A Heuristic Approach to Explore: The Value of Perfect Information.

This article introduces the myopic-value of perfect information (myopic-VPI), a new heuristic decision model designed to address multiarmed bandit (MAB) problems, which involve balancing exploration and exploitation under uncertainty. Myopic-VPI simplifies decision-making by ranking alternatives and computing a one-dimensional integration to estimate the expected value of exploration, avoiding the computational complexity of dynamic programming (DP) approaches. Through extensive simulation experiments, myopic-VPI is shown to significantly reduce computational time compared to other heuristics such as one-step lookahead (OSL), knowledge gradient (KG), and index strategies, while achieving competitive accumulated utility, especially when the discount factor is below 0.85 or when agents exhibit risk aversion. Empirical estimation using consumer scanner data from the diaper category demonstrates that myopic-VPI fits observed choice behavior on par with more computationally intensive models like index and near-optimal strategies, but with substantially lower estimation time, suggesting its practical value as a fast-and-frugal heuristic for modeling exploration-exploitation tradeoffs.