Bias in Planning Algorithms


Does bias exist in planning algorithms? If so, how does bias manifest, and how important is this bias? Answering this question requires a formal, mathematical definition of bias. We formally define bias as the distance between the probability distributions of solutions returned by various algorithms, and the uniform distribution over solutions. We show in this paper that deterministic algorithms are inherently biased, as they don’t return all solutions, and that this property holds even when algorithms return a set of plans instead of just one plan. Exceptions are problem instances or problem classes for which only a single solution exists. We show that entropy is a proxy for the more complex and more expensive distance measurement between pairs of probability distributions. We then discuss changing the definition of bias to compare the probability distributions of properties of sets of plans instead of individual plans. We show the property set bias is smaller than the bias of actual plans. We then describe a roadmap for future investigations of bias in planning.

Workshop on Reliable Data-Driven Planning and Scheduling