Data-Driven Feedback Beyond Next-Step Hints

Intelligent tutors have been shown to be as e↵ective as hu-man tutors in supporting learning in many domains. Al-though they can be very e↵ective, the construction of intel-ligent tutors can be costly. One way to address this problem is to use previously collected data to generate domain models to provide intelligent feedback to otherwise non-personalized tutors. These data-driven methods for providing next-step hints have been successful in providing feedback to students in procedural problem solving tutors. We seek to expand on next-step hints with other data-driven methods. We outline three di↵erent interventions, all of which can be generated using previously collected student data. 1