R&D managers' choices of development policies in simulated R&D environments

There have been numerous clinical investigations of the decision making process in development, however no such systematic study has been performed under experimentally controlled conditions, thus making generalizations about such behavior difficult. This paper describes the results of some experiments, based on a research and development (R&D) game that examined R&D managers' proclivities for various optims that hedge against uncertainty and the overall quality of their choices with respect to the Bayesian decision-theoretic model. Thirty R&D managers participated in the experiments, which were conducted at an industrial R&D facility. R&D managers exhibited a higher proclivity for parallel approaches in information acquisition, but a lower propensity for parallel approaches in terminal actions with respect to the Bayesian norm. Subjects also preferred simultaneous (parallel) to sequential information acquisition, albeit the latter was preferable. These results were significant at the 5-percent level or better. Subjects' choices, which generally conformed to the principles and practices found in the R&D literature, led to substantially suboptimal behavior as compared to the Bayesian decision-theoretic norm. This strongly suggests the need to introduce decision theory into R&D management practice through industrial training programs, a recommendation voiced by the R&D managers themselves.