The present note will be concerned only with Sir Partha Dasgupta's recent article in this journal (Dasgupta Reference Dasgupta2005). What is more, it will concentrate on those parts of the article which contain a serious misreading of Hilary Putnam's position on the entanglement of facts, theories and values. These philosophical matters can perhaps be clarified for economist readers (they should require no clarification for philosophers) by considering, to begin with, Dasgupta's interpretation of the Bergson–Samuelson position. What (Bergson) Burk (1938) and Samuelson (1947) were doing, according to Dasgupta, was to establish ‘the ethical foundations of the subject. . .over five decades ago’ (Dasgupta Reference Dasgupta2005: 221–2).Footnote ² Thus a major theme of the article is heard at once: economics is supposedly based on sound ethical foundations, and these can be traced (it is supposed) to specific work written long ago, and hence needing no augmentation. These ethical foundations, it is claimed, ‘are now regarded to be a settled matter’ (2005: 222).

This theme is continuously repeated – thus we are told that: “modern economics is built on broad ethical foundations . . . But since the founda-tions themselves were settled decades ago, research economists don't find it necessary to rehearse them over and over again . . . So, the ethical foundations of modern economics are regarded as unspoken assumptions in research publications” (2005: 225). The founding fathers who allegedly laid down these foundations are explicitly mentioned in the statement that “over a half-century ago (Bergson) Burk (1938) and Samuelson (1947) established the foundations of policy evaluation on a broad ethical structure” (2005: 226). Dasgupta realized that “merely to refer to (Bergson) Burk (1938) and Samuelson (1947) won't do” (Ibid). But it is on this allegedly ethical foundation, which he considers to be “broad and strong” (Ibid) that he builds, in order “to offer an account of the ethical foundations of modern economics” (2005: 227).

We shall offer just two arguments in this note in response to the position sketched above. First, that it deeply misunderstands what Bergson and Samuelson sought to do, and the historical conditions which dictated their aims, making nonsense in particular of Samuelson's allegedly clear grasp of what was necessary in order for welfare economics to escape the (then) deadly fire of logical positivism. And secondly that Dasgupta's travesty of the Bergson–Samuelson argument is a major reason for Dasgupta's later complete misreading (throughout his article) of Putnam's position on the entanglement of facts, theories and values, and for his inability to grasp the major role of Quine in the scientific aspects of the triple entanglement arguments.

WHAT BERGSON AND SAMUELSON ACTUALLY DID

The crucial point is that Bergson, in his brilliant youthful paper, went to great pains to avoid having to adopt any ethical foundations, and to show that it was possible for a ‘welfare’ economist to give a precise account of the effects of any economic policy without endorsing the values of its proponents. For his part, Samuelson, who was well aware of the influence at that time of logical positivism on the economics profession, explicitly endorsed the (then) fashionable position that “ethical value judgments have no place in scientific analysis” (Samuelson [Reference Samuelson1947] 1983: 219). He would have been horrified at the suggestion that economics was “built on broad ethical foundations” (2005: 225). It was for showing how to avoid exactly this that he so admired Bergson's work.

He was brutally blunt about this: “the scientist does not consider it any part of his task to deduce or verify (except on the anthropological level) the value judgments whose implications he grinds out” (Samuelson [Reference Samuelson1947] 1983: 210). The economist might study the efficiency (or lack thereof) of the economic arrangements of the headhunters of Borneo, or of the British National Health Service. In either case Samuelson's pellucidly clear position entailed that the economist so engaged would be freed (by Bergson's work) from any danger of being involved in the values of Bornean warriors or British bureaucrats.

