## Using Similarity Criteria to Make Issue Trade-Offs in Automated Negotiations (2002)

### Cached

### Download Links

- [www.ecs.soton.ac.uk]
- [www.iiia.csic.es]
- [www.iiia.csic.es]
- DBLP

### Other Repositories/Bibliography

Venue: | Artificial Intelligence |

Citations: | 97 - 8 self |

### BibTeX

@ARTICLE{Faratin02usingsimilarity,

author = {P. Faratin and C. Sierra and N. R. Jennings},

title = {Using Similarity Criteria to Make Issue Trade-Offs in Automated Negotiations},

journal = {Artificial Intelligence},

year = {2002},

volume = {142},

pages = {205--237}

}

### Years of Citing Articles

### OpenURL

### Abstract

Automated negotiation is a key form of interaction in systems that are composed of multiple autonomous agents. The aim of such interactions is to reach agreements through an iterative process of making offers. The content of such proposals are, however, a function of the strategy of the agents. Here we present a strategy called the trade-off strategy where multiple negotiation decision variables are traded-off against one another (e.g., paying a higher price in order to obtain an earlier delivery date or waiting longer in order to obtain a higher quality service). Such a strategy is commonly known to increase the social welfare of agents. Yet, to date, most computational work in this area has ignored the issue of trade-offs, instead aiming to increase social welfare through mechanism design. The aim of this paper is to develop a heuristic computational model of the trade-off strategy and show that it can lead to an increased social welfare of the system. A novel linear algorithm is presented that enables software agents to make trade-offs for multi-dimensional goods for the problem of distributed resource allocation.

### Citations

4143 |
Artificial Intelligence: A Modern Approach, Pearson Education, 2nd edition
- Russell, Norvig
- 1995
(Show Context)
Citation Context ... of the negotiation). Even if assigning prior probabilities was practically achievable for interactions that are repeated (hence permitting the use of probability update mechanisms such as Bayes rule =-=[47]-=-), the same is not true for encounters in open systems. In such environments the prior probabilities may simply be wrong, a fact that is exacerbated by the one-off nature of encounters which prevents ... |

1807 |
The Sciences of the Artificial
- Simon
- 1996
(Show Context)
Citation Context ...iencies only arise when beliefs are probabilistically inferred, leading to a process of negotiation. Generally, the theory is silent with respect to the actual computational rationality of the agents =-=[51]-=-. However in the real world, to know a solution exists is not to know what the solution is. The perfect rationality of all agents, although useful in designing, predicting and proving properties of a ... |

1149 |
Linear and Non-Linear Programming
- Luenberger
- 1984
(Show Context)
Citation Context ...tract that is most likely to be accepted). Attempting to approximate the preference structure of an agent based upon its negotiation be5 Numeric methods, such as active sets, can handle such problems =-=[26]-=-. However, with this method as the number of decision variables increases then so does the complexity of the computation involved in solving the quadratic problem. Therefore, active sets become unlike... |

1130 |
Vision: a computational investigation into the human representation and processing of visual information
- Marr
- 1982
(Show Context)
Citation Context ...their visibility. There are various physiological characteristics of the human visual field, distribution of cones and rods, that ensure some colours are better perceived when moving away than others =-=[28]-=-. Green is the colour with the worst visibility and yellow and cyan are those with the best visibility. Given these three criteria, the colour domain can be modeled in the following way (functions are... |

1094 |
The contract net protocol: high-level communication and control in a distributed problem solver
- Smith
- 1980
(Show Context)
Citation Context ...social welfare function (the product of agent utilities, or the Nash solution) for task, state and worth oriented domains [43]. Similarly, Sandholm, in addition to extending the Contract Net protocol =-=[52]-=- with decision theoretic mechanisms, developed a computational model of leveled commitments and coalition formation based on principles of cooperative game theory. On the other hand, Kraus developed n... |

1068 |
Features of similarity
- Tversky
- 1977
(Show Context)
Citation Context ...soning, and interpolation) it has received comparatively little attention in the context of logical models of reasoning. It has, however, been used in work on psychological studies of human behaviour =-=[55]-=-, mathematical work on graded extensions of equivalence relations [56, 64], and as a model of approximate reasoning [22, 59]. From the philosophical perspective, Niiniluoto relates similarity with the... |

