Research as a stochastic decision process

Abstract

This article presents a novel approach to research projects involving high uncertainty, which the author claims can substantially improve productivity. The approach involves modeling the research process as a stochastic decision process, with the goal of maximizing the probability of eventual success while minimizing the expected time spent. The author argues that traditional strategies for tackling uncertain projects, such as completing tasks in order of difficulty or quickest completion time, are often counterproductive. Instead, the author advocates for prioritizing tasks based on their informativeness per unit time, which can be quantified using methods such as expected time saved or failure rate. The author also emphasizes the importance of de-risking tasks by seeking information about their feasibility early on. This can be done through prototyping, simulations, or running experiments. Overall, the author provides a systematic framework for approaching uncertain research projects, which can help researchers allocate their time and effort more effectively. – AI-generated abstract.