Towards Understanding Negotiation Strategies

Abstract

With modern computers becoming more powerful, automated negotiation is coming more to the forefront of research. Software agents are able to aid human negotiators, help them achieve better outcomes and even negotiate on their behalf. An active topic in the field of automated negotiation concerns the development of negotiation strategies. While negotiation strategies and their components have been investigated before, there is to our knowledge no substantial work that analyzes the relations between the components. This thesis investigates these relationships, contributing to the field in two ways. First, it introduces the BOA architecture, which defines and isolates the key components of a negotiation strategy. To validate this architecture we have shown that many well-known and state-of-the-art negotiation strategies fit into the BOA architecture. Second, it investigates how the interactions of these key components influence an agent's performance. We examined the effects of combining different components and we determined as well which key component contributes the most to the agent's performance. We have reached the important conclusion that combining better performing components does lead to better performing negotiation strategies. Furthermore, we found a significant difference in the amount of contribution each key component makes to the performance of a negotiation agent. Throughout our research we have also found that the interactions between components play a large role as well in the performance of a negotiation strategy and that these interactions should not be overlooked.