Abstract
For the study of automated negotiation, which though constituting a very specific form still belongs to the realm of negotiation, the framework of negotiation research — with necessary adaptations and inevitable restrictions — is applicable. Negotiation research in general — though some approaches focus only on components and their relations, neglecting others as discussed in Chapter 1 — studies how the preconditions of the negotiation and the context in which it takes place influence aspects of the negotiation process, which in turn shapes the outcome of the negotiation (see Figure 3).
This is a preview of subscription content, log in via an institution.
Preview
Unable to display preview. Download preview PDF.
References
However, it is also possible to determine the negotiation problem in an automated ‘metanegotiation’ before the actual negotiation, where the content of this meta-negotiation is the negotiation object of the subsequent actual negotiation. Mechanisms that alter the negotiation problem during the negotiation are currently under development [65, 185], but were not used in simulation studies yet, and actually are not elaborated sophisticatedly enough to use them in simulations at the moment. At the current state of the art of automated negotiation the issues and settlement options, as well as the preferences over this negotiation object have to be constant during the negotiation and communicated by the user to its software agent. An alternative to meta-negotiations for determining the negotiation object would be an automated determination by the system used for automated negotiation. For this purpose the software agent has to elicit the preferences of the user (about all issues and options regarded to be important by the user) as well as their outside option — their ‘best alternative to a negotiated agreement’ (BATNA) — before the negotiation. This preference information then has to be communicated by the software agents to a main routine of the negotiation system that is in charge of determining the negotiation object. With preference information for all negotiation parties the system constructs a negotiation object so that only the conflicting issues and only feasible options remain by the following procedure: (i) issues of importance to only one party are settled for the best option for this party in advance, (ii) issues where the parties indicate the same best option are settled for this best option, (iii) the option spaces for the remaining issues, for which conflicting interests exist are set to the upper and lower bound indicated by the parties reservation levels so that only feasible settlements remain in the negotiation object, and finally (iv) in a last step all agreements that do not satisfy the participation constraints of either party (i.e. afford lower utility than the BATNAs of the parties) are deleted from the set of possible agreements. As the preferences used for constructing the negotiation object have to be indicated to the software agents in a first step, which follow them during the automated negotiation, an user only would penalize himself by indicating wrong preferences. E.g. in case of misrepresented reservation levels or BATNAs no zone of possible settlements could be the outcome of the automated negotiation object determination by the system and therefore no agreement can be reached even if according to the true preferences possible settlements would be found, which then would be better than the actual BATNA, in case of misrepresented issue weights or misrepresented partial utility values for options the agreement of the automated negotiation might change to an inferior one as the agent follows these wrong preferences.
As mentioned in the introduction we focus on simulation studies in this review as analytical models apply restrictive assumptions and operative systems for automated negotiation are not available yet — besides few simplistic experimental systems.
Searching in abstracts is possible in JSTOR, EconLit, and ABI Informs/ProQuest.
For scholar.google in many cases there existed multiple results for the same publication.
These 37 publications build the basis for the detailed review in the subsequent section and are marked with an asterisk in the bibliography.
In Figure 3.3 we only illustrate the development until 2006 as the literature search was performed on July 9th, 2007 and we therefore have no complete record of publications in 2007. Furthermore note that given the fast development of research on automated negotiation this review — conducted at the very beginning of this book project — could be criticized as already out-dated. However, recently published simulation studies [119, 125, 169] in scientific journals relevant for the field like Group Decision and Negotiation or Decision Support Systems in the few first months of 2009, on the one hand indicated that the field remains highly important, and on the other hand that the major shortcomings detected in the subsequent review still are unresolved.
In their classification Jennings and colleagues term this concept ‘negotiation object’ [96]. However, they only consider single-issue negotiations in their studies, where objective and subjective valuation of the single issue coincide, when considering multi-issue negotiations also the preferences over these multiple issues have to be considered, and consequently ‘negotiation problem’ is the more adequate term for this component of automated negotiations.
The reduction of complexity achieved through discretizing issues with continuous options refers only to the number of possible solutions of the negotiation problem — which becomes finite while being infinite otherwise — still the negotiation problem could remain complex in other aspects.
If the negotiation object is allowed to be manipulated by the software agents the interaction protocol also has to determine the rules for negotiation object manipulation, i.e. the rules that govern the addition or elimination of issues and options during the negotiation process.
Consult [113] and [86] for information on auctions and the ultimatum game, respectively.
In the Zeuthen-Nash bargaining protocol negotiation ends if both parties refuse to make a concession in one round.
As briefly discussed in the Section 3.1.1.3 there exists some first attempts to make software agents elicit their users’ preferences or learn them from previous experience [83, 134], besides just asking the user to input them.
Consult the Section 3.1.2.3 for details on the progression of negotiations determined by the interaction protocol.
Branches of game theory investigate for a given negotiation problem and interaction protocol what the optimal (equilibrium) strategy of rational agents should be, given rational behavior of the opponent — strategic approach in bargaining theory —, or how interaction protocols should be configured to achieve a desired outcome for a given strategy and negotiation problem — mechanism design [21, 45, 170].
Moreover other than just additive utility models have to be investigated. Additive utility models are easy to handle and could be a good approximation of the user’s real preferences [99], which justifies to use them in simulation systems, however, operative systems should support alternative models that might fit the user’s real preferences better.
Furthermore operational validation is not possible as operative systems are not available yet, but current research focuses on system design, and therefore the real system’s and simulation’s outputs cannot be compared.
Alternatively they have to compensate inferior outcomes with correspondingly lower costs.
Model validation considerations contribute a further argument in favor of termination by the agents — in real world there also exist no (absolutely) fixed deadlines. However, to avoid infinite negotiations, the negotiation protocol could overrule the agents decisions if they fail to make progress and still not terminate the negotiation. Here one rule adopted from the Zeuthen-Nash bargaining game could be, that the protocol terminates negotiations if the two agents both reject the last offer of the opponent — see Chapter 4.
For example the fair concession making approach proposed by [12] in his’ simple model of negotiation’ could be applied for this purpose. Basing on the egalitarian norm of reciprocity [12] proposes that if the opponent makes an unfairly small concession, an agent should stop to make further concessions and wait until the opponent catches up. A concession is unfairly small if the reduction of utility between the last two offers of the opponent is smaller than the reduction of utility between the last two offers of the focal software agent.
Author information
Authors and Affiliations
Rights and permissions
Copyright information
© 2010 Springer-Verlag/Wien
About this chapter
Cite this chapter
Filzmoser, M. (2010). Automated Negotiation. In: Simulation of Automated Negotiation. Springer, Vienna. https://doi.org/10.1007/978-3-7091-0133-9_3
Download citation
DOI: https://doi.org/10.1007/978-3-7091-0133-9_3
Publisher Name: Springer, Vienna
Print ISBN: 978-3-7091-0132-2
Online ISBN: 978-3-7091-0133-9