T. W. Sandholm and F. Ygge. On the gains and losses of speculation in equilibrium markets. In Proceedings of the Fifteenth International Joint Conference on Artificial Intelligence, pages 632--638, Nagoya, Japan, Aug. 1997.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....in a structured engineering way in the different components. 1.1.1 The coordination mechanism component The coordination mechanism (or high level interactions) is the component that describe under which conditions interaction between agents is possible. This may be mechanisms for auctions [82, 97], or negotiations [53, 54] It may also be mechanisms used to coordinate cooperative tasks, as in e.g. contract nets [86] and multi agent planning and organizational structures [92] In the sense of game theory, we consider the structure of a game, to be the coordination mechanism from the ....

....chapter 5 are given. An Iterated Hawk and Dove Game Conflict resolutions are typically resolved by an appropriate negotiation process. In a multi agent setting, there have been several proposals of such negotiation processes. A recent and promising approach are models of computational markets [82, 93]. In this case, and other proposals, the negotiation is modeled as an interaction protocol between the agents involved. An important issue in the agent theories is whether we have centralized control or not. In the latter situations we refer to the agents as autonomous. We propose in this paper ....

T.W. Sandholm and F. Ygge. On the gains and losses of speculation in equilibrium markets. In Proceedings of IJCAI 1997.

....game theoretic issues but only algorithmic ones. In his seminal paper, Vickrey [25] argued for the importance of incentive compatibility and rst analyzed the incentive compatible second price auction. The main motivation is to free the bidders from strategic considerations. It has been argued [21, 6, 22, 23, 12] that this is especially important in computerized settings. Our rst result in this paper is a full characterization of incentive compatible on line auctions: An on line auction is called based on supply curves if before receiving the i th bid, b i (q) it xes some function (supply curve) p i ....

T. Sandholm and F. Ygge. On the gains and losses of speculation in equilibrium markets. In Proc. of the 15th Int. Joint Conference on Articial Intelligence (IJCAI-97), 1999.

....submitted as resource supply and demand functions to an auctioneer, who computes prices by solving the resulting system of equations, such that supply equals demand. If the market only clears once, there is little room for speculative behavior among the agents, as has been demonstrated recently [34]. However, if the market clears several times or if the bidding length is randomized, there is again room for time based speculation, and it can be beneficent to bid aggressively initially or near market clearing time. Clients might not want their agents to reveal their private valuations to an ....

Tuomas Sandholm and Fredrik Ygge, On the Gains and Losses of Speculation in Equilibrium Markets, Proc. of the Fifteenth International Conference on Artificial Intelligence, IJCAI '97, Morgan Kaufmann, 1997.

....is relatively small and there is little to gain by strategic behavior. Although it can be advantageous for an agent to bid strategically, this is not always the case. Indeed, it is well established that pursuing aggressively strategic behavior can lead to pitfalls in uncertain environments [Sandholm and Ygge, 1997; Wellman and Hu, 1998] Alternatively, some have used tournaments as a framework for developing and evaluating candidate agent policies in complicated domains. The Santa Fe Double Auction Tournament [Rust et al. 1994] provided some unexpected insights into effective policies in continuous ....

....I calculated a fixed value so that, excluding all other consumers, there exists a positivesurplus solution for this consumer with 0.9 probability. I determined consumer values 1 It is well established that pursuing aggressively strategic behavior can lead to pitfalls in uncertain environments [Sandholm and Ygge, 1997; Wellman and Hu, 1998] Proposers of particular bidding behaviors (strategic or not) in particular environments still face the burden of demonstrating reasonableness. 56 a1 a2 a3 a8 a9 1 3 8 2 a4 4 a10 9 a5 a6 5 6 a11 a13 15 a14 a12 10 a7 7 c1 a15 3.73 Figure 6.3: ....

