Learning is required for enhancing the performance of agents in a situation when there is no complete information about the preferences or decision-making processes of other agents. Negotiation agents and dynamic pricing agents often do not have complete information about the preferences, deadline, reserved price, and decision-making processes of other agents. In automated negotiation as well as dynamic pricing systems, learning is essential because in adjusting and adapting their strategies during negotiation and trading, agents are more likely to achieve better outcomes and increase their payoffs. Even though there are many existing e-negotiation agents for e-commerce, many of them were not designed with the ability to adapt and enhance their strategies through learning.
Learning approaches such as reinforcement learning, belief-based learning, and evolutionary learning can potentially enhance the performance of negotiation as well as dynamic pricing agents. Negotiation agents using reinforcement learning enhance their performance through trial-and-error interactions with other agents. In belief-based learning, agents keep track of the history of previous actions of other agents and form beliefs about what other agents will do in the future based on past observations. Based on these observations, they tend to select a bestresponse strategy that maximizes their expected payoffs given the beliefs they have formed. In evolutionary learning, agents evolve and derive effective strategies using genetic algorithms.
This special issue brings together researchers in multiagent learning, automated negotiation, and game-theoretic approaches of multiagent systems to present the latest research results in developing advanced negotiation agents as well as dynamic pricing systems. It serves to highlight recent research achievements in multiagent learning techniques and Pareto-search methods for automated negotiation and dynamic pricing systems.