Induction by Any Other Name: A Response to Blaug’s Assessment of Friedman
by Shaun Terry
I would argue that Blaug is right to be skeptical of Friedman. However, I would argue that he lets Popper off a bit too easy.
Blaug rightly argues that Popper and Friedman are not making the same exact argument. However, they do not seem to completely avoid each other’s problems. Friedman’s position represents a sort of instrumentalism, but Blaug takes issue with Friedman’s position on assumptions:
“The idea that unrealistic ‘assumptions’ are nothing to worry about provided the theory deduced from them culminates in falsifiable predictions carried conviction to economists long inclined by habit and tradition to take a purely instrumentalist view of their subject.”
I agree that there are problems with such assumptions, but I contend that Popper runs into some of the same issues, while it might be argued that Friedman’s argument creates problems that Popper does not run into.
When Friedman describes the case of the leaves to argue for the utility in assumptions, he extrapolates to say the least. Friedman seems to assert that an assumption can serve as shorthand for positive facts, but this method presents problems. Even if the use of assumptions could be interpreted as shorthand, if we give as much credit as possible to Friedman, the fact is that the step between the observation of facts and the assumption is inductive.
What makes a rule? It is one thing to say that we have observed specific leaf density patterns; it is quite another thing to say that this is reflective of the behavior of leaves in general. This is a case of begging the question. Because some leaves demonstrate a pattern, should we presume that all leaves demonstrate this pattern? How many observations do we need to make before we can make such an assumption? The pattern is surely reflective of the reality that has been demonstrated within the set of observations, but it cannot necessarily lead to any presumption of a general case.
The case of the leaves makes an oddly anthropomorphic assumption about individual leaves, as well. Do leaves “seek” anything? What does that mean? To my knowledge, plants do not possess something like a central nervous system, so this characterization seems quite misleading, on top of the fact that it could likely only be guesswork. One gets the sense that there is accessibility in a false description of leaves such that they are likened to organisms that seek. Humans can maybe understand leaves as having motivations, but something seems to be amiss in Friedman’s characterization of the leaves, just as in assumptions made in economics. Friedman is clear when he states, in referring to economic assumptions, “It is more compact and at the same time no less comprehensive.” It is less comprehensive. Stating that leaves seek to maximize exposure to sunlight necessarily implies that leaves seek the behaviors that will maximize exposure to sunlight. We have no way to be sure of this because we cannot assume to know all possible sunlight-seeking behaviors. Therefore, the behaviors of the leaves may not be better at optimizing exposure to sunlight than alternative options are. In economics, Friedman might have us believe that assumptions are inclusive of a set of rules, but this approach is sure to breed misleading characterizations, as well as other problems. Moreover, the leaves’ presumed seeking may not be the motivating factor or even a motivating factor.
Whether or not leaves “seek” anything cannot be assumed based on their behaviors. We have no way to know the motivations of leaves, and explaining it in this way seems dishonest. Similarly, if an economist wants, they might make an assumption based on the assumption of political economy that V.38 of Mill’s “On the Definition of Political Economy; and on the Method of Investigation Proper To It,” makes. That someone chooses to go into investment banking because “mankind is occupied solely in acquiring and consuming wealth” is an assumption that can easily be justified. Popper points out that we can find justifications for things if we look for them, but that this does not make them true. It may be that Mill’s assumption is correct, but we have no way to know. Friedman’s characterization of the leaves is attractive in that it makes the case of the leaves more accessible. Similarly, Mill’s description of “mankind’s” motivation has some gravity that comes from its simplicity and potentially from its apparent obviousness. But these kinds of inaccurate characterizations risk breeding dangerous thoughtlessness, if not because of the appeal in their simplicity.
Something that we might agree is unattractive in science is an instance such that some people would be led to an erroneous conclusion, especially insofar as such a conclusion could lead to further erroneous conclusions. As an example, if someone concluded that lighting a boat on fire would surely be the fastest way to get them from the shores of Europe to the shores of North America and with most assured safety, they might conclude that the most efficient thing to do would be to build their ship out of the most flammable materials that they could find. In the economics realm, if we conclude that growth of income and wealth is what is what humans prefer, we might engage in policies designed to maximize growth. If we are wrong to conclude that growth is preferable, then such policies are erroneous but can also lead to unattractive policies based on that assumption and possibly lead to disastrous outcomes.
