My
research topic is,"Enriching Student Model in an Intelligent Tutoring
System". To understand this research topic, I will first explain what
is Intelligent Tutoring System (ITS) and then what system (ITS) I used
in my research, what is student model and how we have enriched (made it
better) the student model.
ITS is a computer-based self-learning system in which students learn a
specific topic, for example, how to make a kite, how to run cessna
aircraft, how to code in Java etc. An Intelligent Tutoring System (ITS)
provides personalized learning content to students based on their needs
and preferences. An ITS consists of learning content, student model
and adaptation engine. Student models are constructed from the log
files available in the ITS. Students’ interaction with ITS, such as
responses to questions, number of attempts at a task, and the time taken
for various activities (such as responding or reading) are captured in
the ITS log file. Student models also typically contain information
such as student’s previous knowledge and background from which it is
possible to infer student’s cognitive states.
Mindspark: a commercial mathematics ITS developed by Educational Initiatives India, is used in our research to implement and test our theory-driven approach. Mindspark has been incorporated into the school curriculum for different age groups (grades 3 to 8) of students. Mindspark is currently implemented in sixty schools and being used by 30,000 students on an average of four sessions per week with each session ranging from 25 to 30 minutes long. In Mindspark, students learn mathematics by answering questions posed by the system. Mindspark provides detailed feedback and explanation upon receiving the answer from the student. Mindspark consists of a sequence of specially designed learning units (clusters), which contain questions on concepts that make up the topic. Each topic consists of questions of progressively increasing complexity levels. Mindspark cover a wide range of topics in school-level mathematics such as linear inequalities, matrices, quadratic equations, fractions, decimals, and polygons. Mindspark adaptation logic selects the questions to be presented based on a student’s answer to a previous question and his/ her overall performance in the topic which allows the student to move at his/her own pace. If a student performs poorly in the current topic (for example, the student does not answer sufficient number of questions correctly), he/she will be moved to a lower complexity level in the same topic. In Mindspark, if a student answers three consecutive questions correctly, he/she receives a Sparkie (extra motivational points). If a student answers five consecutive questions correctly, then he/she receives a Challenge Question (tougher question from higher complexity level). If the student answers the Challenge Question correctly, he/she receives five Sparkies. Every week, the highest Sparkie collectors (Sparkie Champ) are identified and their names are published on the Mindspark website.
My
research is on how to enrich the student model by adding effective
states to the student model and it is briefed in following steps. In
our research, we consider frustration as a cognitive emotion to be
predicted when the students interact with Mindspark and mitigate it when
identified.
Phase 1 Operationalize the theoretical definitions of frustration for Mindspark Log data and develop a model to predict frustration of the students when they interact with Mindspark. The model is developed and tested using human observation. We have collected 11 students' facial expressions when they interact with Mindspark. A linear regression model is developed using the features constructed from log data and trained the model using observed facial expression and then tested using 10-fold cross validation method. Status: Completed.
Phase 2 Find the existing data mining approaches used to predict frustration in ITS and apply all the methods on Mindspark log data. Compare the results with our theory-driven approach. We collected 16 more students’ facial expressions and compared our approach with existing approaches. Our approach gave comparable results with existing approaches and the causes of frustration is clearly shown in our approach. Status: Completed.
Phase 3 At once the frustration is predicted in Mindspark, we pop-up a motivational message to motivate students to continue the session instead of getting more frustrated. We created messages to motivate, to attribute the result to external factor, to praise the student's effort and to show sympathy for their failure. We have developed set of algorithms to select messages based on causes of frustration. Currently we are integrating the theory-driven approach, which we developed in Phase 2, in Mindspark code to show the motivational messages to student whenever our model predicts that they are frustrated. Status: In progress.