many claims about the world once widely agreed upon, are no longer held as true. It appears to be no longer widely held that the earth is flat, for example, or that the sun revolves round the earth, and so on. What does ‘unnecessary’ really mean here? Similarly, what does ‘so widely agreed’ really amount to? By all people? Economists? Economists adopting an axiomatic approach?
agreed upon, such as, perhaps, ‘agents have preferences’, tend to have little or no analytical content without first being supplemented with other, more contentious claims or, in Hahn’s words, before being ‘idealised and strengthened’. The result, however, and as Hahn acknowledges, is not something that can masquerade as a claim so widely agreed upon as to make further argument unnecessary:
However at the end we shall have to agree that the genuine axiom: persons have some preferences has been idealised and strengthened by theorists beyond the point at which it commands universal consent (Hahn, 1985, p. 7).
In practice it appears that much of the ‘idealising and strengthening’ involves convenience-assumptions that are designed merely to achieve mathematical tractability. It is in this sense that the axiomatic structure becomes particularly arbitrary, or ‘a priori’, with the result that ‘theorising’ relinquishes its contact with realism. Not surprisingly, Hahn eventually observes that no-one can understand accounts such as general equilibrium theory and take them to be descriptive. Needless to say, however, many economists do treat them as descriptive, as Hahn critically observes (p. 14). However, those intimately involved in developing axiomatic ‘foundations’ seem eventually to acknowledge, and to embrace, with Hahn, the conclusion that ‘the aim is not realism...’ (p. 13), but something else. 15
To reiterate, foundationalist accounts do not seem relevant from a realist standpoint where the object is to increase knowledge of real-world economic processes. Empiricist orientated accounts, which claim a primary role for data and data-analysis, free of judgement and interpretation, are untenable; rationalist orientated accounts emphasising a primary role for a priori reasoning, if they are to be free from similar error, seem destined to end up with a criterion of relevance other than realism of analysis.
An additional implication of the viewpoint set out above is that knowledge does not merely accumulate in a straightforward or piecemeal fashion. Hey (1983a, p.1) observes that knowledge ‘in economics is accumulated in much the same way as in any other discipline, or indeed in everyday life’. Keynes makes much the same point in A Treatise on Probability. The above account is to some extent supportive of this observation but emphasises a particular aspect of the process of knowledge acquisition: that progress in knowledge is not a purely quantitative affair. The view that ‘scientific development’ is a piecemeal process whereby ‘items have been added, singly and in combination, to the ever growing stockpile that constitutes scientific technique and knowledge’, of course, has been effectively criticised by Kuhn (1970, p. I). Kuhn, in fact, does
15. In fact it seems that the objectives of the research
project associated with general equilibrium theory have been found to be
untenable, and, consequently, the project itself must be regarded as having
failed (see, for example, Ingrao and
acknowledge a ‘normal’ or ‘quantitative’ phase of the development of science. However, during this quantitative phase anomalies and problems are generated, and their accumulation leads, ultimately, to a crisis situation whereby a qualitative change occurs. In fact there are reasons for suspecting that such qualitative changes occur relatively frequently in social analysis such as economics. First, (well controlled) experimentation is rarely possible so that the independent effects of various factors upon each other cannot usually, if ever, be completely isolated and, second, such ‘laws’ or ‘structures’ as are discerned are historically contingent and liable to change (see Lawson, 1985). Under these circumstances a qualitative revision of ‘truths’ may be a frequent occurrence. Of course, the suggestion is not that a qualitative change will mean a total or arbitrary shift in perspective; there is always continuity as well as discontinuity. But the point remains that progress in knowledge has a qualitative as well as a quantitative moment.
The account outlined in the previous sections bears further implications that cannot be considered here. However, the examples included above should suffice both to suggest that the account set out is not without consequence for the conduct of economic analysis in general, and also to indicate the sort of analytical implications that follow.
The nature and extent of knowledge held by individuals is central to economic analysis, an observation that seems increasingly to be acknowledged by economists. In fact, Rutherford (1984) suggests that the ‘problem facing economists amounts to nothing short of explaining the state of knowledge or expectations’ (p. 377) and Loasby (1986), amongst others, observes that it ‘is now becoming widely recognised that many of the central unresolved problems in economics turn on questions of knowledge’ (p. 41). The ways in which economists might go about addressing such questions are, of course, various, and numerous epistemological issues and frameworks might be emphasised. In this paper I have focused upon the observation that individuals in all walks of society appear to hold onto numerous beliefs as if they are, in fact, true or certain; and I have suggested that this is an observation that realist economics ought to accommodate. I have also suggested that accounts which suppose that such beliefs must be true or certain in some purely absolute and indisputable way are inappropriate to realist analysis, as are accounts which, in contrast, take the view that such beliefs have no absolute element at all but are merely relative. Consequently, I have put forward a ‘relative/absolute hypothesis’ whereby knowledge, or true belief, has the characteristic of being both absolute as well as relative. Clearly, the viewpoint, as set out, is at a rather general level and may entail various different paths of construction and development. However, as noted above, even at this general level the account seems to bear implications of some consequences for current economic analysis.
