Measuring the Cost of Bothering Users in Intelligent Systems

摘要

Intelligent systems can benefit from interactions with their users, in order to learn about the users' needs, preferences and goals. However, people immediately run into a problem: too many interactions can cause displeasure to users. To satisfy their users, intelligent agents must try to reduce unnecessary interruptions while obtaining the maximum amount of information from users. There can be significant costs associated with such interactions and it is important to find the right balance. This paper discusses the model of bother cost. Then simulation experiments show that the system that uses a bother cost model has more benefit to the user and modeling bother cost is indeed worthwhile. Hence, in the design of agents that automate the process of negotiation between users and the websites that wish to obtain personal information from those users, the agents must be careful not to bother users too much with asking questions.