A practical theory of fungibility

We formalize 'degrees of fungibility' by differentiating goods according to both their underlying attributes and the perceived value and/or usefulness of those attributes to a value assessor. This allows us to distinguish between goods that appear to be 'exactly the same' from those goods that appear to be 'nearly the same'. Such a distinction is of particular importance in the design space of digital goods, which may exist both natively in the digital space and as surrogates, i.e. as digital representations of physical goods. We provide motivating examples where digital objects are too fungible for certain desired uses, and proceed to develop a formal framework under which degrees of fungibility can be defined and characterized. We close by bridging this framework to applications in machine learning and market design.