923 |
Non-cooperative games
- Nash
- 1951
(Show Context)
Citation Context ...oretic models (in particular that of Rubinstein [45], which has been shown to implement the Nash bargaining solution under some conditions [46] and thus strengthening the support for the Nash Program =-=[31]-=-) that models the negotiation process as a bi-lateral bargaining game, consisting of an alternating and sequential protocol of offers and counter-offers. Our work also borrows from game theory. In par... |

872 | Perfect Equilibrium in a Bargaining Model,” Econometrica
- Rubinstein
- 1982
(Show Context)
Citation Context ...y managed by assuming a finite horizon of interactions. These interactions also follow the rules of an alternating sequential protocol in which the agents take turns to make offers and counter offers =-=[45]-=-. The protocol terminates when the agents come to an agreement or when one of them withdraws from the negotiation. One implication of the above assumptions is that it is not possible to pre-compute an... |

819 |
The bargaining problem
- Nash
- 1950
(Show Context)
Citation Context ... satisficing approach for their strategy [8, 23]. This is in contrast with deductive models of negotiation where each agent explicitly represents and reasons with the decision tree of the entire game =-=[30, 15]-=-. In this case, a negotiation strategy is then the specification (using for example backward induction [2]) of a sequence of choices for every decision node in the game tree, with the property that bo... |

721 |
Game Theory: Analysis of Conflict
- Myerson
- 1991
(Show Context)
Citation Context ...er to maximise their own utility. A mechanism design solution to this problem consists of centrally designing direct incentive compatible or strategy proof decision rules that have certain properties =-=[29, 43]-=-. Although we acknowledge strategic misrepresentation is a concern in multi-agent systems, we are also interested in the types of decision problems that are not only highly complex in dimensionality (... |

471 |
The art and science of negotiations
- Raiffa
- 1982
(Show Context)
Citation Context ...an offer that, although of the same value to the proposer, may be of greater benefit to the negotiation opponent. This, in turn, should make agreement more likely and increase the overall joint gains =-=[41]-=- between the two agents. The particular heuristic we consider in this paper is based on the degree of similarity between two consecutive choices. The contribution of this work is twofold. Firstly, cur... |

323 | Agents that reason and negotiate by arguing
- Parsons, Sierra, et al.
- 1998
(Show Context)
Citation Context ... is solved through showing the dominance of a truthful bidding strategy [57], to models in which the agents’ argue for positions and aim to persuade their opponents of the value of particular action=-=s [37]-=-. In this work we are interested in conflicting preferences over complex multi-dimensional decision problems involved in the bi-lateral resource allocation negotiation of services [50]. In such duopol... |

294 |
Rules of Encounter
- Rosenschein, Zlotkin
- 1994
(Show Context)
Citation Context ...aining presented here shares with other mechanism design models the explicit design of protocols whose execution is a function of the agent’s strategy [3]. A protocol is a set of “rules of encount=-=er” [43]-=- between the negotiation participants; that is, who can say what, to whom, at what time. Given a protocol, an agent strategy then defines the model that the individual participants apply to act in lin... |

288 | Negotiation Decision Functions for Autonomous Agents, Int
- Faratin, Sierra, et al.
- 1998
(Show Context)
Citation Context ...esources consumed in the negotiation, and the behaviour of the negotiation opponent (for this reason this is called a responsive mechanism since agents react to their prevailing environmental context =-=[12]-=-). This exchange of offers and counter-offers continues until a crossover occurs between the demands of the two agents or one of the agents withdraws. Conversely, reasoning in the trade-off mechanism ... |

282 | Distributed Constraint Satisfaction Problems
- Yokoo, Hirayama
- 2000
(Show Context)
Citation Context ... mechanisms that deliberate over raw values rather than objectives. Similarly, Yokoo and colleagues formalize negotiation as an extension to the classic single agent constraint satisfaction framework =-=[60, 62]-=-, where variables and constraints are distributed among multiple agents. Search algorithms (asynchronous backtracking and asynchronous weak-commitment search) are shown to solve this distributed probl... |