Tuomas W. Sandholm and Fredrik Ygge. On the gains and losses of speculation in equilibrium markets. In Fifteenth International Joint Conference on Artificial Intelligence, pages 632--638, 1997.

....interaction. We also added a mechanism that enables agents to exchange their goods through time, to support the dynamically changing demands of users. We obtained a desirable result by simulations. In previous research based on general equilibrium, theoretical aspects have been mainly discussed[1, 3, 4, 5, 7, 9]. On the other hand, the analysis on the costs of computation and communication, which comes into issue when these mechanisms are implemented in an actual distributed environment, has not been investigated extensively. In this paper, we examine the following two major tradeoffs by applying our ....

T. Sandholm and F. Ygge, "On the Gains and Losses of Speculation in Equilibrium Markets," 15th International Joint Conference on Artificial Intelligence (IJCAI-97), pp. 632--638, 1997.

....[Kalakota and Whinston, 1996] Both goods (e.g. everything from computers to airplane tickets) and information (e.g. real time stock quotes) can already be bought electronically over the Internet. Another application of automated contracting systems is automated markets for electric power [Sandholm and Ygge, 1997, Ygge and Akkermans, 1996] which are a reality in both the United States and in Europe. Automated decision making is another area which provides people with significant support in complex situations; e.g. in medicine, finance, and engineering. In those complex problems, the amount of information ....

....to monitor the contract. 1.1. 2 Contracts that Involve Multiple Agents or Tasks The contract used most commonly in systems which include automated contracting is a contract that can transfer only one task between two agents [Smith, 1980] This contract will be referred to as an original contract [Sandholm, 1997a, Sandholm, 1997b, Sandholm, 1998] Four new types of contracts have recently been introduced [Sandholm, 1996, Sandholm, 1997a, Sandholm, 1997b, Sandholm, 1998] cluster , swap, multiagent, and OCSM contracts. An OCSMcontract is a combination of the other three contracts and the original ....

[Article contains additional citation context not shown here]

Tuomas W Sandholm and Fredrik Ygge. On the gains and losses of speculation in equilibrium markets. In Proceedings of the Fifteenth International Joint Conference on Artificial Intelligence, pages 632--638, Nagoya, Japan, August 1997.

....is to maximize expected revenues. Moreover, our algorithms for matching buyers and sellers are fundamentally based on flow algorithms as encountered in combinatorial optimization and network theory [22] Finally, MARI builds on work done in the area of market oriented allocation mechanisms [23, 24, 26]. We build upon game theory to formulate our problem in economics terminology [25, 15] with optimization heuristics, such as maximization of aggregate surplus, that derive directly from economics theory. Future Work As of now, we have precisely defined MARI s design framework and functional ....

Sandholm, T. and Fredrick Ygge. "On the Gains and Losses of Speculation in Equilibrium Markets." In Proceedings of the Sixteenth International Joint Conference on Artificial Intelligence, p.632-638, Nagoya, Japan. 1997.

....game theoretic issues but only algorithmic ones. In his seminal paper, Vickrey [25] rst argued for the importance of incentive compatibility and suggested his incentive compatible second price auction. The main motivation is to free the bidders from strategic considerations. It has been argued [21, 6, 22, 23, 12] that this is especially important in computerized settings. Our rst result in this paper is a full characterization of incentive compatible on line auctions: An on line auction is called based on supply curves if before receiving the i th bid, b i (q) it xes some function (supply curve) p i ....

T. Sandholm and F. Ygge. On the gains and losses of speculation in equilibrium markets. In Proc. of the 15th Int. Joint Conference on Articial Intelligence (IJCAI-97), 1999.

....equilibrium) they were still likely to be worse off than had they behaved as simple competitors. The reason is that the strategic agents may had no way to verify their conjectures about actions not taken, and so they often missed favorable opportunities. In a related model, Sandholm and Ygge [21] also found that competitive agents could outperform their strategic counterparts under uncertain conditions, as the more sophisticated policies were more sensitive to errors in models formed. Clearly, these types of experiments merely scratch the surface. We cannot realistically expect to obtain ....

