Purpose
The purpose of this paper is to introduce a new agent-based simulation model of bilateral negotiation based on a synthesis of established theories and empirical studies of negotiation research. The central units of the model are negotiators who pursue goals, have attributes (trust, assertiveness, cooperativeness, creativity, time, etc.) and perform actions (proposing and accepting offers, exchanging information, creating value, etc).
Design/methodology/approach
Methodologically, the model follows the agent-based approach to modeling. This approach is chosen because negotiations can be described as complex, non-linear systems involving autonomous agents (i.e. the negotiators), who interact with each other, pursue goals and perform actions aimed at achieving their goals.
Findings
This paper illustrates how the model can simulate experiments involving variables such as negotiation strategy, creativity, reservation value or time in negotiation. An example simulation is presented which investigates the main and interaction effects of negotiators’ reservation value and their time available for a negotiation. A software implementation of the model is freely accessible at https://tinyurl.com/y7oj6jo8.
Research limitations/implications
The model, as developed at this point, provides the basis for future research projects. One project could address the representation of emotions and their impact on the process and outcome of negotiations. Another project could extend the model by allowing negotiators to convey false information (i.e. to bluff). Yet another project could be aimed at refining the routines used for making and accepting offers with a view to allow parties to reach partial settlements during a negotiation.
Practical implications
Due to its broad scope and wide applicability, the model can be used by practitioners and researchers alike. As a decision-support system, the model allows users to simulate negotiation situations and estimate the likelihood of negotiation outcomes. As a research platform, it can generate simulation data in a cost- and time-effective way, allowing researchers to simulate complex, large-N studies at no cost or time.
Originality/value
The model presented in this paper synthesizes in a novel way a comprehensive range of concepts and theories of current negotiation research. It complements other computational models, in that it can simulate a more diverse range of negotiation strategies (distributive, integrative and compromise) and is applicable to a greater variety of negotiation scenarios.
A Deep Reinforcement Learning Approach to Concurrent Bilateral Negotiation