Reconciling Some Conflicting Evidence on Decision Making under Uncertainty

Laboratory experiments concerning decision under uncertainty tend to uncover systematic violations of Bayesian rationality. When models that posit Bayesian rationality are compared to non-experimental data, though, they fit the data well. One possible explanation is that an agent's global pattern of choices may not be rationalizable, but that the pattern may satisfy weak conditions sufficent to rationalize the limited range of choices required by any particular decision protocol. Examples of such patterns are constructed here. An agent who adopts a protocol acts rationally, but an experimenter induces irrationality by imposing distinct protocols in various phases of the experiment.