231 | Automated negotiation: prospects, methods and challenges
- Jennings, Faratin, et al.
- 2001
(Show Context)
Citation Context ... parties. The parties first verbalise contradictory demands and then move towards agreements. The prevalence and importance of automated negotiation can be seen in the large number of proposed models =-=[8, 18]: ra-=-nging from auctions in which the agents’ pricing decision problem is solved through showing the dominance of a truthful bidding strategy [57], to models in which the agents’ argue for positions an... |

200 |
Fun and Games. A Text on Game Theory
- Binmore
- 1992
(Show Context)
Citation Context ... upon the agreement. The generative model of bargaining presented here shares with other mechanism design models the explicit design of protocols whose execution is a function of the agent’s strateg=-=y [3]. A -=-protocol is a set of “rules of encounter” [43] between the negotiation participants; that is, who can say what, to whom, at what time. Given a protocol, an agent strategy then defines the model th... |

200 |
Negotiation behavior
- Pruitt
- 1981
(Show Context)
Citation Context ...onflict outcome and is often expected to have an important bearing on the outcome of negotiation. For example, in multi-issue negotiations the middle point on each issue often becomes the focal point =-=[40]-=- and the negotiators then try together to find other agreements that are better for both. In section 4.3 we use the axiomatic extension of the reference point as the point that is pareto optimal and l... |

199 |
Theory of Value: An Axiomatic Analysis of Economic Equilibrium
- Debreu
- 1959
(Show Context)
Citation Context ... agent a utility that is at least as large as the agent can guarantee for itself from the conflict ��� outcome . Pareto optimality is defined for a bargaining ���¤����� � game=-= in the following manner [6]. 2 Such a perfect curvili-=-near shape is only obtained under the assumptions that the utility functions of the agents are perfectly concave and differentiable. 4sb x 1 a w 2 O a a x 2 ��� ��� ��� ��� �... |

186 | Iterative combinatorial auctions: Theory and practice
- Parkes, Ungar
- 2000
(Show Context)
Citation Context ...h the final choices and the complete sequence (sub-game) of choices are often in equilibrium [45]. However, because representation and reasoning under such a system can be computationally intractable =-=[23, 25, 36]-=- we have been involved in developing approximating decision models for a more limited type of agent that has no explicit representation of the entire game tree. Then, rather than computing the best re... |

179 | Coalitions among Computationally Bounded Agents
- Sandholm, Lesser
- 1997
(Show Context)
Citation Context ...of trust. This algorithmic approach contrasts with the cooperative (axiomatic) and non-cooperative approaches of game theory that have been highly influential in the mechanism design tradition of MAS =-=[23, 24, 43, 48, 58]-=-. Rosenchein and Zlotkin used cooperative game theory to design negotiation mechanisms that maximize the social welfare function (the product of agent utilities, or the Nash solution) for task, state ... |

160 |
Other solutions to nash’s bargaining problem
- Kalai, Smorodinsky
- 1975
(Show Context)
Citation Context ...ict outcome utility [30]. One interpretation of the Nash bargaining solution is that agents are motivated by equity or proportional cooperation [27]. Another solution concept is the Kalai-Smorodinsky =-=[21] which-=- modifies one of Nash’s axioms (independence to irrelevant alternatives to individual monotonicty) and is interpreted as endogenously providing more weighting to the “needier” player [27]. Howev... |

157 |
Similarity relations and fuzzy orderings
- Zadeh
- 1971
(Show Context)
Citation Context ... negotiation decision variables, in the presence of information uncertainty and resource boundedness for multi-dimensional goods. The algorithm itself operates by using the notion of fuzzy similarity =-=[64]-=- to approximate the preference structure of the negotiation opponent and then uses a hill-climbing technique to explore the space of possible trade-offs for the one that is most likely to be acceptabl... |

130 | Negotiation and Cooperation in Multi-Agent Environments
- Kraus
- 1997
(Show Context)
Citation Context ...the above assumptions is that it is not possible to pre-compute an optimal negotiation strategy at design time. Rather the agents need to adopt a heuristic and satisficing approach for their strategy =-=[8, 23]-=-. This is in contrast with deductive models of negotiation where each agent explicitly represents and reasons with the decision tree of the entire game [30, 15]. In this case, a negotiation strategy i... |

