Enriching the Student Model in an Intelligent Tutoring System

The ability to identify emotions had been a uniquely human trait, till artificial intelligence equipped machines with the same trait, though not at an equal level of complexity. Identification of emotions using human interaction with the computer, and addressing cognitive needs based on those emotions has a wide range of applications today.

Ramkumar Rajendran, a Research Scholar at the IITB-Monash Research Academy in Mumbai, is currently researching the development of a mathematical model for Intelligent Tutoring Systems, which predicts and addresses the emotions of students, such as frustration, while they interact with the system. An Intelligent Tutoring System (ITS) adapts the learning content based on a student’s performance, previous knowledge and background.

The cognitive process in computer-based learning is influenced by the learner’s affective state to a large extent. For example, frustration with a learning system may cause a learner, in this case a student, simply to give up on the learning process. If these affective states can be identified and predicted by the ITS, the role of affective states like frustration in the learning process can be minimized to a large extent. Ramkumar Rajendran’s research proposes a model to predict and address in real time, an affective state, frustration, based on data from the student’s interaction with the computer. This is done using theoretical definitions which allow the causes of frustration to be identified.

Predicting the students’ affective states during their interaction with the system is a challenging problem in the area of education research, and is the focus of several current research efforts. The methods that have been implemented in the ITS to predict the affective state have thus far included human observation, the learners’ self-reported data of their affective state, mining the system’s log data, face-based emotion recognition systems, analyzing the data from physical sensors (such as a posture analysis seat, and a pressure sensitive mouse), and more recently, analyzing the data from physiological sensors such as electrocardiogram (ECG), electromyography (EMG), and galvanic skin response (GSR). While advances in the above methods are promising in the lab setting, these methods, except data-mining approaches, are not yet feasible in a large scale real-world scenario to predict affective states. Hence data-mining approaches using the data from the log file generated by the system as a record, are feasible for large scale implementation of commercial ITS.

The model developed by Ramkumar Rajendran for predicting frustration is based on the information available in the log file of Mindspark, a math ITS. The feature selection is based on theoretical definitions, and therefore, an informed adaptation is possible. The frustration model is tested by comparing predicted affective states with human observation while the students interact with Mindspark. He enumerates the benefits:

  • It predicts the student’s frustration and addresses it while the students interact with the system, thus increasing the learning time spent with the system.
  • The student’s frustration is addressed by showing motivational messages

In facilitator-led learning scenarios, teachers often check the students’ emotion from their facial expressions in the class, based on which, teaching strategies are adapted. For example, when a teacher perceives that students in the class are bored, he/she may begin to interact with the students using alternative ways, to engage their attention. This identification of emotion, and the reaction to it, are not available in commercial ITS. This, Ramkumar Rajendran says, is what motivates him to research on affective computing—giving emotions to the computer.

Rajendran reflects on his work at IITB-Monash Research Academy, saying, “Our work has been tested only on students in the age group of 10-12. Hence it cannot be generalized to all age groups, without further studies. Our research is preliminary, and a first step in real-time emotion identification in commercial ITS. However, I personally believe that, in a decade or less, the way of interacting with computer-based technology will be entirely different, from how we interact with systems now”.

Established in 2008, IITB-Monash Research Academy is an important collaboration between Australia and India. It offers graduate research scholars the opportunity to study for a dually-badged PhD from both IIT Bombay in India and Monash University in Australia, spending time in both countries over the course of their research.

Research scholar: Ramkumar Rajendran, IITB-Monash Research Academy

Project title: Enriching Student Model in an Intelligent Tutoring System

Supervisors: Sridhar Iyer (IITB), Sahana Murthy (IITB), Campbell Wilson (Monash), Judy Sheard (Monash)

Contact details: ramkumar.r@iitb.ac.in

For more information and details on this technology, email research@
The above story was written by Ms. Sheba Sanjay based on inputs from the research student and IITB-Monash Research Academy.