Human Agent Negotiation

Abstract

Automated agent that can negotiate with human or even act on behalf of human has attracted many researchers for decades. A number of approaches and strategies have been developed to cope with the negations in various contexts. This paper presents the survey of automated negotiation in a general perspective. It captures the core theories and components for which mandatory in every automated negotiation systems. It also discusses weakness and strength of each approach. However, this survey does not aim for theoretical economics background; therefore some of economic theories that may be mentioned will not be argued in depth on this paper.

1 Introduction

Negotiation involves finding a compromise which leads to a mutual agreement on conflict of interests for the participants. The term of automated negotiation has become dramatically significant when computer technologies make almost everything turn into possibility. Automated system agents then have been developed to serve human with various purposes including trading, training, or even providing support to human in organizations in many aspects (The electric Elves has been recently developed to assist human with day-to-day task in an organization see [12]) for example; saving cost for organization by reducing human involvement in operational procurement. [9] Many buy and sell agents for online market have also been introduced to the public (for example; Persona Logic, Firefly, Bargain Finder, Jango, Kasbah, Auction Bot, T@T) [10]

Many years ago, Chavez and Maes have developed Kasbah [11], which is an online multi-agent for buyer and seller with simplicity of negotiation protocol. The negotiation process is straightforward and is not rich with negotiation strategy. However it represented the beginning of automated agent innovation to the world of business.

Figure 1. Kasbah: One of the first agents for online marketplace developed by A.Chavez and P.Maes in MIT Media Lab [10]

For a clearer view on the negotiations in computational context, we may categorize it into three bases by considering the participation level of automated agent in decision making process. The first way is Human-to-human negotiation where the human can make a final decision by themselves without control from the system agents. However the options are provided by a computer system. These software systems are called Negotiation Support Systems (NSS). (more detail regarding NSS can be found in [2]) The second is Agent-to-agent, generally referred to as autonomous agents, where the computer system automatically makes the final decision without human interference. Finally, the Agent-to-human negotiation is the basis where the agents act on behalf of one party to negotiate with the human counterparts. The examination in this paper draws on the latter basis.[11]

Automated negotiation is a very broad subject. There are too many models and theories that have been developed to cover in one single paper. Thus, this paper only represents the observation only on the center of creating automated agents.

In the following section, we will look at the negotiation domain and some of its significant characteristics

2 Negotiation Domain

Many interesting computer science articles have given explanation of negotiation domain but have differing views. In general concept, domain of automated negotiation or in others words the environment where the agents operate, basically encompasses a set of negotiation objects and the issues being covered by those objects. If the negotiation objects contain multi-attribute issues, the negotiation mechanism tends to be able to make complex decision. [9] However, there are other parameters that play important roles in the negotiation environment and could be influences on the participants as well as the strategy being used in the negotiation process.

The negotiation scenario can be either cooperative or non-cooperative. It is cooperative when the negotiators try to maximise the payoffs of all partners, this is said to be "win-win" approach. On the other hand, in a non-cooperative scenario, each of negotiators interact together in order to maximise their own payoffs alone, this is said to be "win-lose" approach. [4,9]

The negotiation also can be conducted between two participants (bilateral negotiations) or among many participants (multilateral negotiations). [9] For example in the general market there can be one seller and one buyer negotiate over an item. In contrast, in the auction there can be many buyers taking part in negotiating over an item.

Time is also