105 | FM96.5 A Java-based Electronic Auction House - Rodrı́guez, Noriega, et al. - 1997 |

96 |
An Introduction to Fuzzy Sets: Analysis and Design
- Pedrycz, Gomide
- 1998
(Show Context)
Citation Context ...s a binary, commutative, associative, non-decreasing operation in [0, 1] with 1 as a neutral element. T-norms play a central role in fuzzy set theory in modeling intersection operations on fuzzy sets =-=[38]. 10 ���s� (iii) is nothing but the usual triangular inequality and becomes a pseudo-metric, while for � � ����� ��� � � ���-=-�� � ����������� ��������� � ����������� � ����§�� ����§�� ������� � ��� � �... |

93 | A Service-Oriented Negotiation Model between Autonomous Agents
- Sierra, Faratin, et al.
- 1997
(Show Context)
Citation Context ...ticular actions [37]. In this work we are interested in conflicting preferences over complex multi-dimensional decision problems involved in the bi-lateral resource allocation negotiation of services =-=[50]-=-. In such duopolistic negotiations, one producer and one consumer have to bargain and come to a mutually acceptable agreement over the terms and conditions under which the producer will execute some a... |

93 |
Multi-agent compromise via negotiation
- Sycara
- 1989
(Show Context)
Citation Context ... dividing the dollar, solution quality is based on a satisficing rather than optimal criteria. Uncertainty in negotiation was also addressed by using decision theoretic models in the Persuader system =-=[53]-=- where multi-attribute utility theory was combined with case-based reasoning in contexts where the agent had no previous cases to reason with. This dual approach is similar to our work in that agents ... |

78 | Using Similarity Criteria to Make Negotiation Tradeoffs
- Faratin, Sierra, et al.
- 2000
(Show Context)
Citation Context ...litative value with � valuation . A theoretical analysis shows that the average time the algorithm takes to complete is linear with respect to the number of decision variables in the negotiation (se=-=e [11] for details of the proof). This 9 As the converge-=-nce is asymptotic to the value ������������� ��� , if we had a situation with ��� ����� � we could not guarantee reaching the iso-curve. Also, the sea... |

73 | Distributed Constraint Satisfaction Algorithm for Complex Local Problems
- Yokoo, Hirayama
- 1998
(Show Context)
Citation Context ...ough resolution over only a single variable and the implicit assumption that agents communicate constraints and modify their local solutions cooperatively. Even when multiple variables are considered =-=[61]-=-, the second assumption greatly helps the search process. However, in open systems, agents are motivated to misrepresent their true constraints for selfish reasons. Our trade-off algorithm implements ... |

65 | Efficient Mechanisms for the Supply of Services in MultiAgent Environments
- Vulkan, Jennings
(Show Context)
Citation Context ... can be seen in the large number of proposed models [8, 18]: ranging from auctions in which the agents’ pricing decision problem is solved through showing the dominance of a truthful bidding strateg=-=y [57], -=-to models in which the agents’ argue for positions and aim to persuade their opponents of the value of particular actions [37]. In this work we are interested in conflicting preferences over complex... |

63 | Designing and building a negotiating automated agent
- Kraus, Lehmann
- 1995
(Show Context)
Citation Context ...of trust. This algorithmic approach contrasts with the cooperative (axiomatic) and non-cooperative approaches of game theory that have been highly influential in the mechanism design tradition of MAS =-=[23, 24, 43, 48, 58]-=-. Rosenchein and Zlotkin used cooperative game theory to design negotiation mechanisms that maximize the social welfare function (the product of agent utilities, or the Nash solution) for task, state ... |

62 |
Essays on the foundations of game theory
- Binmore
- 1990
(Show Context)
Citation Context ...e each agent explicitly represents and reasons with the decision tree of the entire game [30, 15]. In this case, a negotiation strategy is then the specification (using for example backward induction =-=[2]-=-) of a sequence of choices for every decision node in the game tree, with the property that both the final choices and the complete sequence (sub-game) of choices are often in equilibrium [45]. Howeve... |

