Automatic Simulation of Users for Interactive Information Retrieval

Hiring and engaging real users for conducting user studies is costly and time consuming. Therefore, simulating users has been proposed as a resource saving solution. Simulation helps to predict users' performance with retrieval systems and could provide researchers with much needed insights. There is a growing body of research in simulating users but most of the research conducted in the field is based only on data about how real users interact with the Information Retrieval System (IRS) leaving out aspects describing information seeking behaviours and individual differences. Consequently, essential data about users is not accounted thus the models are less realistic. To overcome this issue, an alternative approach is proposed. It concerns looking at user behaviour and analysing all the influencing factors that contribute to the user profile in order to produce representative user models. Thus, the aim of this research is to build real user profiles in order to produce more realistic user simulation. We will achieve that by capturing information seeking behaviour through user studies and selecting those factors and attributes that make our user models more realistic and operational at the same time. Furthermore, the proposed approach will allow to reduce the number of real users needed to evaluate the system.