Tuomas Sandholm and Fredrik Ygge. On the gains and losses of speculation in equilibrium markets. In Proceedings of the Sixteenth International Joint Conference on Artificial Intelligence, pages 632--638, Nagoya, Japan, 1997.

....particular model of agents exchanging information goods [37] and find that whether strategic learning is beneficial (or how much) is highly context dependent. We provide further data distinguishing the cases in our recent experiments within a dynamic trading model [18] Finally, Sandholm and Ygge [28] investigate a general equilibrium scenario very similar to ours. Like us, they find that strategic behavior can be counterproductive when agents have incorrect models. Moreover, their study quantifies the costs of acting strategically and competitively as a function of model error, confirming ....

Tuomas Sandholm and Fredrik Ygge. On the gains and losses of speculation in equilibrium markets. In Proceedings of the Sixteenth International Joint Conference on Artificial Intelligence, pages 632--638, Nagoya, Japan, 1997.

....any information about other agents in the system, the agent will perform worse than those who do use the information. Although the particulars differ, our results are consistent with recent studies of strategic versus competitive behavior in computational markets, conducted by ourselves and others [8, 11]. While behaving strategically can always potentially produce superior outcomes, it can also be far riskier than adopting a less aggressive, competitive approach in assumptions about other agents. 6 Conclusions In this paper we analyze the issue of learning about other agents in a dynamic ....

Tuomas Sandholm and Fredrik Ygge. On the gains and losses of speculation in equilibrium markets. In Proceedings of the Sixteenth International Joint Conference on Artificial Intelligence, pages 632-- 638, Nagoya, Japan, 1997.

No context found.

T. W. Sandholm and F. Ygge. On the gains and losses of speculation in equilibrium markets. In Proceedings of the Fifteenth International Joint Conference on Artificial Intelligence, pages 632--638, Nagoya, Japan, Aug. 1997.

....level. Here we will demonstrate how general equilibrium can be reached both through iterating over prices and iterating over resources. To make market oriented programming successful, it is of vital importance to incorporate available knowledge from economic theory [Wellman, 1996] and game theory [Sandholm and Ygge, 1997]. At the same time it is, as argued before [Ygge and Akkermans, 1996, Andersson and Ygge, 1998] also important to study efficient implementations of computational markets. As many resource allocation problems face real time constraints, one interesting issue is to investigate any time algorithms ....

....[Hurwicz, 1986] Agents can still speculate by giving false utility functions, similar to how agents can speculate when submitting demand functions in the priceoriented case. In small markets, agents that have both detailed and accurate knowledge about other agents can gain by such speculation [Sandholm and Ygge, 1997]. 4 where p i # , 5p i # ,and3p i # are abbreviations for p # #z i # #, 5p # #z i # #,and#5p # #z i # ## ,1 ,respectively. The term p i n 5p i n # P n #=1 3p i # # ,1 3p i n P n #=1 3p i # # p i # , p i n # can be interpreted as the expected price which we denote by ....

T. Sandholm and F. Ygge. On the gains and losses of speculation in equilibrium markets. In Proceeding of the Fiftenth International Joint Conference on Artificial Intelligence, IJCAI 97, August 23--29 1997. In press. (Available from http://www.enersearch.se/ygge).

....its utility through trade with other agents. In such trade the HOMEBOTS are programmed to act competitively (equivalent to acting as a pricetaker) Mas Colell et al. 1995, p. 20, p. 314] i.e. treat prices as exogenous rather than speculate about the effects of its own action on market prices [Sandholm and Ygge, 1997]. Also uncontrollable loads are represented by HOMEBOTS. In this case, however, the HOMEBOT is totally insensible to prices. Either a HOMEBOT can represent one or a number of loads. A HOMEBOT representing an uncontrolled load utilize a suitable method for prediction, see e.g. Holst, 1977, Holst ....

