Intelligent environments in educational settings are aimed at supporting the learning process with an unobtrusive monitoring of the student while doing his/her activities. A desk is a common object in these settings, so if it is enhanced with sensing capabilities, it would enable gathering information of user-object interaction in a natural and unobtrusive way. An intelligent system is needed to analyze that information to reach conclusions faster than with traditional, manual observational process, and to provide timely valuable information. In this article, a design of a system to semi-automatically identify relationships between behaviors and the task performance of learners from user-object interaction logs is presented. The aim of detecting such relationships is to help teachers and students in the learning process, supporting their activities to identify special needs. Its components are designed to address four main functions: data acquisition, behavior identification, student performance identification, and computing relationships between student task performance and behaviors.