Title: An R package for Data-Driven Dynamic Decision Models
Speaker: John J. Nay, Vanderbilt Univ.
Abstract: I'll describe a method I’ve developed with Jonathan Gilligan (Vanderbilt Univ.) for automatically generating models of dynamic decision-making that both have strong predictive power and are interpretable in human terms. I’ll outline applications of the method to empirical data from decades of game theory experiments and international cooperation over trans-boundary rivers. Just plugging in the experimental data as input, we obtain a highly interpretable decision model that predicts human strategic behavior better than all of the existing theoretical models we are aware of. We also demonstrate the method's ability to recover known data-generating processes by simulating data and correctly deriving the underlying decision models. Finally, I will outline the open-source software package, datafsm, that implements the method, which is also available on CRAN, and briefly discuss features of R (closures and functionals, Rcpp, package system, etc.) that motivated us to implement the tool in R.