Foundations of Algorithmic Contract Theory

Description du projet

Étendre la théorie des contrats algorithmiques pour une application dans le monde réel

Les contrats sont des outils économiques par lesquels un «principal» incite des «agents» à fournir un effort rémunéré en son nom. La négociation de contrats en ligne a connu une croissance exponentielle, notamment dans des domaines tels que les services de freelance et les cours en ligne. En outre, l’ampleur et la complexité de ces contrats ont considérablement augmenté, ce qui rend l’utilisation des données possible et essentielle dans les négociations. Il est dès lors nécessaire d’adopter une approche algorithmique de la conception des contrats. La théorie des contrats algorithmiques, proposée dans un article pionnier de 2006, peut répondre à ce besoin. Le projet FACT, financé par le CER et initié par un coauteur de l’article original, vise à faire progresser la théorie des contrats algorithmiques, en améliorant son applicabilité à la négociation moderne de contrats en ligne et en établissant une théorie fondamentale pour la conception de contrats algorithmiques.

Objectif

Algorithmic game theory (AGT) lies at the intersection of algorithms and incentives, offering a framework for the design of algorithms that provide incentives for collaboration in computational markets. AGT has played a pivotal role in shaping vital computing platforms. One of the most influential fields within AGT is algorithmic mechanism design, providing incentives for truthful revelation of private information.
An equally important intersection of algorithms and incentives arises in the design of algorithms incentivizing self-interested agents to exert costly effort in executing their tasks. Incentivizing effort was studied with great success in the economics field of contract theory.
As classic applications of effort incentivization move to computational platforms, including huge computational markets for freelancing services, marketing, and online courses, they grow in scale and complexity and become data-driven. In this evolving landscape, adopting an algorithmic approach to contract design is essential. This includes a language for analyzing the computational complexity of contracts in combinatorial settings, tools for studying simplicity-optimality tradeoffs, and a formalism for analyzing data-driven contracts.
In 2006, I co-authored the pioneering paper applying an algorithmic lens to contract design, marking the inception of the field of algorithmic contract theory (ACT) and foreshadowed future developments. However, it is only in recent years that the scientific landscape and computational platforms have matured to a stage where ACT has become a prominent field in AGT.
Preliminary research in ACT has been enlightening. However, the state of the art of ACT still lags far behind full applicability in Internet applications, largely due to disregard of crucial real-world aspects. The paramount goal of this project is to establish a foundational theory for algorithmic contract design that will bridge this gap and inform the design of real-world applications.