It could be attractive for an economist to say that the models are only meant to make predictions and that the assumptions do not matter, but surely this cannot be true. Friedman describes how some rules sometimes apply and some rules sometimes do not:
“Each occurrence has some features peculiarly its own, not covered by the explicit rules. The capacity to judge that these are or are not to be disregarded, that they should or should not affect what observable phenomena are to be identified with what entities in the model, is something that cannot be taught; it can be learned but only by experience and exposure in the ‘right’ scientific atmosphere, not by rote.”
There are two problems I see with this passage. First, it directly contradicts what Friedman says, on pages 33 and 34 of “The Methodology of Positive Economics,” about complexity and science. Friedman points out that science has demonstrated a tendency to be able to explain diverse and seemingly disjoint phenomena through simple theories. So if he is justifying the use of simplified assumptions in economics because science finds ways of simplifying things, then how is it that he is simultaneously defending the oversimplified nature of assumptions in economics by virtue of the world being too complicated? Which is it? Is the world actually complicated to a point that it prohibits simple explanation or can the world be explained by simple descriptions? The second, and more crucial, point is this: Popper clearly expressed a desire to avoid fitting theories to cases; Friedman seems to prescribe that we alter our theories such that some rules apply and some do not, based on the necessities predicated by the particular reality being faced. Here, Friedman advocates for loyalty to assumptions that are justified by their abilities to predict, but if the assumptions change based on the circumstances, is this not the very thing that Popper is opposed to? Further, Friedman makes quite a dangerous suggestion: he proposes that there is a level of discernment afforded by experience. This expertise coincides with the ability to correctly apply and judge the model, meaning that there is a small set of people who hold the key to the interpretation and application of this knowledge, many or all of whom, one supposes, might be inclined to protect the methodologies that they are taught and that they are inclined to engage in (as well as which earn them their livings). If we are to believe Friedman’s conclusions about the expertise needed in order to judge the propriety of the applications of these methods, the methods of this elite class can be insulated from some forms of criticism.
In order to dismantle the notion that assumptions are worth paying attention to, Friedman, on page 32 of “The Methodology of Positive Economics,” lists some of the ways by which a theory on a wheat market would need to be accurate in order to be fully realistic. There are two problems with his contention: the first is that the fact that we have a hard time accounting for conditions of the wheat market does not make those factors any less important; the second, and related, point is that this rationalization for incomplete theories allows a justification for not trying harder to obtain a more complete account of the relevant points, even if Friedman’s intention is not to discourage this work. We should grant to Friedman that the challenges in constructing a perfectly accurate theory are likely prohibitive and that Friedman is likely not intending to argue against efforts to improve theories. However, if we are to be as generous to Friedman as we possibly can, the least we can say is that he perhaps unwittingly discourages improvement of theories in terms of their assumptions and their completenesses. What is the value in a theory that is not truly accurate, compared to one that is? Friedman is right to point out that this is complicated, but this is good reason for not committing to assumptions — not for justifying their uses.
To contextualize this point, if Friedman thinks that we should adjust our models to fit our cases, as I demonstrated above, and if he finds it unimportant that we account for all relevant factors, then it is easy to see why what he advocates for is not necessarily better than what Marx or Freud produced. Some might assume that Friedman’s saving grace should lie in his reverence for predictive power, but the concept of predictive power in economics has its own flaws, and more to the point, predictive power is less convincing if the theory can be altered post hoc such that discrepancies can be explained away, “saving” the theory from the results. This is exactly one of the dangers in claiming that: 1) not all rules in a theory need apply to all cases and 2) you need be an expert to be able to determine when a rule or set of rules need apply or not apply.
The point about the problems with predictive power in economics is a subtler problem in Friedman’s treatment of assumptions. In previous papers, I have pointed out how it is reasonable that assumptions made in economics lead to distortions of markets such that markets are influenced by the assumptions, even leading to possible self-fulfilling prophecies. In those arguments, I have highlighted a few such cases, so I will not be restating my case here, except to make one fine point. Friedman points out at different points in “The Methodology of Positive Economics,” that economics must adapt to discoveries in psychology and that some advances in economics depend on additions and adjustments to psychology. With that in mind, I find it important to note that decision-making is not as significantly logical a process as we may have been led to imagine. A 2008 study by Soon, Brass, Heinze, and Haynes, titled, “Unconscious determinants of free decisions in the brain” shows that much of the decision-making process in humans is unconscious, which is cause for greater concern over the issues of distorted markets than I may have given credit for. Apart from that, it should be sufficient to say that treating assumptions in economics as though they are legitimate runs the risk of distorting markets.