59 |
Automated Service Negotiation Between Autonomous Computational Agents
- Faratin
- 2001
(Show Context)
Citation Context ... parties. The parties first verbalise contradictory demands and then move towards agreements. The prevalence and importance of automated negotiation can be seen in the large number of proposed models =-=[8, 18]: ra-=-nging from auctions in which the agents’ pricing decision problem is solved through showing the dominance of a truthful bidding strategy [57], to models in which the agents’ argue for positions an... |

56 | Market-aware agents for a multiagent world
- Wellman, Wurman
- 1998
(Show Context)
Citation Context ...of trust. This algorithmic approach contrasts with the cooperative (axiomatic) and non-cooperative approaches of game theory that have been highly influential in the mechanism design tradition of MAS =-=[23, 24, 43, 48, 58]-=-. Rosenchein and Zlotkin used cooperative game theory to design negotiation mechanisms that maximize the social welfare function (the product of agent utilities, or the Nash solution) for task, state ... |

49 | Constraint-Directed Negotiation of Resource Reallocations
- Sathi, Fox
- 1989
(Show Context)
Citation Context ...d and, as we have shown in section 4.2, this appears to be a better choice in uncertain environments. The process of negotiation has also been modeled as a distributed constraint satisfaction problem =-=[1, 49, 62]. -=-In the work of Sathi and Fox, agents’ objectives are represented as constraints together with their associated utilities. Strategies (e.g. composition, reconfiguration and relaxation operators) are ... |

48 | Bargaining with limited computation: Deliberation equilibrium
- Larson, Sandholm
(Show Context)
Citation Context ...h the final choices and the complete sequence (sub-game) of choices are often in equilibrium [45]. However, because representation and reasoning under such a system can be computationally intractable =-=[23, 25, 36]-=- we have been involved in developing approximating decision models for a more limited type of agent that has no explicit representation of the entire game tree. Then, rather than computing the best re... |

47 |
2000) “A multi-attribute utility theoretic negotiation architecture for electronic commerce
- Barbuceanu, Lo
(Show Context)
Citation Context ...ction. In particular, for � ������� � � ����� ������� � � � § � (Lukasiewicz t-norm) property 8 A triangular norm, t-norm for short, is a binary, c=-=ommutative, associative, non-decreasing operation in [0, 1] with 1 -=-as a neutral element. T-norms play a central role in fuzzy set theory in modeling intersection operations on fuzzy sets [38]. 10 ���s� (iii) is nothing but the usual triangular inequality and ... |

35 |
Searching for Joint Gains in Multi-party Negotiations
- Ehtamo, Kettunen
- 2001
(Show Context)
Citation Context ...ne emanating from � the similarity based trade-off execution trace, and the line joining ������§ � to ��§���� � the pareto-optimal line. The pareto-optimal line was com=-=puted using the weighted method [41, 7]. The -=-output of the algorithm, ��� , is shown in figures 5 and 6 (top row) as the end point of the execution trace and for 6 (bottom row) as the explicitly marked points (since there is no trace). For... |

30 |
Autonomous Agents for Business Process Management” Int
- Jennings, Faratin, et al.
- 2000
(Show Context)
Citation Context ...a social choice. These protocol and agent assumptions were necessary in order to design a negotiation system for the types of real world problems we have been involved in: business process management =-=[19]-=-, telecommunications network management [9], and e-commerce [35, 42]. These assumptions are as follows. Firstly, agents have only limited information about their negotiation opponent. Although a mecha... |

29 |
A Primer in Game Theory, Harvester Wheatsheaf
- Gibbons
- 1992
(Show Context)
Citation Context ... possible outcomes lie on the pareto-optimal line. Furthermore, assuming linear conflicting value functions for the negotiation participants, the sum of each outcome is § (i.e., it is a zero-sum game=-= [13]). In this scenario, the Nash barg-=-aining solution is easily computed as the mid point (and most equitable) of both agents’ value function (i.e., at ��������������� � ). Given the single decision varia... |

