Sometimes Decision Making Requires Thinking in Reverse

In “Decision Making: Going Forward in Reverse”, Einhorn and Hogarth discuss how analyzing past information can be an important way of dealing with the future. The authors discuss how managers are constantly using both of these types of analysis, they do not understand the differences between the two. Since the two processes need to be handled differently, managers often make bad decisions when they don’t understand how they are supposed to handle them.
The authors claim that when thinking backwards decision makers must begin by trying to find a cause and effect relationship. They describe how most decision makers begin this investigation by trying to identify an unusual event or occurrence that may explain the current situation. They then go on to analyze their theories and determine whether or not they seem to adequately explain the situation or resolve the problem. However, there are a number of possible explanations that may explain the situation, so it is difficult to identify which, if any, of the explanations is appropriate. Therefore, they usually try to experiment to narrow the possible explanations and speculate as to which one is most appropriate.
The authors discuss how decision makers often look for links between causes and effects by looking for similarities between them. They use the example how in early medicine physicians believed that jaundice would be cured by a yellow remedy. Obviously, some of the associations that end up coming up end up not making sense. Therefore, people end up considering different levels of association between cause and effect based on cues. The four categories are: causes come before effects, causes and effects occur at approximately the same time, causes vary with effects, and causes may resemble effects. These cues may provide decision makers with a sense of which direction they should go in to investigate.
The authors suggest several different approaches that decision makers can use to think backwards more effectively. One approach is that they can use a number of different metaphors, to guard against the flaws associated with any single metaphor. Decision makers should also use more than one cue, and should also sometimes try to promote creative thinking by going against the cues. Decision makers should also consider how many connections and possible links there are between a cause and effect, and consider that the more links between the two the weaker the chain may be. Finally, decision makers should consider different explanations. These theories should be tested experimentally when possible. However, when experimentation is not possible, decision makers can imagine the situations and circumstances involved to get a better idea of what might happen and how the cause and effect might work.
The authors spend the second half of the article talking about how decision makers think forward. They discuss how most people have a hard time doing so accurately. Interestingly, people tend to have more faith in human judgment than in statistical models, despite the fact that statistical models tend to be much more accurate. Nonetheless, there are a number of reasons why humans tend to not have faith in statistical models.
The primary reason that humans tend to be skeptical of models is that they cannot adequately understand all the variables and the relationships between them. The errors produced by models tend to show up consistently, as opposed to the errors produced by humans which vary. As a result, the inaccuracies in models tend to be more visible and stand out better in the minds of humans. Humans often try to extrapolate their own patterns which they believe can be more accurate than the models they would alternatively use.
Decision makers also are skeptical of models because they believe that they these models are static. In order to fight this bias, models need to be updated and improved based on new information and links that have been learned. One problem that the authors suggest is that it is important to separate accurate predictions from the effects caused by those predictions. They illustrate through the example of the president of the United States making a statement about the economy going through a recession. If a recession does result, it is important to understand if the president actually predicted the recession correctly, or if his statement itself caused the recession.
The final reason why humans tend to avoid using models is that models are often thought to be more costly than they are of value. The authors argue that even though it is difficult to measure the cost-benefit tradeoff of using a model, the models usually will eventually be worth the cost if they are used enough. They illustrate this by showing how AT&T used models to reduce bad debt which was costing them over $100 million a year.
This article was similar to “Competing on Analytics”, (Davenport et. al,2005) because both articles discussed how statistical models can be an enormous value to decision makers. Both articles stress the limitations of human judgment, and how we must consider models that may be capable of producing superior results. Similarly, the article “Automated Decision Making Comes of Age” (Davenport, 2005) discusses the use of computer operated decision making and how it can yield many benefits that cannot be realized from human judgment. All three of these articles stress how even though carefully constructed models can yield better results than human beings, they are still not widely accepted and are met with skepticism by decision makers in the real world.