A number of research groups have exploited utility curves to model interaction with distributed systems. For example, they have been used to construct the models of subjective value that support “intelligent”advice giving systems. They have been integrated into ATM architectures to ensure that users’ Quality of Service requirements are met by underlying network protocols. They have also been used to represent and reason about the risk aversion and risk preference that users exhibit when retrieving resources from remote servers over unreliable networks. However, much of this previous work has rested upon implicit assumptions about properties of the preference relation that underpins modern consumer theory. This paper examines the mathematical basis of the preference relation. The analysis helps to identify the implications that preference axioms have for the application of consumer theory to interface development.