William Easterly Reviews Daniel Kahneman’s Thinking, Fast and Slow

Daniel Kahneman has spent his career studying how people behave both rationally and irrationally, for which he has won a Nobel in economics and largely founded the field of behavioral economics.  As development economist William Easterly notes in his Financial Times review, there “have been many good books on human rationality and irrationality, but only one masterpiece. That masterpiece is Daniel Kahneman’s ‘Thinking, Fast and Slow’.”  (The FT review appears to be public.)

Kahneman presents our thinking process as consisting of two systems. System 1 (Thinking Fast) is unconscious, intuitive and effort-free. System 2 (Thinking Slow) is conscious, uses deductive reasoning and is an awful lot of work. System 2 likes to think it is in charge but it’s really the irrepressible System 1 that runs the show. There is simply too much going on in our lives for System 2 to analyse everything. System 2 has to pick its moments with care; it is “lazy” out of necessity …  Of course, Kahneman is one of the fathers of the field of cognitive biases, and most of the book is indeed spent on the mistakes made by System 1. We get probability and uncertainty terribly wrong, usually leading to overconfidence and mistaken decisions. We react to identical situations differently depending on what is already on our minds. Even worse, we don’t know what we don’t know.

The observation of systematic cognitive bias leads to calls for policies of paternalism, expert guidance to correct for mass cognitive biases.  But as Easterly emphasizes, the experts also have systematic cognitive biases:

Kahneman’s endorsement of “libertarian paternalism” contains many good ideas for nudging people in the right direction, such as default savings plans or organ donations. But his case here is much too sweeping, because it overlooks everything the rest of the book says about how the experts are as prone to cognitive biases as the rest of us. Those at the top will be overly confident in their ability to predict the system-wide effects of paternalistic policy-making – and the combination of democratic politics and market economics is precisely the kind of complex and spontaneous order that does not lend itself to expert intuition.

I just purchased this book, on Easterly’s recommendation.  But on the basis of the best-seller Nudge, which argued explicitly for libertarian paternalism, I agree with the criticism of “libertarian paternalism.”  The problem with Nudge, as many critics pointed out, was that it had one or two terrific examples of where re-setting the default behavior would have a huge and positive effect on outcomes without compromising liberty very much.  Everyone properly focuses on the default opt-out of the company 401(k) plan, rather than opt-in, for example.  Fair enough.  But many of the other applications seemed to critics, and to me, as “shove” rather than “nudge.”  So I will be curious to read this book and see if it seems to me any different.

Comments are open.  I am curious whether commenters who have read this literature (whether this book, or Nudge, or others in this category) agree that there is a problem with “shove” over “nudge.”  I am also interested in how commenters address the question of experts having their own cognitive biases – ones that are just as powerful, and which often have large consequences for everyone else, as those of the masses whom they manage.  Bill Easterly, as those who know his work would note, has reasons for wanting to emphasize that expert biases can easily be far more important than those of ordinary people.  But in that case, how does one get out of the cognitive bias-skepticism regress, in which there is always another layer of skepticism that undermines the ability to act?  Easterly does point to a particular feature of Kahneman’s views – the formation of judgment, the ability to use the lessons of “system 2” to train “system 1.”

Finally, we might ask whether and in what ways that differs from what Aristotle and deliberate inculcation of character to the “proper” formation of judgment.  Skepticism, Aristotle and the virtue-ethicists today might say, is essential – but ultimately too easy.  There’s always another layer of skepticism; eventually it becomes a paralyzing cost, rather than a benefit, to action.  Skepticism is easy, judgment is hard.

(Update:) Thanks for the thoughtful comments.  I am reminded of an excellent essay (it is short, 14 pages, and clear, elegant, and accessible) by NYU’s Richard Pildes a couple of years ago on Cass Sunstein’s body of work which, of course, is central to this discussion.  Here at SSRN, and here is the abstract:

In both constitutional law and public policy, Cass Sunstein’s work has entailed a search for the largest common denominator that justifies government action. In constitutional theory, Sunstein developed the concept of “incompletely theorized agreements” as a model for how judges ought to decide cases. In public policy analysis, Sunstein’s work has reflected a similar commitment to maximizing consensus and reducing conflict. While Sunstein’s conception of minimalist adjudication has been thoroughly explored, less attention has been paid to the underlying political vision that structures his view of the proper role of the state and the desirable forms of public policymaking.

In this tribute, I explore and challenge the structure of Sunstein’s political vision. Two ways of seeing this vision exist. The first is the way in which Sunstein presents it: as a profound new alternative capable of transforming current politics and transcending political polarization and conflict. Sunstein himself calls his vision “a real Third Way,” a post-partisan conception that provides a synthesis of Franklin Delano Roosevelt’s New Deal liberalism and Ronald Reagan’s new conservatism. The second way is almost diametrically the opposite. Perhaps this conception actually reveals how chastened and minimalist political aspirations are limited to being in our era.

Based in behavioral law and economics, the centerpieces of Sunstein’s political vision are default rules and information disclosure. This is a vision focused on changing the means by which government acts. This focus, however, then raises the question: how much can or should politics focus primarily on the means of government action, rather than what ends government ought to pursue? Or, to the put the question in terms of Sunstein’s own stated ambitions, can it really be the case that the major political critique of the New Deal that was effectively launched in the Reagan years was simply a critique about the means of public policy, as opposed to the proper role of the state and the ends for which government ought to act? Should we see Democrats and Republicans, liberals and conservatives, as so divisively polarized today merely because they disagree about what means government ought to use in pursuing policy objectives – objectives that, we are presumably to believe, all sides actually share? If this vision actually is the “real Third Way” in contemporary politics, it is worth asking what that tells us about the possibilities for democracy today.

