Tim Harford is the finest expositor of economic ideas today. That is a view I have held for a long time. While his first few books stayed within the economic view of puzzles his last two have moved beyond that tradition. Adapt looked at how failure breeds innovation. And now he has produced Messy a book that celebrates, well, mess. Now I would have loved to be snarky at this point and call the book a mess (ha ha) but it isn’t. It is a beautifully organised set of stories to remind us how perfection is the enemy of the good and optimisation in the short-run can harm you in the long run. It is not something I can easily argue against since my book, The Disruption Dilemma, made essentially the same point albiet confined to the less sexy issue of management.
Let me, however, do more than other reviewers have done and honor the work by taking it seriously. The stories are all very well and good but what are we to make of all this? If we treat this as an economist would Harford is saying that if we have an objective function, B(1-m) that is a function of the share, m, of ‘mess’ (where m = 1 is complete disorganisation, or maximum entropy, and m = 0 is complete order), then B is not monotonically increasing. That is, even if we prefer complete organisation to complete mess, B(0) > B(1), there may be a point m* which maximises B; in other words, rather than subtitle his book “The Power Disorder to Transform our Lives” we would subtitle it more accurately that “Some Mess Can be Optimal” (and just watch the books fly off the shelves).
Even there the argument is somewhat more subtle because in Harford’s mind there is a role for time and for uncertainty. B arises in each period and we choose m at each point of time. Most of the time, nothing exciting happens, and it is best to choose m = 0 to maximise B. But sometimes, with probability s, something bad happens and you need to do something that requires m > 0 in the past. In this situation, the best course of action is to keep m > 0 so as to deal with the exciting event.
This is abstract but let’s put it in concrete terms as Harford did in this except in The Guardian. In that except, Harford examines automatic piloting systems that allow pilots to relax during flights. The interpretation of these systems is that they make pilot life simple (setting m = 0) which is, of course, great for any given flight. However, what if the automated system shuts off either accidentally or by design as it did for the doomed Air France Flight 447 a few years ago. Because the pilots lacked experience, they had trouble (in this case fatal trouble) in dealing with the situation of flying through a storm without automation. How might they have got that experience? By spending more time flying the plane. In other words, setting m > 0. Harford, like many others, sees this has a problem arising from automation — people lose skills.
But what are we to do here? How high do we have to set m to avoid the issues in Flight 447. There are millions of flights where the pilots get the benefit of B(0). What level of m will be required to move the needle on this small probability event and will the costs be too high? Harford doesn’t tell us and he doesn’t actually even phrase the question. After all, what we are talking about is trading off millions of pilot flight hours and utility with the lives of a couple of hundred passengers (and the utility reduction we all get from flying when we know it is not perfectly safe). No one other than economists like to even consider those trade-offs.
My point here is that I am not convinced on automation. To be sure, I agree that de-skilling is real and we lose options as a result of it. I am just not convinced that it is worth not using the technology because of it. The problem is that the shape of B matters. In particular, it may be multi-peaked with what we might call an “uncanny valley” where a little more mess does not good but lots of damage. That means we might not be talking about a little more mess in our lives as Harford claims (relying implicitly on concavity) but instead trading off a lot of mess versus a lot of order.