....and the loads are managed optimally at the given prices. However, there is a small potential gain for the utility in speculating about its own effect on prices. In most cases though, the potential gain is very small compared to the potential losses caused by uncertainty about the market, cf. [Sandholm and Ygge, 1997]) Therefore, acting competitive is a very reasonable thing to do, and it is often the optimal behavior. 6.4 Cost Benefit Analysis Since every agent is facing the market price, it is able to compute an approximate value of its profits from the ability to control its load. If consequences on ....

T. Sandholm and F. Ygge. On the gains and losses of speculation in equilibrium markets. In Proceeding of the Fiftenth International Joint Conference on Artificial Intelligence, IJCAI 97, August 23--29, 1997. (Available from http://www.enersearch.se/ygge).

....we randomly chose producer costs uniformly over [0; 1] and for each producer, based its bid on 500 Monte Carlo samples over the costs of other producers in its instance. We also 5 It is well established that pursuing aggressively strategic behavior can lead to pitfalls in uncertain environments [12, 16]. Proposers of particular bidding behaviors (strategic or not) in particular environments still face the burden of demonstrating reasonableness. a1 a2 a3 a9 a10 1 3 9 2 a4 a5 a6 a7 a8 4 5 6 7 8 a11 a12 a13 a14 a15 a16 1 0 1 1 1 2 a17 a18 1 3 c1 a19 a20 1 4 a21 ....

T. W. Sandholm and F. Ygge. On the gains and losses of speculation in equilibrium markets. In Fifteenth International Joint Conference on Artificial Intelligence, pages 632--638, Nagoya, Japan, 1997.

....that acts competitively treats prices as exogenous, that is, the impact on the prices due to its own behavior is negligible (Varian, 1996, p. 516) This is a very reasonable assumption if the market is of at least moderate size and or if there is uncertainty about the behavior of the other agents (Sandholm Ygge, 1997). 305 Ygge Akkermans Real World Micro economic theory Economist Computer scientist Human agents on human markets Computational agents on computational markets Study Generate Study Generate Figure 2: A simplified view on the relation between economics and computer science with ....

....should have the opportunity to make trade o#s between comfort and economical value. It has been shown that generally speaking it is very di#cult to benefit from speculation if the number of agents on the market is su#ciently large, and or if uncertainty about the other agents behavior exists (Sandholm Ygge, 1997). Taking several time periods into account. As each o#ce serves as a storage for heat, agents can for example gain by using relatively much resource during certain hours when the total resource need is small. That is, a more realistic approach to the building control problem is to use ....

Sandholm, T. W., & Ygge, F. (1997). On the gains and losses of speculation in equilibrium markets. In Proceeding of the Fifteenth International Joint Conference on Artificial Intelligence, IJCAI 97, pp. 632--638.

....infinity: with infinitely many agents (of comparable size) each agent is best off acting competitively since it will not affect the prices. However, in markets with a finite number of agents, an agent can act strategically, and potentially achieve higher utility by over under representing [66], 36, pp. 220223 ] 25] In doing so, the agent has to speculate how its misrepresentation affects the market prices, which are simultaneously affected by how other agents respond to the prices which changed due to the first agent s strategic actions. In other words, general equilibria do not in ....

....market prices, which are simultaneously affected by how other agents respond to the prices which changed due to the first agent s strategic actions. In other words, general equilibria do not in general correspond to strictly rational, strategic equilibria of game theory. This section is based on [66]. We analyze how much an agent can gain by speculation. Standard lies are also presented via which an agent can drive the market to a solution that maximizes the agent s gains from speculation, and looks like a general equilibrium to the other agents and the adjustor. These results are independent ....

[Article contains additional citation context not shown here]

Tuomas W Sandholm and Fredrik Ygge. On the gains and losses of speculation in equilibrium markets. In Proceedings of the Fifteenth International Joint Conference on Artificial Intelligence, pages 632--638, Nagoya, Japan, August 1997.

