Professor Gigerenzer is Director of the Max Planck Institute for Human Development in Berlin. He is famous for his work on rationality, decision making under risk, and heuristics. His new book, Heuristics: The foundations of adaptive behavior (Oxford University Press), explains the importance of simple heuristics for everyday life and the world of business.
Eva: Can you give an example of such a simple heuristic?
Gerd: Many social scientists think of a decision as an optimization problem, for which people try to select the best solution based on full information. I think this does not reflect the way people make decisions in an uncertain world. For instance, people cannot use all information, simply because they do not have enough time. Fortunately, we can use shortcuts that work amazingly well, often better than models that use all available information. A good example is what we dubbed the ‘recognition heuristic’. We asked American students which city had the larger population, Detroit or Milwaukee. 60 % of the Americans got the answer correctly. When the study was replicated in Germany, 90 % gave the correct answer. How can one make better inferences with less knowledge? Precisely because the Germans could rely on the recognition heuristic: If you have heard of Detroit but not of Milwaukee, then you can infer that the recognized city is larger. The Americans could not use this rule of thumb, they knew too much.
Eva: If you are saying that such heuristics work so well, does that imply that the rare people who do optimize and use all information actually face a loss?
Gerd: Optimizing works best in a world where all risks are known for certain (as in lotteries, the stock-in-trade of decision research), but not necessarily in an uncertain world. An uncertain world is one in which not all alternatives, payoffs, and probabilities are known – the world of business, for instance. We have shown that heuristics are often more accurate and faster in uncertain worlds than optimization methods such as multiple regression and non-linear algorithms such as neural networks. The reason is that simple models tend to be more robust than complex models with many free parameters, and are less hurt by overfitting. The big challenge is to understand the nature of the worlds in which less is more.
Eva: In a recent article with Nathan Berg, you criticize ‘as if ‘ behavioral economics by saying that it does not add realism, only extra parameters. But to an economist, realism is not at stake- they want good predictions.
Gerd: Predictions are important, but we are more curious. We want a process model of cognition, instead of predictions from a black box. Bringing psychological realism to economics was a goal of behavioral economics, but its most prominent theories, from prospect theory to inequity aversion, have been content with adding a few parameters to the neoclassical “as-if” models. You should also be aware of how the word prediction is used in economics; often it is not about forecasting, but rather about fitting parameters to a dataset without predicting anything.
Eva: I recently saw a talk by Toshio Yamagishi, who said that there is one parameter that does predict what people do in different experimental games: social risk aversion. He found that the extent to which people tried to avoid socially ambiguous situations was predictive of what they do in other games. Does that fit with your view?
Gerd: I haven’t heard that talk. But it is certainly a good idea to test a model by predicting behavior across different games. Much of the work only fits parameters to each new game or study. Even better, one should test competing models, not just one.
Eva: So how do business people, who receive direct financial feedback on their decisions, react to the heuristics approach you propose?
Gerd: Business people are extremely interested. Optimization theories are of little help to them: they have to survive in an uncertain world. My book “Gut Feelings” got the Swiss award for the best business book. Recently, the department of economics at Bielefeld University invited two successful local entrepreneurs and two academics, Reinhard Selten and me, for a debate. The consensus between the four of us was that much of what is useful for a company is not taught, and much what is taught is not useful. The business entrepreneurs emphasized that they have to rely on heuristics and gut feelings and would like to get some guidance on these matters. For instance, they face the problem to predict which of their customers are active and which are inactive in a database of ten thousands of customers. For this problem, simple heuristics that rely on only one good reason (e.g., if a customer did not made a purchase in the last nine months, out, otherwise in) have been shown to predict better than the complex Pareto/NBD model featured by marketing science. At the end, all four of us told the department to teach smart heuristics. The audience loved it, the faculty not quite as much. The president of the university, an engineer, invited me on the spot to give another lecture to the entire university. We still have a long way to develop economics so that it becomes more relevant for real business.