25 |
A logic of knowledge and belief for recursive modeling: A preliminary report
- Gmytrasiewicz, Durfee
- 1992
(Show Context)
Citation Context ...ur is difficult. The most common means of doing this is to construct an explicit model of the negotiation opponent and then update and refine this model in the light of subsequent interactions (e.g., =-=[2, 14]-=-). However, such models are difficult and computationally demanding to construct (especially for multi-dimensional goods), they are not well suited to situations where an agent negotiates with many op... |

20 | Representing aggregate belief through the competitive equilibrium of a securities market
- Pennock, Wellman
- 1997
(Show Context)
Citation Context ...gents trade commodities for given announced prices. Prices are then iteratively lowered or increased with excess supply or excess demand respectively, until the market clears at a general equilibrium =-=[5, 39]. -=-The First Fundamental Theorem of Welfare Economics then states that given consumers’ preferences are well behaved, trading in perfectly competitive markets implements a Pareto-Optimal allocation of ... |

19 | Automated Negotiation for Provisioning Virtual Private Networks using FIPA-Compliant Agents
- Faratin, Jennings, et al.
- 2000
(Show Context)
Citation Context ...sumptions were necessary in order to design a negotiation system for the types of real world problems we have been involved in: business process management [19], telecommunications network management =-=[9]-=-, and e-commerce [35, 42]. These assumptions are as follows. Firstly, agents have only limited information about their negotiation opponent. Although a mechanism can theoretically be designed to incen... |

18 |
Search Algorithms for Agents
- Yokoo, Ishida
- 1999
(Show Context)
Citation Context ...d and, as we have shown in section 4.2, this appears to be a better choice in uncertain environments. The process of negotiation has also been modeled as a distributed constraint satisfaction problem =-=[1, 49, 62]. -=-In the work of Sathi and Fox, agents’ objectives are represented as constraints together with their associated utilities. Strategies (e.g. composition, reconfiguration and relaxation operators) are ... |

18 |
Bargaining: Formal Theories of Negotiation
- Young, ed
- 1975
(Show Context)
Citation Context ...e one-off nature of encounters which prevents the update of prior distributions. Secondly, the formulation of decisions based on subjective expected utility introduces the silent out-guessing problem =-=[63]—t-=-he agent designer’s choice of probabilities is based on guesses about the probable choices of others, whose choice in turn is dependent on the guesses about the probable choices of the first, and so... |

17 |
A framework for social motives
- R, Messick
- 1976
(Show Context)
Citation Context ...o player � for � � � settlement and � is player ’s conflict outcome utility [30]. One interpretation of the Nash bargaining solution is that agents are motivated by equity or proportional =-=cooperation [27]. -=-Another solution concept is the Kalai-Smorodinsky [21] which modifies one of Nash’s axioms (independence to irrelevant alternatives to individual monotonicty) and is interpreted as endogenously prov... |

15 | Decentralized method for computing pareto solutions in multiparty negotiations
- Heiskanen
- 1999
(Show Context)
Citation Context ...more recently, autonomous computational systems. Negotiation in operational research is viewed as an optimization problem solved through the design of (mostly centralized) optimal solution algorithms =-=[7, 17, 23, 54]-=-. These algorithms, based on mathematical programming techniques, are often optimal because: a) the geometry of the solution set is assumed to be described by a closed and convex set (therefore there ... |

15 |
Identifying Pareto-optimal settlements for two-party resource allocation negotiations
- Teich, Wallenius, et al.
- 1996
(Show Context)
Citation Context ...more recently, autonomous computational systems. Negotiation in operational research is viewed as an optimization problem solved through the design of (mostly centralized) optimal solution algorithms =-=[7, 17, 23, 54]-=-. These algorithms, based on mathematical programming techniques, are often optimal because: a) the geometry of the solution set is assumed to be described by a closed and convex set (therefore there ... |

14 |
A Logic for Approximated Reasoning”, in: The
- Ying
- 1994
(Show Context)
Citation Context ... It has, however, been used in work on psychological studies of human behaviour [55], mathematical work on graded extensions of equivalence relations [56, 64], and as a model of approximate reasoning =-=[22, 59]-=-. From the philosophical perspective, Niiniluoto relates similarity with the broader area of analogical reasoning [34]. Finally, although similarity has been frequently used to model case-based reason... |