Document Type


Publication Date



Traditional intelligent tutoring systems have been successful at fostering learning, but very few systems have been built due to the high cost of creation. It has been reported that it takes between 100 to 1000 hours to produce a single hour of tutoring content. Our previous research reported a reduction in the cost of authoring content through the use of pseudo-tutors, constructs that mimic cognitive tutors but are limited in scope to a single problem. Although the extreme reduction in complexity allowed teachers with no background in intelligent tutoring systems to build effective tutoring content, building multiple questions within a particular skill set required significant repetition of content. In the current work, we add some complexity back into the system by allowing teachers to generalize pseudo-tutors through the use of variables that can alter the contextual and numerical data used in the problem.We report evidence that variabilization reduces the cost of authoring similar skill problems by a factor of two. Further, this factor increases linearly with the number of instances of the problem created. We also suggest that the additional complexity is not a hindrance to teachers adopting the system and some repetition of tutoring content is acceptable to students.