The PIs use a mechanism design approach to analyze the informational properties of experimental trials, with the goal of improving upon the standard randomized controlled trial (RCT) method. Their concept of a selective trial uses a dominant strategy mechanism to generalize RCTs by allowing experimental participants to express their preference over treatment through costly signals that increase the probability of being chose for the treatment group. A selective trial makes the implicit selection possibility in an ordinary RCT an explicit part of the experimental design.
The research has four parts. First the PIs demonstrate how straightforward extensions of standard experimental protocols can produce important information. Second, they demonstrate that these extensions work as predicted in controlled experimental settings. Third, they use the same basic concepts to consider methods for conducting dynamic experiments. Fourth, they consider how to adjust the experimental design for situations where there are informational or productive externalities between experimental subjects.
Randomized controlled trials are becoming increasingly important in the social sciences. Improved experimental design that carefully takes into account the incentives and beliefs of experimental subjects will improve the usefulness of field experiments.
Broader Impacts: These methods are also useful for medical research. The method is particularly suited to considering medical non-compliance, and may provide a more accurate way to predict the effects of medications. The PIs will be making the power formulae, protocols, and software for data analysis available to a wide community.
