Following the operationalization, the effects of the proposed determinants on learning are summarized in the hypotheses below in terms of quality of learning, where “better learning” is associated with “better outcomes”. Also, it is possible to use the commonly used measure of convergence to some equilibrium where “better learning” is, eg., associated with faster convergence.

Click to enlarge image

Discussion

The contingent learning approach (CLA) aims to identify the factors that influence learning in order to explain and predict adaptive behavior. Every situation has its specifics, and there are always several factors that influence learning simultaneously, though not necessarily in the same direction. That is, the effects summarized in the above hypotheses may support or offset each other so that the net effect of the relevant determinants in a given situation is not always straightforward, but must be analyzed carefully.

Two implications follow: First, it may sometimes not be difficult to find or make up examples where single hypotheses of the CLA are expected to be falsified, because other learning determinants, or additional factors (such as the notion of fairness in some games) are dominant for the behavior in that example. Therefore, it is desirable to test the CLA hypotheses under ceteris paribus conditions. That these conditions can be identified, and that the determinants and effects can theoretically be distinguished may be a methodological advantage of the CLA.

However, in some cases it may be difficult to provide a strict ceteris paribus analysis or test since determinants and effects may be interwoven, and outside factors may be important (such as framing effects). Second, it therefore follows that the theorist’s task is to untangle the net effect of the determinants on learning, and analyze them as commonly done in comparative statics.

As in standard economics, a theorist who wants to model a situation where learning is involved may carefully choose and include the determinants for learning that appear to be relevant in that situation. Hence, the CLA aims to enrich economic theory by providing useful tools for model building, but does not make the art of doing so obsolete.

It follows from this short discussion that the CLA hypotheses cannot be assumed to be generic statements that hold in every instance. Rather, they reflect general tendencies that are likely to influence learning in the direction indicated by the hypotheses in many situations. The range of economically relevant situations in which they allow to explain and predict behavior remains to be explored, however.

Prof. Tilman Slembeck

Next: Concluding Remarks

Summary: Index