Modeling user preferences in the design of decision-making agents

这项研究的重点是设计代表用户做出决策的智能代理，例如在购买、销售和交易商品、服务和信息的电子市场中。这需要有能力对用户的偏好进行建模，在面对不完全信息时做出决定，并与其他竞争代理进行谈判。 如果用户已经对自动化系统建立了足够的信任，这样的代理不仅可以帮助用户代表用户做出某些决定，而且它还可以是帮助用户自己更好地了解问题和可用的选择、向自己提出问题以及确保她自己的信念和偏好是一致的有价值的工具。随着用户与信息过载的问题作斗争，预计这种自动化代理的重要性将继续增加。 这项工作涉及为智能代理开发改进的技术，以用于代表用户作出决策，并使其行为适应每个用户的偏好和目标。这个研究项目将包括几个子项目，包括偏好启发中的有效查询选择工作，未知偏好预测工作和自动谈判中的对手建模工作。

This research focuses on the design of intelligent agents that make decisions on a user's behalf, such as in electronic markets where goods, services and information are purchased, sold and traded. This requires an ability to model a user's preferences, to make decisions when faced with incomplete information, and to negotiate with other competing agents. Not only can such an agent aid in making certain decisions on behalf of a user, provided that the user has developed sufficient trust in the automated system, but it can also be a valuable tool for helping the user herself to gain a better understanding of problems and of the choices that are available, to ask herself questions, and to ensure that her own beliefs and preferences are consistent. As users struggle with the problem of information overload, it is anticipated that such automated agents will continue to gain in importance. This work involves the development of improved techniques for intelligent agents to use in making decisions on behalf of users and in adapting their behaviour to the preferences and goals of each individual user. This research project will include several subprojects, including work on effective query selection in preference elicitation, on predicting unknown preferences and on opponent modeling in automated negotiation.