One of the present authors was a young member of Lord Robbins's department at the London School of Economics in the days when logical positivism was still vigorously debated by the young turks there. At that time Lionel Robbins's faculty and graduate seminar was regularly visited by the leading American neoclassicists, like Paul Samuelson and Will Baumol (who would later become a friend of both of the present authors). There was also an economics seminar which met at Oxford, Cambridge, and London and to which the young people went. So, aside from reading the articles and books which were coming out, we heard a good deal from the people who wrote on welfare economics then. We had a pretty good idea of what was being said. The present author who was at LSE in those days, however, has no memory whatever of the Bergson–Samuelson work being interpreted as having provided an ethical foundation for economics. A glance at an economist who was also an Oxford philosopher, and who was writing then, confirms the belief based on memory. Ian Little wrote: “The fact that the conclusions of welfare economics are inevitably, in part at least, value judgments raises the question of whether it would be best to take the view that welfare economics simply draws out the logical conclusions of a set of consistent value axioms which are laid down for the welfare economist by some priest or parliament or dictator” (Little [Reference Little1950] 1957: 80, emphasis added). He adds in a note: “The chief representatives of this view seem to be Professor A. Bergson in ‘A Reformulation of Certain Aspects of Welfare Economics’, Q. J. E., Feb. 1938, and Professor P. A. Samuelson in Foundations of Economic Analysis, Ch. viii.’ (Ibid).

Present-day readers need not be surprised that American economists in the late 1940s and 1950s should have embraced with enthusiasm the arrival of the Bergson–Samuelson position, which allowed them to avoid having to provide ethical foundations. One of the present authors remembers being asked by a famous American mathematical economist at that time: “Do you know why we invented the term ‘behavioral science’? We were afraid that writing of ‘social science’ would get us in trouble with some backwoods congressmen!”

Bergson–Samuelson, with a clear bill of health from logical positivism, offered a safe haven, secure from the storms of the cold war. But how can Dasgupta possibly see this as foundation for the place of values in economics today? To begin with, it does no justice to the vivid role of values in several important previous works of Dasgupta himself (as we shall explore later in another place).

THE ENTANGLEMENT OF FACT, THEORY AND VALUE

There is no awareness of entanglement of fact and theory, much less of fact and value, in Bergson–Samuelson. Samuelson certainly knew enough logical positivism to have sniffed out any such, had he suspected its existence. But the analytic/synthetic dichotomy, combined with the fact/value dichotomy ruled supreme then. In 1947, W. V. O. Quine's first deadly barrage on the outer defences of the fact/theory dichotomy had not yet begun, nor had Putnam's later advance, through the breach that would be cleared by Quine, toward the assault on the fact/value dichotomy.

So far from having provided values for economics, Samuelson knew that what he was promoting was an elegantly simple and coherent way in which, it appeared, an economist could avoid commitment to any values. The older welfare economics had been open to philosophical attack by positivism, and also it had made social policy claims which opened it to political danger. With Bergson–Samuelson, welfare economics was discreetly and painlessly filleted.

Dasgupta occasionally gives the impression that Putnam should be satisfied with the way in which Dasgupta is treating values (and in earlier work Dasgupta had endorsed the idea of entanglement). But the position on values which he takes in his article (2005), and which we have been examining in this note, has absolutely nothing to do with Putnam's position. Putnam is nowhere concerned with analysing an economist's efforts to maximize the attainment of someone else's values – and in a supposedly value-free manner, as was the hallmark of the Bergson–Samuelson tradition! On the contrary, Putnam's position, developed in a series of works since the 1980s, is concerned with sustaining the claim that all scientific discourse is an entanglement of facts, theories and values. What this work shows is that all scientific discourse is doubly entangled, and that the entanglement of facts and theory blows a hole in the defences needed in order to keep out the entanglement of fact and theory with values. This is a long story, and only one or two rather elementary aspects of it can appropriately be sketched here.

An aspect of entanglement which may speak to the interests of economists concerns certain features of the construction of economic models. Even undergraduate texts on microeconomics often devote a page or two to the discussion of how economic model building requires (as does other scientific model building) that one simplify nature. The situation is sometimes compared to the making of maps. A map of the Rocky Mountains designed for hikers will need no data on electoral districts, mineral deposits, population density and so on. A map for a student of geology, on the other hand, will not need to show hiking trails, but will need other information. The reader is then blithely told that scientific models (and maps) are value-free!

