The value of Value of Information (VoI): When and how to use simpler or heuristic voI approaches

Abstract: Any decision process must address whether to seek further information, which at least implicitly requires VoI issues. However, many important decisions, including research prioritization, biosurveillance, or study design, contain multiple levels of uncertainty which make formal models of VoI computationally intractable, or at least make such analysis unreasonably expensive to produce. Moreover, simpler approaches ranging from simple heuristics, to EVPI, to expert judgement, can provide most or all of the value. This means that while VoI is necessarily positive, the value of VoI analysis is often negative. The seminar presents a staged process which was developed for considering when and how to do VoI. It allows consideration of whether near-optimal decisions can be found without additional research at each stage. This practical approach is intended to be scalable to the full range of VoI questions, and is indeed applicable to decision processes generally.

David's presentation was based on Chapter 5 of his dissertation.