Factors and Business Impacts in Human-Computer Negotiations

Negotiation commonly takes place where there are competing interests. Negotiations require a substantial amount of cognitive effort and time commitment. Artificial Intelligence (AI) has recently been experiencing a dramatic rise. AI and computer agents may significantly affect how negotiations are conducted. Agents can exhibit human-like behavior and follow the preferences of the principals and predefined strategies, goals, and constraints. For example, some companies already used computer sales assistant to help customers and even negotiate the price and other features online. The purpose of this thesis is to contribute to the transformation of the negotiation process from human vs human to human vs computer, in the context of e-commerce. By investigating various factors that influence human-computer negotiations and the impact of these factors on negotiation outcomes, the current thesis can shed light on the cognitive process underneath human-computer negotiation in the context of online purchasing. The work of this thesis is organized into three major components. As its first component, this thesis conducted a thorough search of state-of-the-art literature on human-computer negotiation and proposed a framework for future studies. Based on prior research, a list of various kinds of computer agent attributes that may influence negotiation results and the relationships between these factors and negotiation outcomes were proposed. In addition to computer agents’ attributes, this essay included past literature that studied human participants’ individual differences and the influence of such differences. Based on the Technology Acceptance Model (TAM), this essay investigated the development of human-computer negotiation and human participants’ acceptance and perception of a computer agent. At the end of the first essay, an overall research framework is presented. Based on the framework of essay 1, an experiment was conducted in essay 2 to investigate how various agent strategies, tactics and configurations influence the outcomes of negotiations. Specifically, essay 2 investigated the effects of negotiation tactics (concession pattern/curve), synchronous vs. asynchronous modes, and solution-search mechanisms (search between multiple issues or dive into one issue at a time) on the subjective and objective outcomes of human-computer negotiations. A 3×2×2 experiment was conducted where the subjects could negotiate the purchase of a mobile plan with computer agents acting as sellers. In this experiment, three time-based negotiation concession patterns and two solution-search mechanisms were employed in synchronous vs. asynchronous mode. On the other hand, the negotiation results were evaluated from multiple levels. Specifically, not only the overall result at group level but also the result at individual level were included in this research. On the individual level, in addition to objective measurements, subjective measures of negotiation results, such as usefulness and intention to use, were also adopted. A model was generated and tested based on TAM and a so-called TIMES framework (Task, Individuals, Mechanism, Environment, and System). Essay 3 investigates a construct named “implicit power” and the influence of implicit power in the context of online purchasing where humans negotiate with computer agents. Implicit power refers to perceived power gained indirectly through hints in the exchange of offers. In most of the prior research, when researchers talked about power, they meant the kind of power that can be gained directly through communication during negotiation. But there is another kind of power that is implicitly perceived by the other party through ways other than communication and influences negotiations as well. After introducing implicit power, a model was built to test the influence of implicit power of both negotiation parties: humans and computers. Specifically, a 2×4×3 experiment was conducted. Several aspects of implicit power were studied, including anchoring, agent avatar image power, and the power of human subjects’ personality. In the experiment, the subjects negotiated the purchase of a laptop with computer agents acting as sellers. Two anchoring conditions and four different avatar images were used to test the influence of computer agents’ implicit power. As the source of human’s intrinsic power, the participant’s personality (Social Value Orientation) was also tested in three different types: prosocial, individualistic, and competitive. This research proposed the concept of implicit power and studied the influence of several kinds of implicit power. The model built in this research shows a good ability to explain the variance in the dependent variable (R-square: 0.44)