Figure 1 - uploaded by Catholijn M. Jonker
Content may be subject to copyright.
Source publication
This paper presents two experiments that contribute to the comparison of human- versus computer behavior in the domain of multi-issue negotiation. The experiments are part of an ongoing endeavor of improving the quality of computer negotiators when negotiating against human negotiators. The validity of the experiments was tested in a case study of...
Contexts in source publication
Context 1
... the top level, SAMIN consists of three components: an Acquisition Component, an Analysis Component and a Presentation Component, see Figure 1 Here, the solid arrows indicate data flow. The dotted arrows indicate that each component can be controlled separately by the user. ...
Context 2
... the Presentation Component is used to present the results of the analysis in a user-friendly format. Furthermore, SAMIN maintains a library of properties, templates of properties, bid ontologies, and profile ontologies (not shown in Figure 1. The working of the three components will be described briefly in the next subsections. ...
Similar publications
The design and development of open multiagent sys-tems is one of the main areas in multiagent research. Human soci-eties have successfully coped with a similar issue by creating insti-tutions, which can broadly be understood as formal or semi-formal frameworks that provide commonly understood collections of rules within which people can interoperat...
In this research we explored a methodological proposal whose main purpose is to show and clarify the use of Information Technology and Communication in teacher education, using Conversational Software Agents and Social Network Analysis. In this context, the real-time work with the teacher is emphasized, and is feasible therefore to generate a pract...
Nowadays, with the expansion of Internet, there is a need of methodologies and software tools to ease the development of applications where distributed (human or software) agents can participate. Electronic institutions of multiple agents, can play a main role in the development of systems where normative specifications play a vital role. Electroni...
Citations
... It is a core activity in human society and widely exists in social and organizational settings. Automated negotiation [14] involves intelligent agents negotiating on behalf of humans, aiming to not only save time and effort for humans but also yield better outcomes than human negotiators [8]. Automated negotiation can play an important role in application domains, including supply chain, smart grid, digital markets, and autonomous driving. ...
Understanding an opponent agent helps in negotiating with it. Existing works on understanding opponents focus on preference modeling (or estimating the opponent's utility function). An important but largely unexplored direction is recognizing an opponent's negotiation strategy, which captures the opponent's tactics, e.g., to be tough at the beginning but to concede toward the deadline. Recognizing complex, state-of-the-art, negotiation strategies is extremely challenging, and simple heuristics may not be adequate for this purpose. We propose a novel data-driven approach for recognizing an opponent's s negotiation strategy. Our approach includes a data generation method for an agent to generate domain-independent sequences by negotiating with a variety of opponents across domains, a feature engineering method for representing negotiation data as time series with time-step features and overall features, and a hybrid (recurrent neural network-based) deep learning method for recognizing an opponent's strategy from the time series of bids. We perform extensive experiments, spanning four problem scenarios, to demonstrate the effectiveness of our approach.
... Section 7 concludes the paper with future work directions. Bosse and Jonker [2005] develop the current benchmark of artificial negotiators. They analyze human and computer behavior in multi-issue negotiation by conducting two different set of experiments. ...
Designing agents aiming to negotiate with human counterparts requires additional factors. In this work, we analyze the main elements of human negotiations in a structured human experiment. Particularly, we focus on studying the effect of negotiators being aware of the other side’s gain on the bidding behavior and the negotiation outcome. We compare the negotiations in two settings where one allows human negotiators to see their opponent’s utility and the other does not. Furthermore, we study what kind of emotional state expressed and arguments sent in those setups. We rigorously discuss the findings from our experiments.
... Aspire is also a negotiation support system and not a fully automated negotiation system. Bosse and Jonker (2005) compare the dynamics of agent negotiation with human negotiation. For this purpose, the negotiation process is formalized by introducing states. ...
... Linear utilities are assumed in most of the works in negotiation literature (e.g. Yu et al. 2013;Matos et al. 1998;Yan et al. 2007;Klein et al. 2003;Restificar and Haddawy 2004) but there are many works that consider non-linear utility functions (Bosse and Jonker 2005;Sánchez-Anguix et al. 2013;Klein et al. 2003;Ito et al. 2007;Lai et al. 2008). ...
... Apart from the above heuristic models, several other heuristics have been used in the literature. The opponent model proposed in (Jonker et al. 2007;Bosse et al. 2005;Jonker and Robu 2004) use a "guessing" heuristic in which preference order of an opponent is predicted. It is based on the heuristic that an opponent will be more willing to concede on less important preferences and vice versa. ...
Negotiation is a complex process. The decision making involved in several stages of negotiation makes its automation complex. In this paper we present a lifecycle model of a negotiation agent in which we identify the individual components that comprise automated negotiation and the interactions between those components. We present a survey of methods used in the automated negotiation literature fitting them to the components of our lifecycle model. While discussing the opponent modeling component, we present the taxonomy of opponent models. The lifecycle model is generic enough to accommodate most of the frameworks in the literature. To this end we fit the methods used in some of the automated negotiation frameworks in the literature to the lifecycle.
