Haptic Interfaces have been used as cooperative systems to reproduce and simulate human actions. The Haptic Interface (HI) can be used as a tool capable of interacting dynamically with the operator using Reactive Robot (RR) technology. The RR systems provide to the HI the capability of interpreting the human actions and exert a more intelligent force feedback strategy. This paper deals with an application of the RR theory to teach people to write Japanese characters, using a desktop HI designed at PERCRO. In this article, we will outline the main features of the RR system. Human action identification was implemented using Hidden Markov Models. The recognition system was tested using a Polhemus sensor. An initial control law to restrict the user to move along a predefined trajectory using the HI is described.