One of the important questions about Sunstein’s judicial theory as distinguished from the larger political theory is whether “minimalism” should govern in both contexts.  My own view is that this is the correct view of how a judiciary should act and decide – but for the important reason (drawn less from political theory of liberalism than Weberian social theory of legitimacy) that courts and judges are “embedded” within structures of politics and society, cabining their necessary counter-majoritarianism within larger majoritarian structures and within a larger society in which these institutions have social legitimacy.

But those concerns are not the same as those that majoritarian political institutions.  The executive in particular has to address in many matters in which a larger strategic vision for how to go forward is required.  National security is one of them; success in national security requires commitment to a longer term and more “enveloping” strategy than minimalism entails.  Minimalism counsels small, tactical, incremental, cautious moves – which is wise for judging but can be disastrous in matters in which a strategic vision is the only way to success.  The problem with the strategic vision is that it requires making commitments and guesses about long run things that will not necessarily command agreement; it involves bets on the future, in which some bets rule out others.

That’s the nature of strategy.  As Philip Bobbitt once remarked to me about fighting terrorism as an example of national security as strategy, the problem of applying cost benefit analysis in any serious way to these problems, as a consensus way forward, is that the serial minimalism of CBA leads to narrow incrementalism.  It emphasizes exceedingly linear thinking in a strategic realm in which offensive victory is usually defined by envelopment in time and space.  This kind of serial minimalist thinking precludes grand strategy of the kind that political institutions in a democracy must take on, in a way that judges do not.  While it might be true in special contexts that the best “grand strategy” is, in fact, serial incrementalism, much of the time it reflects not a strategic decision but instead adherence to a method that is, in Bobbitt’s phrase, “relentlessly tactical.”  I’ve sometimes called it “serial catastrophe prevention” as a “strategic” policy that really is merely tactical.  E.g., everyone can agree that we need to protect the airport from the next terrorist attack and so we can agree on concrete barriers to prevent the suicide bomber driving a vehicle.  But we can’t agree on a longer run strategy beyond that minimal tactical level agreement aimed at serial prevention of the “next attack” to come up with any larger, on-offense strategy for ending the threat, or even that we ought to seek such.

I’ve framed it here as a problem for strategy in national security, and I’ve also framed it here as a question of strategic bets where the desired strategic outcome is not itself at issue; the problem is means to those ends.  There are lots of areas outside of national security in which long run policy strategies are crucial, and require big and not always minimal assumptions.  And, as a separate issue, there are also many areas in which there will not be consensus around the strategic ends themselves – different conceptions of values.  I don’t think the answer can always be to go for the minimalist approach; it too easily produces the worst of all worlds.  In these matters, that’s why we have democratic and majoritarian process – not purely majoritarian by any means, but not consensus either.  I talk about some of this stuff in a frankly not well organized article – my initial foray into some of this stuff in the national security area, but then I got pulled into other things, like drones – at SSRN here.

My article is long and alas confused in places, so I’d suggest Rick Pildes. But here is the abstract to my piece:

This 2007 article (based around an invited conference talk at Wayne State in early 2007) addresses risk assessment and cost benefit analysis as mechanisms in counterterrorism policy. It argues that although policy is often best pursued by agreeing to set aside deep foundational differences, in order to obtain a strategic plan for an activity such as counterterrorism, foundational differences must be addressed in order that policy not merely devolve into a policy minimalism that is always and damagingly tactical, never strategic, in order to avoid domestic democratic political conflict. The article takes risk assessment in counterterrorism, using cost benefit analysis, as an example of a foundational disagreement that cannot easily be elided. Examining an extreme, indeed crude, recent example of cost benefit analysis applied to the risks of terror and the costs of counterterrorism – John Mueller’s widely noticed Overblown – the article suggests that cost benefit analysis, at least applied in this way, runs roughshod over other important values in counterterrorism policy, such as justice, but in addition, makes radical yet unstated assumptions about what cost benefit analysis seeks to compare in establishing counterterrorism policy or estimating the risks and costs of terrorism – unstated assumptions that, in fact, assume the conclusion. The article notes that cost benefit analysis tends to promote a policy-minimalizing “event specific catastrophism” – seeking above all to prevent simply the next, serial terrorist attack, with however no greater strategic vision. Indeed, the article says in conclusion (as Philip Bobbitt has noted) cost benefit analysis is “relentlessly tactical,” not strategic; it also tends toward serial ‘event specific catastrophism’ as its analytic frame; and it is a method of evaluating proposed courses of action, not generating them, and hence promotes a strategically questionable tendency to reaction as a response to terrorism. This article presents these ideas in brief fashion, however, as the first draft in a larger project on cost benefit analysis and counterterrorism, and it does so by reference to a book that is unabashedly crude in its approach to both cost benefit analysis and terrorism/counterterrorism. The critical project will extend beyond this particular article, which is in effective a a first pass at developing a critique. It is also an article that does not extend beyond events of early 2007 (when the original address was given) and should be read in that light.