... Whether the negotiating agent can outperform the human counterpart is another concern for researchers in the field of human-computer negotiation [7,29,56,57]. Extensive experiments have been conducted to test whether the proposed negotiation agents possess enough intelligence to handle complicated negotiation situations. As discussed in Section The Meta-Requirements, the biggest obstacle in protecting the seller agent's benefit is the non-monotonic concession with bad faith in the impasse situation. ...
... In real-world negotiation, time (equivalent to the number of negotiating rounds in our study) is a major concern of the negotiators [7,14,30], because how long the negotiation takes directly relates to the negotiator's psychological feeling as well as how much cost the parties can bear. According to [56], a competitive strategy negotiator does not make substantive concessions until the deadline approaches; the negotiation process tends to take longer time to complete. ...
... Utility obtained by the negotiators when a negotiation finishes is an important metric that reflects how much the parties gain from the negotiation [14,30] and has been used as an evaluation metric in many studies [7,28,56,61,62]. As a reasonable inference of Hypothesis 2, as the agent with portfolio strategy could outperform human in negotiations, we have the reason to believe that the implementation of the proposed portfolio strategy agent will result in higher seller agent utility compared with those of the single-strategy agents, and thus we have the following: Hypothesis 4: The portfolio strategy seller agent would obtain higher utility than the one implementing the single strategy (competitive, collaborative, or selection) when negotiating with a human buyer. ...
Human–computer negotiation has the potential to play an important role in today’s highly dynamic online environment, especially in business-to-consumer e-commerce transactions. However, the lack of research on effective automated negotiation algorithms to respond to human buyers’ strategic and/or tactic offers has limited the development of automated human–computer negotiation systems for real-world applications. Intelligent software agents that are capable of dynamically adjusting their negotiation strategy in response to human buyers’ offers can greatly improve the negotiation experience of human buyers. In this study, guided by design science principles, we design a portfolio strategy model, which implements four negotiation strategies (i.e., time-dependent, behavior-dependent, dynamic time-dependent, and impasse resolution) as the core of our software agent for negotiating with human buyers. To evaluate this novel model, we implement a prototype of the system and compare it with three benchmark single-strategy models (i.e., competitive, collaborative, and selection) in human–computer negotiation experiments. The results show that our model not only enables the software agent to outperform its human counterpart but also significantly increases the settlement ratio and the joint outcome of both parties.
... Due to the expensive nature of negotiation, attention to automating the process has gained considerable traction in the past twenty years [8], since the development of Contract Net Protocol by Smith in the 1980s [75]. It spurned the promise of finding better outcomes than human negotiators [9,18,32,55,79,41]. ...
Negotiation is a process where agents aim to work through disputes and maximize their surplus. As the use of deep reinforcement learning in bargaining games is unexplored, this paper evaluates its ability to exploit, adapt, and cooperate to produce fair outcomes. Two actor-critic networks were trained for the bidding and acceptance strategy, against time-based agents, behavior-based agents, and through self-play. Gameplay against these agents reveals three key findings. 1) Neural agents learn to exploit time-based agents, achieving clear transitions in decision preference values. The Cauchy distribution emerges as suitable for sampling offers, due to its peaky center and heavy tails. The kurtosis and variance sensitivity of the probability distributions used for continuous control produce trade-offs in exploration and exploitation. 2) Neural agents demonstrate adaptive behavior against different combinations of concession, discount factors, and behavior-based strategies. 3) Most importantly, neural agents learn to cooperate with other behavior-based agents, in certain cases utilizing non-credible threats to force fairer results. This bears similarities with reputation-based strategies in the evolutionary dynamics, and departs from equilibria in classical game theory.
... To our knowledge, the first experiment pairing humans with agent counterparts in negotiation settings was reported in (Byde, Yearworth, Chen, & Bartolini, 2003). In another study an experiment was designed for comparing the performance of agents vs. humans in agent-human negotiations (Bosse & Jonker, 2005). The findings suggested that agents were able to achieve more fair agreements. ...
Human-computer negotiation have attracted a substantial attention in the recent past. This study is part of an ongoing endeavor of testing the outcomes of computer agents when they negotiate electronically with humans. Specifically, the paper studies the effects of various time spans and negotiation tactics on the process and outcomes of human-software agent negotiations. The paper reports the results of the experiments where the subjects negotiated a purchase of the mobile plan with software agents acting as sellers. Three various time-based negotiation tactics were employed in three different time span (the time period set for the negotiations) settings. The results suggest that both concession-making tactics and the time span have significant effects on the outcomes of the negotiations, including agreement utilities and agreement rate.
