Successful management of dynamic decision making tasks is widely thought to depend on the adequacy of the decision makers' "mental models" of the dynamic environment. A mental model is usually taken to be some representation of the decision maker's knowledge and understanding of the dynamic task. This project will explore mental models -- their measurement, their correctness, and their relation to performance. The central question to be addressed is: To what degree does the ability of humans to perform in dynamic environments depend on their understanding of the system to be managed? Understanding of the task is indicated by the correspondence between the mental model and the system model. This question has not been addressed in previous research on dynamic decision making because a) the dynamic systems used were simple and were described in a way that would maximize understanding of the systems, and b) system understanding, i.e., correspondence between mental models and system models, was not assessed. Therefore, it has not been possible to determine whether suboptimal performance was due to lack of understanding of the system (an inadequate mental model), lack of an ability to manage the system even though it is understood (inability to use the model as an effective guide to action), or both. This project is the result of collaboration between judgment researchers and system dynamicists. A dynamic policy simulation model of the JOBS program (a social services program intended to get people off of welfare and into the workforce) has been developed and validated. A user interface for the model has also been developed and it has been found to be a feasible basis for decision making simulations. In this project, the JOBS model will provide a platform for realistic simulation of dynamic decision making in a public policy context. Subjects will include policy experts (county social services commissioners and staff) and novices (graduate students in public administration). Three distinct but interrelated types of mental models ("ends" models, "means" models, and models of system functioning) will be assessed before and after the policy simulation using two different methods. Model coherence and correctness, and the relation between performance and quality of each of the three types of models will be evaluated.
