Negotiating multiple deals is an essential day-to-day activity for many businesses. Procurement, for instance, typically represents one of the largest expense items for businesses worldwide. Today, 95% of European businesses are still negotiating their goods and services entirely unaided by computers, which has been shown to lead to significantly less efficient outcomes, increased costs, and highly labor-intensive processes. Enabling the automation of general-purpose multi-deal negotiations would therefore have an enormous impact on the competitiveness and profitability of businesses world-wide. However, currently available algorithms are not yet capable of performing multiple complex and interdependent negotiations at the same time. This so far underexplored research challenge calls for solutions and methods beyond the state-of-the-art research in auctions, game theory, or bilateral negotiation. It requires new asynchronous negotiation strategies that can reach multiple interdependent deals, as well as novel mathematical coordination mechanisms that are able to steer pro-actively toward a desirable aggregate outcome. The challenges of multi-deal negotiation call for 1) a mathematical model and protocol for multi-deal negotiation algorithms that can strike multiple partial deals with multiple partners; 2) coordination techniques for making optimal trade-offs regarding expected agreement utility; and 3) multi-deal negotiation strategies that can provide online probability estimates of the expected outcome. Altogether, such a research endeavor would deliver the fundamental underpinnings for general-purpose multi-deal negotiation algorithms, thereby paving the way for future systems for domains ranging from procurement and energy to ethics and transportation.

, , , ,
Coordinating Multi-deal Bilateral Negotiations
23rd International Conference on Autonomous Agents and Multiagent Systems, AAMAS 2024
Centrum Wiskunde & Informatica, Amsterdam (CWI), The Netherlands

Baarslag, T. (2024). Multi-deal negotiation. In Proceedings of the International Conference on Autonomous Agents and Multiagent Systems (pp. 2668–2673).