El debate acerca del efecto facilitador en problemas de probabilidad condicional: ¿Un caso de experimentación crucial?

Abstract

Se ha mostrado en la literatura de Psicología Cognitiva que las personas generalmente tienen dificultades para resolver problemas de probabilidad condicional . Sin embargo, también se ha mostrado que, bajo ciertas condiciones, las respuestas mejoran de manera significativa. Desde mitad de la década de 1990 hubo un gran debate acerca de cómo dar cuenta de dicho efecto facilitador . Se han propuesto dos hipótesis rivales, la hipótesis de frecuencias naturales que dice que el efecto facilitador se debe a presentar la información de manera frecuentista, y la hipótesis de conjuntos anidados que dice que dicho efecto se debe a la clarificación de las relaciones de conjunto del problema. En este artículo intenta- remos clarificar el debate y analizar la evidencia empírica relevante. La pregunta a responder es la siguiente: ¿Se ha producido alguna experimentación crucial en favor de alguna de las dos hipótesis? Nuestra respuesta será negativa, aun que reconociendo que la hipótesis de conjuntos anidados parece hasta ahora, mejor respaldada que su rival.

In the early ‘70s, Tversky and Kahneman founded a research program in Cognitive Psychology called Heuristics and Biases . This program found extensive evidence that shows that people tend to commit reasoning errors when making judgments under uncertainty. A particular case is that people tend to fail when reasoning about conditional probability pro- blems, that is, problems that ask for the probabi- lity of some event given the fact that another event has occurred (e.g. the probability of rain ing given that it is cloudy). But in the mid ’90s, Gigerenzer and other evolutionary psychologists came along and gave an important turn to the state of the art. They showed that if the condition al probability problems used in the literature are framed in a different way, people’s performance greatly improves. More specifi- cally, if the problems present the information under a specific format called natural frequency format, around 50% of participants get the cor - rect answer. Since the mid ́90s researchers en- gage in an important debate on how to account for such a facilitation effect . There are two main proposals, one by the Evolutionary Psychology Program and the other by Heuristic and Biases Program. The natural frequency hypothesis supported by the Evolutionary Program basically says that the natural frequency format is the responsible factor for the improvement in people’s performance. The Heuristic and Biases Pro gram, in turn, has proposed the nested-set hypothesis to explain the facilitation effect. The basic idea is that natural frequency versions tend to make transparent the relevant subset relations of the problem. When people see clearly the set relations involved in this kind of problems (the argument goes) they tend to use correctly base rates and thus, their performance improves. They point out that, according to this view, the success of the frequency effect does not have to do with natural frequency formats per se. They predict that any format whatsoever that make the relevant set relations clear will show the same effect. The key question is, then, as follows. Is this a case of crucial experimentation in favor of one of our rival hypotheses? In other words, is there an experiment or a series of experiments such that our rival hypotheses predict opposite results, so that we can claim one of them as victorious over the other? The empirical evidence on the matter is mixed. Some studies seem to support the natural frequency hypothesis while others seem to support the nested-set hypothesis. We will then try to clarify this debate by focusing on the diverse strategies and techniques used in the literature to settle the dispute. We will argue that the right strategy to discriminate between both hypotheses is to use genuine probability problems with a clarified set structure and see wheth er these conditions elicit or not a performance comparable to the natural frequency effect. Within this general strategy, we review the literature and found that there are three tech- niques, namely, the improved wording tech- nique, the natural chance technique and the gra- phical representation technique that seem to pro- voke a performance as good as the one elicited by natural formats, giving, thus, a stronger sup- port for the nested set hypothesis. However, a careful analysis shows that neither the improved wording technique nor the chance technique has provided both consistent and clear results in favor of the nested-set hypothesis. As for the gra- phical representation technique, the evidence still seems very slim. The improvement in performance was shown in two studies that worked with only one problem each. Furthermore, neither of these problems seems completely ade- quate. Thus, we do not think the last word about the matter has been said and more empirical work is needed to settle the issue