But the whole construction of a map (or of a model) crucially involves evaluations of what is important and what is not. The model (or map) can be expected to fit the interests of those who pay for it very well – like a suit made in Savile Row. But if you are poor, or uneducated, or of a different caste or class or gender, it may not fit your most vital needs at all well.

So in economic models, mathematical assumptions concerning continuity, or about convexity (for example), and factual assumptions concerning the information possessed by agents in the model, may be instrumental in giving the model certain value properties – or in concealing the fact that certain values will actually not be promoted if the policies which the model appears to support are carried out. Any economist will be able to think of numerous examples from the recent past.

Philosophical readers may find it useful to note that the values in an economic model are often to be found best by probing for what has been left out. An elegant example of such probing can be found in Amartya Sen's critique (Sen Reference Sen2002: 121–258) of the usual axiomatizations of ‘rational’ choice theory. What he does is essentially to show how changes in the axiomatization allow one to bring certain socially important kinds of choice making (usually left out) onto centre stage and under a spotlight. (Compare, Walsh Reference Walsh2007, forthcoming.)

This raises an interesting question as to the place of values in axiomatic, formal work in economics. Economists have been using the methods of formalist mathematics since the 1940s, but when discussing the place (or lack of place) of values, in their field they almost never speak or write of the entanglement of facts, theories and values. It is as if they believed that the fact/value dichotomy can be expected either to stand or to fall on its own legs. But Putnam has for a number of years been at pains to lay out the reason why we are unavoidably dealing with an entanglement of facts, theories and values. It is like a three legged stool – all three legs are needed, or it falls over. What is more, the leg which broke first – and fatally – was the fact/theory dichotomy. This, of course, was largely the result of W. V. O. Quine's demolition of the inflated metaphysical analytic/synthetic dichotomy which had been a vital part of the foundation of logical positivism. (See Quine [Reference Quine1951] 1953: 20–46 and Putnam [Reference Putnam1975] 1979: 33–69; 2002: 7–18, 29–30.)

As James Conant (in one of his excellent editorial introductions to a work of Putnam's) puts the matter “Putnam wants to argue not only that the normative and descriptive dimensions of our paradigmatically ethical concepts are hopelessly entangled (and cannot be separated into distinct ‘components’), but that, upon careful reflection essentially the same point can be seen to apply to our most fundamental epistemic and scientific concepts as well” (Conant, Introduction to Putnam Reference Putnam and Conant1994, lxi). He cites Putnam: “Many of our key notions – the notion of understanding something, the notion of something's making sense, the notion of something's being capable of being confirmed, or infirmed, or discovered to be true, or discovered to be false, or even the notion of something's being capable of being stated – are normative notions, and it has never been clear what it means to naturalize a normative or partly normative notion” (Putnam Reference Putnam and Conant1994: 260).

Another reason why the entanglement of theory and values is not often seen by economists may be that they expect to see values in a natural language, but hope to have left values behind when they leave English (or French, or Chinese) aside and move into a constructed language (like convex set theory). The logical positivists, it should be recalled, regarded the fact that natural languages were peppered with value words and expressions as one of the reasons why (they believed) all sciences must withdraw to the chaste cloisters of an artificial language in which, by construction, values supposedly could not appear. This titanic project of constructing such a language and translating science into it, of course, sank from its own monstrosity without need for an iceberg. But some of the reductionist point of view which cherished such dreams may still be around. A serious study of Sen's mathematical work should be enough to make it clear that mathematics, rather than offering a pure Alpine air in which values cannot live, is instead a beautiful environment for cultivating (for example) the ethical implications of social choice (see Sen Reference Sen2002, especially).

It remains to say a word or two about economic orthodoxy. We believe that our view of Samuelson's interpretation of Bergson's approach to welfare economics would have been considered orthodox at LSE (and at equivalent American institutions) in the 1940s and 1950s. We also believe that Samuelson's respect for logical positivism was widely (though not universally) shared by neoclassical economists at that time. But the argument in this note does not depend on any position concerning what (if anything) counts as orthodox neoclassical economics at the present time.