Experimenting with Democracy
TECHNOLOGICAL ACCELERATION increases the capacity to measure the consequences of government action. The resulting social knowledge of which policies work can improve government performance and decrease social disagreement. Government itself can be a catalyst for increased social knowledge if government transforms itself into a better instrument for social learning by adopting rules that will better examine which government programs are successful. These include rules encouraging policy experimentation through decentralization and randomization, providing incentives for improved research practices, and making government data more transparent and accessible. The information age permits us to foster a more experimental politics.
Empirical social science attempts to discover the causes of social behavior. One cause of social behavior is social policies. Insofar as empirical investigators show how various policies affect social behavior, they create social knowledge that can improve policies on the assumption that there is a consensus on the behavior to be either encouraged or discouraged. Thus, like natural science, empirical social science seeks the causes of things. But the nature of social phenomena makes this task difficult. Experiments can be designed to isolate the causes of natural phenomena, but social science generally faces the difficulty of trying to infer causes from a welter of real-world data.1
For instance, suppose we want to know the effect of merit pay awarded to teachers on the basis of improvement in class test scores on a particular kind of human behavior to be encouraged: student learning. There are a variety of possible theories about merit pay. On one hand, the theory in support is that it will provide incentives for teachers to teach better, improving long-term educational outcomes for their students. On the other