Aaron Brown

All Things Are Difficult Until They

Become Easy
Lessons learned from which temperature stays constant, as well as the set-
ting that will cause the temperature to move to the
Dorner’s The Logic of desired level at a reasonable rate. That allows you to
choose your next setting, after which you wait five
Failure. minutes again and recalibrate. This process should

By calflier001 [CC BY-SA 2.0 (https://creativecommons.org/

allow you to iterate to stability at the correct tem-

T
perature with a few calm, calculated moves.
his column’s title is from the great Note that this is a simple situation, with one

licenses/by-sa/2.0)], via Wikimedia Commons

medieval Persian poet Abū-Muhammad control variable, one output, a linear response,
Muslih al-Dīn bin Abdallāh Shīrāzī, and a clearly defined goal. Participants were under
better known by his pseudonym no stress, had full control, and the process had
Saadi. been fully explained to them. There was only one
The good participants differed from the bad unknown, the sensitivity of the rate of temperature
ones in that they tested their hypotheses. The bad change to dial movements, and it could be quickly
participants failed to do this. For them, to propose and easily inferred from simple experiment.
a hypothesis was to understand reality; testing that Now compare this to real situations with many
hypothesis was unnecessary. Instead of generating potential control actions, complex responses, and
hypotheses, they generated ‘truths.’ multiple goals. Add stress and lack of information,
The quotation above is from one of the most Virtually everyone – and participants were multiple hard-to-estimate unknowns, and multiple
important risk management books, The Logic of drawn from a range of populations including decision makers with complex relations and con-
Failure, by Dietrich Dorner. Two of the author’s engineering students and quantitative academics – flicted interests. What chance is there that policy-
experiments in particular are worth memorizing ruined the meat, and ruined it in the same way. Even makers can correctly calibrate the optimal amount
and considering before making any complex plans. with repeated practice, few of them learned to avoid of a carbon tax to reduce carbon emissions in
One is a computer simulation of a thermostat in the error. The refrigeration dial was first turned up, cost-effective ways, the proper level of a minimum
a meat locker. Simple mechanical devices have per- as the temperature was too high. When the tempera- wage, or the level of interest rates that supports
formed this task reliably for centuries. The trick is ture did not immediately start to fall, the dial was robust economic growth without excessive infla-
to increase refrigeration when temperature is high, turned up more. Participants continued to turn it tion or asset bubbles? If decision makers cannot
and decrease it when temperature is low, to keep the up until the temperature was at the desired level, at individually keep meat from spoiling in a computer
meat cold but above freezing. Prolonged high or low which point it fell rapidly, more rapidly than it had simulation, how can they be expected to perform
temperatures ruin the meat. risen initially. The refrigeration was then turned collectively in anything other than disastrous ways?
In this simulation, participants were shown a down, and turned down more until the temperature The second revealing experiment was more
temperature gauge and a refrigeration dial, the latter was at the desired level but rising even more rapidly. complicated. Participants had a simulated village to
with no markings. The story was that the thermo- Clearly, the correct approach is to note the rate manage with multiple control options. They were
stat had broken and the refrigeration level had to be of temperature increase at the initial dial setting, not given explicit goals. Nearly all participants man-
set manually until it could be fixed. Initially, the dial then turn up the refrigeration by any amount aged to kill everyone within ten simulated years.
was set to the middle position, and the temperature (there’s no way to guess a reasonable amount as The paths in this experiment were predictable
was too high and rising. Participants were told that the refrigeration dial has no markings). Then do as well. Participants began by noting and solving
there was an approximate five-minute lag between nothing for five minutes – that’s the hard part. problems. For example, death rates from disease were
changing the refrigeration level with the dial and After five minutes, you can note the change in rate high, so many participants began by delivering vac-
seeing the full effect of the temperature difference of temperature change for the amount you moved cinations and other health services. Of course, this
on the thermometer. the dial, which allows you to estimate the setting at caused the population to increase, so participants had

8 wilmott magazine
to dig more wells and use farmland more intensively. minor other goals, rather than minimizing the total Roosevelt remains a hero, not because it’s clear that
This quickly led to resource depletion and famine. number of long-term deaths. his policies shortened the Depression or mitigated
In the words of the author, they: “…acted without Another important lesson, related to the per- its bad effects, but because he took innovative and
prior analysis of the situation; failed to anticipate side ceived need to attack every problem immediately, is aggressive action. Herbert Hoover was a smarter and
effects and long-term repercussions; assumed the the creation of negative goals. “We must prevent” or more experienced man, and a better person, but his
absence of immediately negative effects meant that “we cannot allow” are very dangerous words because engineering training led him to support only actions
correct measures had been taken; and let overin- they may define undesirable states of the world. If that he was confident would help.
volvement in ‘projects’ blind them to emerging needs you can envision an attainable, sustainable positive Another example is that, even with well over a
and changes in the situation.” They assumed that outcome, it might at least be achievable, and might half-century of overwhelming evidence that active
every problem demanded a solution, and when those be good if you achieved it. If you instead rule out managers overtrade and charge too much for their
solutions created new problems, participants brought large classes of outcomes without understanding the activity, and individual investors lose by switching
in more solutions until they were overwhelmed. tradeoffs, you may be left with no good choices. Since strategies, most investors still switch among active
Here, too, the proper solution is clear. You the 2007–2009 crash, a lot of undertakings proceed managers: chasing prior returns and closing barn
watch for a while, to figure out how things are from the assumption that we must avoid future crises. doors after the horses are gone. It is nearly impossible
working and what the equilibrium forces are. You But there has never been a dynamic, inclusive econo- for most people, even with their life savings literally at
determine a plan for a better equilibrium situation. my that did not have periodic crises. It might be pos- stake and clear empirical evidence, to avoid reacting to
That plan involves an optimal population – which, sible to design one, but probably not by starting with market events, recent performance, and good stories.
in turn, implies a relation between birth rates and the negative goal of avoiding all crises. To paraphrase Maurice Switzer,1 in risk man-
death rates – and the correct amount of supporting These issues are exacerbated by politics. Elected agement it can be better to do nothing at the risk
resources, which must be sustainable. Tradeoffs officials and their top appointees cannot afford to of allowing catastrophe than to take ill-considered
have to be made. Different people would choose do nothing when problems arise. Even if they do not action and remove all doubt of it. However, with
differently, but no one thinks it’s good to kill off the know which measures will help, in fact, even if they foresight and careful observation, it is sometimes
entire population. And you can’t just jump to your are confident that any reaction will have net negative possible to take calibrated actions that improve the
desired situation; as with the meat locker, you have consequences, they must be perceived as taking chances of acceptable outcomes. These can even be
to make small changes and wait to see equilibrium strong action. “We can’t just stand by and let...” is random actions, as in the first movement of the dial
effects, with the additional complication that you never true; you can always stand by, and the alter- in the meat locker simulation. The actions will near-
will learn more about interactions and what villag- native that results from action may be worse. Doing ly always start earlier than intuitive managers prefer,
ers want as you try to improve things. Finally, you nothing creates the perception of not caring. change more gradually, and reverse sooner. They are
should think hard about whether or not it’s worth Andrew Mellon lives in infamy for his 1931 paths of exploration to reach a tolerable state rather
trying any improvement at all; after all, things are advice in reaction to the onset of the Great than checklists to ensure success. Moreover, they
initially in a tolerable equilibrium and any attempt Depression: “Liquidate labor, liquidate stocks, liq- nearly always require advance consideration and
at improvement risks uncontrollable outcomes. uidate the farmers, liquidate real estate.” One reason planning; brilliant ideas that occur to people only
The author supplements these simulation exper- is the unfortunate modern connotation of “kill” for after disaster are unreliable in execution.
iments with analysis of real crises, from Chernobyl “liquidate.” But what the sentence actually means is Failure is rarely an unpredictable, exogenous
to AIDS. One important point is that people are take no government action and let the price of labor event; there is a logic to failure. Risk managers can-
rewarded for the short-term effects of their actions. fall until there is full employment; stock prices fall not predict or prevent disasters, but they can try to
That means that ignoring safety protocols that are in until investors are willing to commit new capital; replace the logic with something that might work.
place to prevent rare but catastrophic events gener- agricultural prices fall until supply equals demand;
ally leads to people being perceived as efficient and and real estate prices fall until there are buyers for endnotes
having practical knowledge that trumps abstract foreclosed properties.
learning. Solving problems quickly is valued, even Whether short-term shock treatment in fact In 1906 Maurice Switzer wrote: “It is better to remain
silent at the risk of being thought a fool, than to talk
if the solutions create more problems in the future leads to the least amount of total pain and the quick-
and remove all doubt of it.” Variants of the quote have
or entail low-frequency, high-impact risks. The est and most robust recovery is controversial. It does
been attributed with scant evidence to Mark Twain,
operators at Chernobyl, for example, were regarded not mean that the government cannot work to assure Abraham Lincoln, and others, and there is even Biblical
as the best team in the Soviet Union precisely for adequate minimum resources to everyone, just that version: “Even a fool, when he holdeth his peace, is
their efficiency and rapid problem-solving. Early it should not try to keep prices up as a means of sup- counted wise: and he that shutteth his lips is esteemed
research and public health efforts to control AIDS porting people or institutions. However, the wisdom a man of understanding.” (Proverbs 17:28, King James
were directed at making marginal improvements for of that approach is irrelevant because no politician translation).
relatively few people and accomplishing relatively who advocated it would remain in office. Franklin

wilmott magazine 9