MAINA: Mechanism for Automated Inference of Negotiable Attributes from Contrasts in Declarative Representation

Representation of business rules in contracts has been widely studied in AI literature. Courteous Logic Programs (CLP) developed at IBM explicitly incorporates prioritized conflict handling into its declarative semantics [1][2]. Business Rules Markup Language (BRML), an XML encoding of CLP, features the representation with simple parsing, easy integration into the world-wide-web, and most importantly robust interchangeability with heterogeneous commercial rule systems [6]. However, a problem that has not been fully addressed in CLP/BRML is the inference of negotiable attributes from contracts, which is of particular interest in automated trading where agents need agree upon terms of contracts through negotiation. Challenges agents often face in such situations are the ignorance of non-predefined negotiable attributes and the difficulty of finding them. This mini project then aims at exploiting appropriate representation to allow automated inference of negotiable attributes. To this end, contract format is first defined in CLP syntax subject to two light-weight restrictions. Then, a mechanism MAINA is developed to discover directly, indirectly, and weakly negotiable attributes by applying direct intersection, condition matching, and dynamic rule modification on negotiating contracts. Successful in identifying negotiable attributes, the changes to CLP/BRML incur a small overhead and retain the two most decent properties of CLP/BRML – computational tractability and high-level human understandability. Due to the novelty of this work, no detailed analyses were done to study the complexities of MAINA, yet the implementation is indeed efficiency-aware.