This paper analyzes coalitions among self-interested agents that need to solve combinatorial optimization problems to operate e ciently in the world. By colluding (coordinating their actions by solving a joint optimization prob-lem) the agents can sometimes save costs compared to operating individually. A model of bounded rationality is adopted where computation resources are costly. It is not worthwhile solving the problems optimally: solution quality is decision-theoretically traded o against computation cost. A normative, application- and protocol-independent theory of coalitions among bounded-rational agents is devised. The optimal coalition structure and its stability are signi cantly a ected by the agents ' algorithms ' performance pro les and the cost of computation. This relationship is rst analyzed theoretically. Then a domain classi cation including rational and bounded-rational agents is in-troduced. Experimental results are presented in vehicle routing with real data from ve dispatch centers. This problem is NP-complete and the instances are so large that|with current technology|any agent's rationality is bounded by computational complexity. 1

This paper presents eMediator, an electronic commerce server prototype that demonstrates ways in which algorithmic support and game-theoretic incentive engineering can jointly improve the efficiency of ecommerce. eAuctionHouse, the configurable auction server, includes a variety of generalized combinatorial auctions and exchanges, pricing schemes, bidding languages, mobile agents, and user support for choosing an auction type. We introduce two new logical bidding languages for combinatorial markets: the XOR bidding language and the OR-of-XORs bidding language. Unlike the traditional OR bidding language, these are fully expressive. They therefore enable the use of the Clarke-Groves pricing mechanism for motivating the bidders to bid truthfully. eAuctionHouse also supports supply/demand curve bidding. eCommitter, the leveled commitment contract optimizer, determines the optimal contract price and decommitting penalties for a variety of leveled commitment contracting mechanisms, taking into account that rational agents will decommit strategically in Nash equilibrium. It also determines the optimal decommitting strategies for any given leveled commitment contract. eExchangeHouse, the safe exchange planner, enables unenforced anonymous exchanges by dividing the exchange into chunks and sequencing those chunks to be delivered safely in alternation between the buyer and the seller.

A Dissertation Presented by TUOMAS W. SANDHOLM

The Vickrey auction has been widely advocated for multiagent systems. First we review its limitations so as to guide practitioners in their decision of when to use that protocol. These limitations include lower revenue than alternative protocols, lying in non-private-value auctions, bidder collusion, a lying auctioneer, and undesirable revelation of sensitive information. We discuss the special characteristics of Internet auctions: third party auction servers, cryptography, and how proxy agents relate to the revelation principle and fail to promote truth-telling.

Automated negotiation and coalition formation among self-interested agents are playing an increasingly important role in electronic commerce. Such agents cannot be coordinated by externally imposing their strategies. Instead the interaction protocols have to be designed so that each agent is motivated to follow the strategies that the protocol designer wants it to follow. This paper reviews six component technologies that we have developed for making such interactions less manipulable and more e cient in terms of the computational processes and the outcomes: 1. OCSM-contracts in marginal cost based contracting, 2. leveled commitment contracts, 3. anytime coalition structure generation with worst case guarantees, 4. trading o computation cost against optimization quality within each coalition, 5. distributing search among insincere agents, and 6. unenforced contract execution. Each of these technologies represents a di erent way of battling selfinterest and combinatorial complexity simultaneously. This is a key battle when multiagent systems move into large-scale open settings.

Combinatorial auctions where bidders can bid on bundles of items can lead to more economically efficient allocations, but determining the winners is NP-complete and inapproximable. We present CABOB, a sophisticated optimal search algorithm for the problem. It uses decomposition techniques, upper and lower bounding (also across components), elaborate and dynamically chosen bid-ordering heuristics, and a host of structural observations. CABOB attempts to capture structure in any instance without making assumptions about the instance distribution. Experiments against the fastest prior algorithm, CPLEX 8.0, show that CABOB is often faster, seldom drastically slower, and in many cases drastically faster—especially in cases with structure. CABOB’s search runs in linear space and has significantly better anytime performance than CPLEX. We also uncover interesting aspects of the problem itself. First, problems with short bids, which were hard for the first generation of specialized algorithms, are easy. Second, almost all of the CATS distributions are easy, and the run time is virtually unaffected by the number of goods. Third, we test several random restart strategies, showing that they do not help on this problem—the run-time distribution does not have a heavy tail.

Collective choice settings are the heart of society. Game theory provides a basis for engineering the incentives into the interaction mechanism (e.g., rules of an election or auction) so that a desirable system-wide outcome (e.g., president, resource allocation, or task allocation) is chosen even though every agent acts based on self-interest. However, there are a host of computer science issues not traditionally addressed in game theory that have to be addressed in order to make mechanisms work in the real world. Those computing, communication, and privacy issues are deeply intertwined with the economic incentive issues. For example, the fact that agents have limited computational capabilities to determine their own (and others') preferences ruins the incentive properties of established auction mechanisms, and gives rise to new issues. On the positive side, computational complexity can be used as a barrier to strategic behavior in settings where economic mechanism design falls short. Novel computational approaches also enable new economic institutions. For example, market clearing technology with specialized search algorithms is enabling a form of interaction that I call expressive competition. As another example, selective incremental preference elicitation can determine the optimal outcome while requiring the agents to determine and reveal only a small portion of their preferences. Furthermore, automated mechanism design can yield better mechanisms than the best known to date.

Procter & Gamble put into practice CombineNet’s approach to building sourcing networks, called expressive competition. At its heart is a vision that looks past lowest-price reverse auctions and combinatorial package bidding toward a highly expressive commerce relationship with suppliers. It enables suppliers to make electronic offers that express rich forms of capabilities and efficiencies. As the buyer, P&G also uses an expressive language to state constraints and preferences. The detailed expressions of supply and demand are brought together via an advanced optimization engine to decide the optimal allocation of business to the suppliers. By March 2005, over a period of two and a half years, P&G had sourced over $3 billion through expressive commerce and seen $294.8 million (9.6 percent) in recommended savings. In the process, P&G’s suppliers benefited from the winwin approach: expressive competition matched demand to the most efficient means of production (rather than squeezing suppliers’ profit margins) and removed the exposure risks in making offers. Beyond direct monetary savings, the benefits included the redesign of supply networks with quantitative understanding of the trade-offs and the ability to implement in weeks instead of months.

Drawing from our experiences of designing and fielding over 800 sourcing auctions totaling over $60 billion, I will discuss issues that arise in very-large-scale generalized combinatorial auctions, as well as solutions that work (and ones that do not). These are by far the largest (in terms of the number of items as well as the number of side constraints) and most complex combinatorial