This is the second of a series of responses by Mario Rizzo and Glen Whitman, authors of Escaping Paternalism, for my Book Club on their treatise.

Inclusive Rationality and Falsifiability

In the first installment of the book club, Bryan raised questions about the falsifiability of inclusive rationality, and he wonders if we can offer a concrete example of a potential falsification.

This is an issue that we wrestled with as authors (as indicated by the passages Bryan quoted).  We don’t wish to present our position as tautological.  We do think that irrationality is possible, even under our permissive definition of the term.  But we also think finding examples that are both systematic and undeniable is extremely difficult.  With apologies, we can’t explain why without delving into the philosophy of science.

In most philosophies of science, even Karl Popper’s, it is not true that everything is directly falsifiable.  Instead, scientific research is typically governed by a paradigm (Kuhn), research programme (Lakatos), or metaphysical research program (Popper).  The purpose of such a programme is not to be directly falsified, but to generate ideas and falsifiable hypotheses.  We believe “inclusive rationality” to be such a programme, and it incorporates many subsidiary questions with testable implications:  whether (and how much) people learn over time; whether (and how) people construct their environments to shape their behavior; whether (and how) people adopt regimes of self-reward and self-punishments; whether these strategies are effective at changing behavior; and so on.  Bryan’s own “rational irrationality” is another example.  Many of these hypotheses have passed the test, but it’s possible that some won’t.  If some of these hypotheses were falsified, the programme could still survive.  But with enough falsifications, it could in principle be overturned in favor of some alternative programme.

Aside from its being a research programme rather than a simple hypothesis, there is another barrier to the direct falsification of inclusive rationality, and that relates to positive versus normative/welfare analysis.  If we just want to know whether people behave in a certain way (do demand curves slope downward?), then no problem – test away!  Likewise, if we just want to know if people behave consistently with a given model (do they act as if maximizing a Cobb-Douglas utility function?), again there’s no problem – test away!  But once we get to normative/welfare issues, “as if” is no longer good enough.  The explanation must be accurate in a more thoroughgoing way.  We have to know something about what specific people really want.  And this, it turns out, is extremely difficult to get at.

Neoclassical economics “solved” this problem via the presumption of preference revelation.  But if, for the sake of argument, we agree with the behavioral economists that actions don’t necessarily reveal true preferences, then how can we know?  To repeat a line from the book that Bryan quotes, “If actions do not reveal preference, then what does?”  It is a conundrum that we do not purport to solve.  We should note that the behavioral paternalists have not solved it, either.

To summarize, the scientific method, great as it is, does not always answer the precise questions that interest us.  Some questions lie beyond its reach, at least for now.  But we can use the scientific method to explore a set of related questions.  We don’t know what people really want, but we can investigate to see what kind of tools they use – or don’t use – to shape their behavior, as well as how effective those tools are.  It turns out that people don’t behave consistently with the old neoclassical model of rationality – but nevertheless, their behavior strongly suggests an ongoing process of trying to achieve their purposes as well as they can, given both environmental and cognitive constraints.

We still owe an answer to the question, “Can you provide a concrete counter-example to inclusive rationality?”  Given the need to get inside someone’s head to learn what they really want, a psychiatrist might be able to offer better answers than we could.  She might point to specific clients who commit the same error again and again, yet cannot manage to break out of the destructive cycle.  Then again, their going to a shrink suggests at least a meta-rational desire to change.

Bryan offers the example of opioid addiction, and we think it might work.  Not in the sense that it’s irrational to take opioids – the initial decision is surely quite rational for many – but that for some users, an emergent addiction may progressively reduce their ability to engage in the type of self-management behaviors we lay out in the book – in essence, robbing their inclusively-rational toolkit of its tools.  The concrete evidence might involve proof that opioid users tend to exhibit few or none of these behaviors, or that these tools fail for them at a much higher rate.  We are far from experts in this field, so we would not want to commit ourselves to this position, but it seems reasonable.

Returning to the main point, we view the practical difficulties of ascertaining inclusively irrational behavior, not as an unnecessary limitation of our analysis, but as a reflection of a fundamental epistemic problem: human beings are purposeful but their purposes are not easily identified. Taking off our economist hats and putting on our policy hats, it seems to us that policymakers ought to adopt a presumption of inclusive rationality. The opposite presumption of irrationality would surely be vastly over-inclusive and a threat to liberty.