The empirical validity, or otherwise, of the account might be gauged by focusing upon various implications that seem to arise from it. For example, and
again at a very general level, the suggested relative aspect of knowledge gives rise to the implication that views held with conviction - perhaps concerning industrial matters or political issues in general - are likely to vary significantly across time and space contexts (the latter including, for example, across countries, between urban and rural centres, and perhaps even between northern and southern regions, and so on). The objective or absolute aspect of knowledge suggests that within relatively localised, or similar, contexts there should be greater conformity of belief. By similar reasoning we should expect a greater variation in such beliefs across ‘open’ communities where the inhabitants partake in a wider range of occupations/social practices than, say, in close-knit, or ‘sheltered’ communities (see, e.g. Lawson, 1981). Also where experiences are most segmented according to factors such as age or gender positions we may expect some influence to be found on the degree of variation of interpretations held. And so on. Indeed such implications seem likely to be borne out, being totally consistent, for example, with the often observed incomprehension shown by one group of individuals to the beliefs and actions of a different group inhabiting, or ‘coming from’, a significantly different environment or situation.
The operationalisation of the account set out above involves recognising that the emphasis on the importance of the social context in which beliefs are formed lends weight to those analyses which attempt to incorporate the results of case-study and other forms of primary research. The general point is that, to the extent that knowledge has a relative aspect, then attempts to understand individual beliefs must take account of the context in which agents are acting and learning: and to the extent that knowledge has an absolute aspect, then it follows that the task of learning about individual beliefs is unlikely to be hopeless or irrelevant. It might also be noted, moreover, that these characteristics of knowledge allow individuals to act in purposeful and capable ways within the context in which they are situated, and, indeed, to be able, in turn, to influence their social environment. Thus, the viewpoint set out above is supportive of a particular analytical account which, perhaps, might be termed societal interactionism. On this view, individual agency and social structures and context are equally relevant for analysis - each presupposes each other. Thus any reductionist account stressing analytical primacy for either individual agents or for social ‘wholes’ must be inadequate. A realist account needs to be thoroughly interactionist.
Various further implications for analysis are contained in the above discussion, but the main conclusion remains, basically, an optimistic one: analysis which involves accounts of the beliefs of individuals, or groups of individuals, seems entirely feasible, despite the problems noted in the course of the discussion - although the other side of the same coin, clearly and to repeat, is that any account that neglects the context of individual action and learning is liable to be deficient.
Date of receipt of final typescript: July 1987
Braithwaite, R. B. (1975). ‘Keynes as a philosopher’, in
Keynes, M. (ed.) Essays on John Maynard Keynes.
Chalmers, A. F. (1980). What Is This Thing Called
Coddington, A. (1982). ‘Deficient foresight: a troublesome theme in Keynesian economics.’ American Economic Review, vol. 72, no. 3, pp. 480-7.
Debreu, G. (1959). Theory of Value: An Axiomatic Analysis of Economic Equilibrium. Cowles foundation monograph, Yale University Press.
Hahn, F. (1985). ‘In Praise of Economic Theory’, the
1984 Jevons Memorial Fund Lecture. Published by
Harrod, R. F. (1972). The Life of John Maynard Keynes. Harmondsworth: Penguin Books.
Hempel, C. G. (1966). Philosophy of Natural Science.
Hey, J. D. (1979). Uncertainty in
------------ (1983a). Data in Doubt.
------------ (1983 b). ‘Wither uncertainty?’ Economic Journal (Conference Papers). pp. 130-9.
Hodgson, G. (1985a). ‘Persuasion, expectations and the limits to Keynes’, in Lawson and Pesaran (1985).
Ingrao, B. and
Kahn, P., Lewis, N., Livock, R. and Wiles, P. (1983).
Picketing: Industrial Disputes, Tactics and the Law.
Keynes,J. M. (1973). The Collected Writings of John Maynard Keynes, vol. viii, The Treatise on Probability. Royal Economic Society.
Kuhn, T. S. (1970). The Structure of Scientific
Revolution, 2nd edition.
Lawson, T. (1981). ‘Paternalism and labour market segmentation theory’, in Wilkinson, F. (ed.), The Dynamics of Labour Market Segmentation. Academic Press.
------------- (1985). ‘Uncertainty and economic analysis.’ Economic Journal, vol. 95, pp. 909-27.
------------- (1987 a). ‘Rule systems and
economic analysis: the case of
------------- (1987 b). ‘Realism and
instrumentalism in the development of econometrics.’ Mimeo,
-------------- and Pesaran, H. (eds.) (1985). Keynes’
Economics: Methodological Issues.
Loasby, B. J. (1986). ‘Organisation, competition, and
the growth of knowledge’, in Langlois, R. N. (ed.), Economics as a Process:
Essays in the New Institutional Economics.
Lucas, R. E. (1972). ‘Econometric testing of
the natural rate hypothesis’, in Econometrics of Price Determination, 0.
O’Donnell, R. (1982). ‘Keynes: philosophy and
economics’, Ph.D. dissertation,
Popper, K. R. (1974). Conjectures and Refutations.
Russell, B. (1905). ‘On denoting.’ Mind,
reprinted in R. Marsh (ed). Logic and Knowledge.
‘Knowledge by acquaintance and
knowledge by description’, in Problems of Philosophy, pp.46-60.
-------------- (1914). ‘On the nature of
acquaintance’, The Monist; reprinted in R. Marsh (ed.), ‘Logic and
------------- (1917). ‘Knowledge by acquaintance and
knowledge by description’, in Mysticism and Logic.
Savage, L. J. (1954). The Foundations of
Winch, P. (1970). ‘Understanding a
primitive society’, in B. Wilson (ed.).
Wolfe, A. B. (1924). ‘Functional economics’,
in Tugwell, R. G. (ed.), The Trend of Economics.