Performance and learning in a simulation of oil industry dynamics

Experimental studies on decision making in dynamically complex simulated systems have shown that poor performance arises from decision makers' failure to correctly interpret feedback information due to the confounding effects of stock accumulation, time delays between decisions and actions, and non-linearities between system elements. This experimental simulation study examines ways to improve performance through online information feedback that clarifies for subjects the task structure of their decisions including cause and effect. In order to test our ideas we devised an assignment for business school students using a microworld simulator representing the global oil industry. Subjects completed a set of six upstream investment tasks and their resulting financial performance was measured and recorded. The repeated measures factorial design included three different treatment groups, who received various task structure feedback during the first three trials only. A fourth control group received no task structure feedback at all. The results show that mean subject performance was significantly greater for the treatment groups during the first two trials, consistent with our view that task structure feedback is helpful. However, in later trials the performance of all groups, including the control group, eventually reached a similar plateau, some 30% below a behavioural benchmark. From these results conclusions are drawn about individual and team learning in complex dynamic systems.