Parametric Adaptive Learning (Draft)

We investigate a general parametric model of adaptive learning. The model includes most of the adaptive learning procedures studied in the literature where agents optimize given their ranking over actions, perhaps allowing for experimentation. It provides a convenient parametric framework to analyze experimental data and to compare the performance of previously proposed learning hypotheses. We show that several “parameter clusters” result in qualitatively similarly behavior, hence making precise the important realtions between the di¤erent parameters. We also identify and analyze some previously uninvestigated parameter clusters which lead to empirically plausible behavior, such as “loss aversion.”