nts have recently emerged as a valuable abstraction for characterizing interactions among autonomous agents at the level of their business relationships. Traditionally, interoperation is approached from the standpoint of data exchange or of messaging. We use commitments to characterize interoperability in high-level terms: at the level of the communications among agents. Specifically, two agents are interoperable if their commitments align. Drawing upon Kant's famous distinction, we distinguish between two kinds of interoperability, constitutive and regulative. Constitutive interoperability takes into account solely the meaning of messages whereas regulative interoperability also takes into consideration message order, occurrence, and data flow. We present a language for specifying agents constitutively and a decision procedure for determining their interoperability. Categories and Subject Descriptors D.2.12 [Software Engineering]: Interoperability; I.2.11 [Artificial Intelligenc...
Amit K. Chopra, Munindar P. Singh