... Pairing humans with agent counterparts in negotiation settings has been reported in [2]. In another study an experiment was designed for comparing the performance of agents vs. humans in agent-human negotiations [3]. In particular, agents were able to achieve more fair agreements. ...
Prior research in assessing the prospects of software agent vs. human negotiations examined the effectiveness of using agents in negotiations with humans. In this paper we study the impacts associated with using a person's photograph for representing an agent in negotiations. An experiment employing photographs of a younger, a mature, and an older male person for representing agents was conducted. The findings suggest significant differences in terms of the agreement utilities achieved by the three groups of agents. Hence, the choice of an image for representing a software agent has a significant impact on the outcomes of human vs. agent negotiations.
... As a result, the negotiation traces of ANAC 2011 showed to a very large extent the stochastic behavior discussed above (see also Figure 1). Only 25% of the negotiation moves were an improvement for the opponent over the previous bid; the other 75% of the moves could be classified as selfish, unfortunate, or silent [4]. ...
In the previous chapter we classified generic acceptance conditions, and we formulated new ones that performed better in an experimental setting. This chapter takes a different approach by devising theoretically optimal solutions. We approach the decision of whether to accept as a sequential decision problem, by modeling the bids received as a stochastic process. We argue that this is a natural choice in the context of a negotiation with incomplete information, where the future behavior of the opponent is uncertain. We determine the optimal acceptance policies for particular opponent classes and we present an approach to estimate the expected range of offers when the type of opponent is unknown. We apply our method against a wide range of opponents, and compare its performance with acceptance mechanisms of state-of-the-art negotiation strategies. The experiments show that the proposed approach is able to find the optimal time to accept, and improves upon widely used existing acceptance mechanisms.
... E-commerce oriented negotiation is increasingly assuming a pivotal role in many organizations, and a number of prominent negotiation models have been developed over the past decades [3]. The human-computer negotiation plays a paramount role in the e-commerce oriented applications, especially in the B2C context where software agents act as business provider [4]. Comparing with the traditional online sales mode where customers view the basic product or service information on the website and often need to negotiate with human salespeople through a "contact us" link, the human-computer automated negotiation system can help business organizations to reduce the labor cost for negotiation and greatly increase the transaction efficiency to the optimum extent. ...
Human-computer
negotiation plays an important role in B2C e-commerce. There is a paucity of
further scientific investigation and a pressing need on designing the software
agent that can deal with the human’s random and dynamic offer, which is
crucially useful in human-computer negotiation to achieve better online
negotiation outcomes. The lack of such studies has decelerated the process of
applying automated negotiation to real world applications. To address the
critical issue, this paper develops a dynamic time-dependent strategy
concession model, that can predict the human’s negotiation behavior during the
process of the negotiation. To demonstrate the effectiveness of this model, we
implement a prototype and conduct human-computer negotiations over 121
subjects. The experimental analysis not only confirms our model’s effect but
also reveals some insights into future work about human-computer negotiation
systems.
... Traditionally, negotiation is a necessary, but time-consuming and expensive activity. Therefore, in the last two decades, there has been a growing interest in the automation of negotiation and e-negotiation systems [18,71,91,99,107], for example in the setting of e-commerce [20,79,105,126]. This attention has been growing since the beginning of the 1980s with the work of early adopters such as Smith's Contract Net Protocol [186], Sycara's PERSUADER [189,190], Robinson's OZ [164], and the work by Rosenschein [168] and Klein [102]. ...
... This attention has been growing since the beginning of the 1980s with the work of early adopters such as Smith's Contract Net Protocol [186], Sycara's PERSUADER [189,190], Robinson's OZ [164], and the work by Rosenschein [168] and Klein [102]. The interest is fueled by the promise of automated agents being able to negotiate on behalf of human negotiators and to find better outcomes than human negotiators [20,55,89,121,126,149,192]. ...
A negotiation between agents is typically an incomplete information game, where the agents initially do not know their opponent’s preferences or strategy. This poses a challenge, as efficient and effective negotiation requires the bidding agent to take the other’s wishes and future behavior into account when deciding on a proposal. Therefore, in order to reach better and earlier agreements, an agent can apply learning techniques to construct a model of the opponent. There is a mature body of research in negotiation that focuses on modeling the opponent, but there exists no recent survey of commonly used opponent modeling techniques. This work aims to advance and integrate knowledge of the field by providing a comprehensive survey of currently existing opponent models in a bilateral negotiation setting. We discuss all possible ways opponent modeling has been used to benefit agents so far, and we introduce a taxonomy of currently existing opponent models based on their underlying learning techniques. We also present techniques to measure the success of opponent models and provide guidelines for deciding on the appropriate performance measures for every opponent model type in our taxonomy.
