A Virtual Assistant for Web-Based Training In Engineering Education Frédéric Geoffroy (1) , Esma Aimeur (2) , and Denis Gillet (1) (1) Swiss Federal Institute of Technology in Lausanne (EPFL) LA-I2S-STI, CH – 1015 Lausanne, Switzerland Phone: +41 21 693-5168, FAX: + 41 21 693-2574, E-mail: d, frederic.geoffroy@epfl.ch, denis.gillet@epfl.ch (2) Université de Montréal, Département IRO C.P. 6128, Succursale Centre-Ville, Montréal (Québec), H3C 3J7 Canada aimeur@IRO.UMontreal.CA Abstract. Experimentation has always been an essential ingredient to sustain the learning activities in engineering education. During traditional laboratory sessions, a huge amount of work is carried out by the assistant who is in charge of supporting and evaluating the students. In a Web-based experimentation setting students ask for more feedback while they work on simulation or remote manipulation. We present in this paper a virtual assistant for Web-based training. The training and the evaluation process are shared between real and virtual assistants in order to deliver a tutoring scheme adapted to Web-based experimentation. 1 Introduction The Swiss Federal Institute of Technology in Lausanne (EPFL) currently supports various new learning technologies projects for promoting active and flexible learning in engineering education. The eMersion project [8] is an initiative integrated in this framework with the main objective of sustaining hands-on practice and active learning through Web-based experimentation. The Web-based experimentation environment implemented at the EPFL features Web-based simulation and remote manipulation facilities. Web-based education is getting an increasing popularity due to its clear benefits: Classroom and platform independence. We know only four Web-based educational systems that have influenced a number of more recent systems, among which ELM- ART [4] and InterBook [3]. If we consider simulation, several Web-based systems exist such as Cardiac Tutor [6], Belvedere [10], and Simquest [9]. The purpose of these systems and the methods used are numerous and varied. For example, [11] use induction to generate feedback in simulation-based discovery learning.