....[Hurwicz, 1986] Agents can still speculate by giving false utility functions, similar to how agents can speculate when submitting demand functions in the price oriented case. In small markets, agents that have both detailed and accurate knowledge about other agents can gain by such speculation [Sandholm and Ygge, 1997]. where p i ff and 3p i ff , are abbreviations for p ff (z i ff ) and Gamma 5p ff (z i ff ) Delta Gamma1 , respectively. The term hpi i , de Thetaned by hpi i = p i n 5p i n i P n fi=1 3p i fi j Gamma1 Theta 3p i n P n fi=1 3p i fi i p i fi Gamma p i n j (9) ....

T. W. Sandholm and F. Ygge. On the gains and losses of speculation in equilibrium markets. In Proceeding of the Fiftenth International Joint Conference on Arti\Thetacial Intelligence, IJCAI 97, August 23#29, 1997. (Available from http://www.enersearch.se/ ygge).

....full commitment protocols of the same type, but the differences between the different leveled commitment protocols are minor. 1. Introduction New technology has increased the importance of automated negotiation in modern society. Two examples applications are automated markets for electric power [15, 18] (which are a reality in both the United States and in Europe) and electronic commerce [8] which has become more viable with the expansion of the Internet and novel methods for electronic payments) Contracts in automated negotiation systems consisting of self interested agents have ....

T. W. Sandholm and F. Ygge. On the gains and losses of speculation in equilibrium markets. In Proceedings of the Fifteenth International Joint Conference on Artificial Intelligence, pages 632--638, Nagoya, Japan, Aug. 1997.

....a local optimum with a higher social welfare than the others. Introduction The importance of automated negotiation systems is increasing as a consequence of the development of technology as well as increased application pull, e. g, vehicle routing systems (Sandholm 1993) and electronic commerce (Sandholm Ygge 1997). A central part of such systems is the ability for the agents to reallocate their tasks. Tasks can interact positively or negatively with each other, so they are preferably handled by the same agent or by different agents, respectively. The agents also have different resources that lead to ....

....at a time (Sen 1993) Smith 1980) We will refer to this type of contract as original (O contracts) By allowing more than one task to be transferred in the same contract, more efficient contracting can be achieved. Hence, four new types of contracts have been recently introduced (Sandholm 1996)(Sandholm 1997a) Sandholm 1997b) Sandholm 1998) cluster (C) swap (S) and multiagent contracts, M) as well as a combination of all three of the above with the original contract (OCSM contracts) The C contracts 3 transfer at least two tasks from one agent to another, while the S contracts let two agents ....

[Article contains additional citation context not shown here]

Sandholm, T. W., and Ygge, F. 1997. On the gains and losses of speculation in equilibrium markets. In Proceedings of the Fifteenth International Joint Conference on Artificial Intelligence, 632--638.

....and M contracts concentrated them on a few agents. Introduction The importance of automated negotiation systems is increasing as a consequence of the development of technology as well as increased application pull, e.g. electronic commerce (Kalakota Whinston 1996) electricity markets (Sandholm Ygge 1997), and transportation exchanges (Sandholm 1993) A central part of such systems is the ability to (re)allocate tasks (or analogously, other types of items, e.g. securities, bandwidth slices, Mega Watt hours of electricity, or collectibles) among the agents. Generally, the tasks have a dependency ....

....there is a risk of being trapped in a local optimum instead of reaching the globally optimal task allocation. Cluster contracts (C contracts) Cluster contracts allow the agents to exchange more than one task in each contract, together with a side payment (Sandholm 1993; 1998) Definition. 3 (Sandholm 1997; 1998) A cluster contract (C contract) is a pair hT i;j ; ae i;j i, where jT i;j j 1. T i;j is the task set that agent i gives to agent j, and ae i;j is the contract price that i pays to j for handling the task set. Swap contracts (S contracts) Even when both O contracts and C contracts are ....

