Some reflections on maxEnt

Abstract

ABSTRACT Two different applications of Jaynes’ maximum entropy framework, maxEnt, are given. The purpose is to show how maxEnt stimulates questions about the relations between the formal structure of the maxEnt model and the practical, possibly political, considerations which arise when deciding the objective function and the constraints. While these points do not lead to computationally complicated models they do raise important questions about the relationship between model form and the environment in which it is used. The first example is from urban and transportation planning. The use of maxEnt to derive the allocation of trips, the gravity model, brought benefits of clarity but also raised an important question with policy implications: should journey cost rightly be a behavioural constraint in the model or should it be considered an output? The policy implications are discussed. The second case was initially stimulated by personnel selection and raises the possibility that it is not always clear what maxEnt should be optimising, what should we be minimally discriminating about? The issue is quite general and is illustrated by the choice of a scheme for water resource allocation.