Study about Intelligent Virtual Agent: Learning How to Make Compromises

The research aims to empower the learning agent to control the way of reaching a goal. More specifically objectives of the study are to explore the possibility of intelligent virtual agents learning to make acceptable tradeoffs only when a goal cannot be achieved if all user-defined requirements on how to achieve it are met. This chapter proposes a modification of the Q-learning algorithm to achieve these goals. In this way, it is expected to take a step towards achieving goals such as modelling shopping therapy, understanding the preferences of others, understanding whether the shopping habit is becoming a problem, detecting problems with cognitive memory, modelling behaviour specific to different age groups. To make the Q-learning agent find the optimal path to the goal by meeting particular complex criteria, the use of measures model (a model of environmental criteria and/or emotional models), represented as a new memory matrix, is introduced. If the goal cannot be reached by following the pre-set criteria, the learning agent can compromise a given criterion. The agent makes the least possible number of tradeoffs and appropriate compromises only to reach the goal. If the criteria are arranged by their level of importance, then the agent can choose more in number and more acceptable compromises instead of unacceptable ones. The modified algorithm was applied to train three different intelligent learning agents, respectively a shopping cart agent, a gift-shopping agent, and a broker. The tests show improvement in their behavior.