[Article contains additional citation context not shown here]

Sandholm, T. W., and Ygge, F. 1997. On the gains and losses of speculation in equilibrium markets. In IJCAI, 632--638.

.... Whinston 1996) Another 1 Supported by NSF CAREER award IRI 9703122 and NSF grant IRI 9610122. Copyright c fl1998, American Association for Artificial Intelligence (www.aaai.org) All rights reserved. application of automated contracting systems is automated markets for electric power (Sandholm Ygge 1997; Ygge Akkermans 1996) which are a reality in both the United States and in Europe. Contracts in automated negotiation systems consisting of self interested agents have traditionally been binding, i.e. impossible to breach. Such contracts do not allow the agents to act efficiently upon future ....

Sandholm, T. W., and Ygge, F. 1997. On the gains and losses of speculation in equilibrium markets. In Proceedings of the Fifteenth International Joint Conference on Artificial Intelligence, 632-- 638.

.... Performance profiles, Negotiation, Contracting, Task allocation 1 Introduction The importance of automated negotiation systems is increasing as a consequence of the development of technology as well as increased application pull, e.g. vehicle routing systems [13] and electronic commerce [8, 18]. A central part of such systems is the ability for the agents to reallocate their tasks. Generally, the tasks have a dependency upon each other, as well as upon the agents. That is, some of the tasks are synergistic and preferably handled by the same agent, 1 Supported by NSF CAREER award ....

T. W. Sandholm and F. Ygge. On the gains and losses of speculation in equilibrium markets. In Proceedings of the Fifteenth International Joint Conference on Artificial Intelligence, pages 632--638, Nagoya, Japan, Aug. 1997.

....level. Here we will demonstrate how general equilibrium can be reached both through iterating over prices and iterating over resources. To make market oriented programming successful, it is of vital importance to incorporate available knowledge from economic theory [Wellman, 1996] and game theory [Sandholm and Ygge, 1997]. At the same time it is, as argued before [Ygge and Akkermans, 1996, Andersson and Ygge, 1998] also important to study effi cient implementations of computational markets. As many resource allocation problems face real time constraints, one interesting issue is to investigate any time algorithms ....

....[Hurwicz, 1986] Agents can still speculate by giving false utility functions, similar to how agents can speculate when submitting demand functions in the priceoriented case. In small markets, agents that have both detailed and accurate knowledge about other agents can gain by such speculation [Sandholm and Ygge, 1997]. where p i ff , 5p i ff , and 3p i ff are abbreviations for p ff (z i ff ) 5p ff (z i ff ) and (5p ff (z i ff ) Gamma1 , respectively. The term p i n 5p i n i P n fi=1 3p i fi j Gamma1 3p i n P n fi=1 3p i fi i p i fi Gamma p i n j can be interpreted as the ....

T. Sandholm and F. Ygge. On the gains and losses of speculation in equilibrium markets. In Proceeding of the Fiftenth International Joint Conference on Artifi cial Intelligence, IJCAI 97, August 23-- 29 1997. In press. (Available from http://www.enersearch.se/ygge).

....[28] This is not the case if agents speculate how their demands and supplies affect the market prices. Recently, Sandholm and Ygge have devised general insincere strategies that allow an agent to drive the market to an equilibrium where the agent s maximal gain from speculation materializes [48]. Speculative behavior in general equilibrium markets has recently been studied in the context of learning by Hu and Wellman [19] Shehory and Kraus [53] analyze coalition formation among rational agents with perfect information in CFGs that are not necessarily superadditive. Their protocol ....

T. W. Sandholm and F. Ygge. On the gains and losses of speculation in equilibrium markets. In Proceedings of the Fifteenth International Joint Conference on Artificial Intelligence, Nagoya, Japan, Aug. 1997.

First 50 documents

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC