A conversation yesterday with Pete Kaminski brought to mind one of the more interesting books I’ve read in the last few years.
The book is called Sources of Power: How People Make Decisions, by Gary Klein, a social science researcher and consultant. The book is badly named. It’s not a neo-Machiavellian business manual; it’s fascinating social science research on how people (really) make decisions, in contrast to how we think people make decisions, influenced by our mental models of people as computers.
Spoiler: people make and share decisions with stories
Klein started out looking to find experimental evidence for the way people make decisions, based on long-established academic work on decision theory. The conventional theory predicts that people compare alternatives, and make a rational choice among alternatives. The conventional wisdom is bolstered by a mental model of the brain as a computer, which methodically compares choices in a decision algorithm.
Klein’s group did fieldwork among experienced professionals who make frequent life-and-death decisions as part of their job — intensive care nurses, fire-fighters, military tank commanders. What they found, to their suprise, is that people’s decision-making process is nothing like the textbook model.
Klein made two interesting discoveries — about how experts make decisions, and how they communicate with co-workers.
Expert decision-makers don’t analyze and evaluate alternative options. Instead, they visualize a picture of the situation in their minds eye, envision a single course of action, and run a mental movie carrying out the action. If the scenario plays out, they quickly act. They never consider a second alternative. If there is a problem with the mental scenario, they visualize a second course of action, and implement it instantly if the mental movie plays through.
The experts mental images are very rich; an experienced fire-fighter will imagine the structural stresses on various parts of the building based on the appearance and sound of the fire. An experienced intensive care nurse will visualize a diagnosis based on few clues about a baby’s skin color, breathing, and vital signs. When the experts visualize the image, they don’t think about the components, one-by-one; they quickly identify “points of leverage” — what aspects of the system are most subject to change.
Post-mortem analysis shows that the “first-choice” decisions by experienced experts are most often right; the second-best choices identified in case study analysis are less good than the expert’s first choice.
Klein’s group found that the rational, comparative method of decision-making was followed, not by experts, but by novices first learning the field. Despite the fact that novices were following a more analytical method, the highly considered choices of novices were more likely to be wrong than the expert’s first choice.
When experts troubleshoot a problem, they also diverge from the textbook model. Instead of analyzing the components of the problem, they tell a story to explain the various facts. When expert mentors teach novices, in the field and by the water cooler, they communicate their lessons by means of stories.
I particularly liked the way Klein summarized the difference between computerized decision making methods (define a closed-problem space, break the space into sub-spaces, search the sub-spaces) and human decision-making (pattern-match based on experience, simulate scenarios using imagination, use an iterative process to reach goal, change the goal if necessary; and communicate learning using stories.
However, there is a significant limitation to Klein’s observations.
The decision-making methods that Klein describes enable experts to make quick, effective decisions under pressure. But those skills support preset strategies. The commander of a fire-fighting unit can save the lives of his team and extinguish the fire; but he can’t say whether the Forest Service should have a fire-suppression policy in the first place. Sometimes the intuition of experts can lead them astray, when the situation calls for a truly novel response.
By the way, I was referred to the book by my friend David Blank-Edelman, the author of Perl for System Administration, who used the book’s concepts in a conference presentation on the similarity between network administration and veterinary medicine.