This brings us back to Popper. As previously stated, it is not the case that Popper’s take on these issues is much better than Friedman’s. I would argue that Popper was simply better able to paper over some of the problems.
Popper states his case on induction on page 29 of “The Logic of Scientific Discovery”:
“Thus if we try to regard [a principle of induction]’s truth as known from experience, then the very same problems which occasioned its introduction will arise all over again. To justify it, we should have to employ inductive inferences; and to justify these we should have to assume an inductive principle of higher order; and so on.”
I am inclined to admit that I am not completely sure what Popper means by this, but of course, one can never know exactly what anyone means by anything, so I will propose that what might be intended is that an assumption that induction is valid requires that we presume that some inductions turn out to be true, and so on. For example: How do I know that all swans are white? All the swans I have observed have been white. Why does that mean that all swans must be white? We have seen that in cases such that all observations of something demonstrate a characteristic, they all continue to possess that characteristic, no matter how many observations we make. How do I know that those things will continue to possess that characteristic? Well, we have seen that this kind of induction has worked effectively in lots of cases in the past. How do I know that there will be more cases for which this kind of induction will rightly apply? Because more always seem to appear. How do I know that more will continue to appear? Because they always have… This may not be exactly as Popper intended, but the crux of his argument seems to be that, no matter how you try to justify induction, there is a step, that relies on pure guesswork, needed to go from the singular to the universal. This is no real logical cause for justification.
But in truth, what Popper proposes seems to amount to something like Induction Lite, which is to say that it is still induction, but does not make one feel as greasy. He goes out of his way to acknowledge that induction is invalid, while he proposes that we tentatively accept the findings of predictive theories, but this notion runs into the same kinds of problems as more easily recognized inductions.
What Popper proposes is pragmatic acceptance, even if “tentative,” of theories that are strenuously tested. This is inductive. If we test something an infinite number of times, the infinite-plus-oneth time might produce a divergent result. There is no logical basis for accepting that the observation of the singular — or even multiple — necessarily implies anything in the general. Acceptance of a theory’s validity assumes that the theory’s assertions hold up over all cases and will continue to do so, even if one acknowledges that there may come a point in time at which the theory will no longer hold, but this necessarily means that the theory cannot be accepted. It may be that there is a black swan out there. Further, it may be that the essential reality will simply change.
Perhaps Popper only means to justify the use of theories in order to make predictions that are falsifiable, but that does not preclude the inductive nature of accepting theories, even if a theory is only to be accepted until the point that it is falsified.
Friedman’s case for induction is sometimes more obvious. Earlier, I referred to pages 33 and 34 of “The Methodology of Positive Economics” to point out that Friedman claims that science is often able to simplify seemingly complicated and disconnected aspects of reality. He goes on to induce from the singular that the general must be true:
“If a class of ‘economic phenomena’ appears varied and complex, it is, we must suppose, because we have no adequate theory to explain them. …Any assertion that economic phenomena are varied and complex denies the tentative state of knowledge that alone makes scientific activity meaningful.”
Apart from Friedman’s contradiction concerning simplicity and complexity of reality and science’s and economics’ relationships to reality, Friedman is merely advocating for induction, here. If science on multiple occasions, has simplified the world, then economics must be able to do the same thing.
The problem of induction is one that has bothered philosophers for a long time. What seems to be the case is that the value in induction gets criticized, only for people to later come along, wanting to redeem induction. Why are we so attached to induction? I suppose that we complicate the matter by basically being inductive creatures. We seem to operate by forming patterns in our minds and operating on them, and this has led well enough to our survival. I believe that there is a strong case for why we might defend induction, based on what feels perfectly natural to us. That said, science is not everyday life, and I agree with those who want to hold science to a higher standard.
But if not induction, then what? I do not have a strong answer to this question. In fact, I do not propose that we need to do anything differently other than that I believe that we should acknowledge these problems in ways that are reflected in our methodologies. Because what is the alternative? To say nothing of Kuhn, Popper’s methodology has its problems, and they seem to be worth addressing. Popper’s proposal that the source of a conjecture does not matter does not seem completely true. Friedman’s arguments, and the responses he elicits, help to clarify this. One might presume that the sky will turn into pink cotton candy at 5am the next day, but it is unlikely that one could find a basis for such a presumption. While there are problems with presuming that the blue sky will remain a blue sky tomorrow, at least it has the virtue of being based on some reason, flawed though that type of thinking might be. So I have some sympathy for what Popper and Friedman are trying to accomplish. Still, it is hard to escape the feeling that they go too far in valorizing inductive techniques and in justifying them.