Data Driven Knowledge Tracing to Improve Learner Outcomes

Course Description: To see learning, we need to look at a detailed, fine-grained decomposition of knowledge to be learned (called Knowledge Components). We will look into how we can infer an individual student’s knowledge of the KCs in the cognitive model, based on performance. Then individualize the curriculum such that each student can master all KCs at their own pace.

In this course, students will learn to apply data-driven modeling to track students’ knowledge growth in a learning technology environment and adapt the instructions based on that. They will also be able to apply different techniques to figure out the best KC model for a task via search.

Targeted Audience: Researchers, Data Scientists

Upon completing the course you will be able to:

  • Describe how data-driven modeling can help individualize learning
  • Implement Item Response Theory, Bayesian Knowledge Tracing, and n-correct model
  • Select the best model for your use based on model attributes (e.g. accuracy, model size, introspectibility)
  • Predict student performance using BKT
  • Apply data-driven tuning of cognitive mastery
  • Explain knowledge component modeling and ways to implement a KC model
  • Perform data-driven KC model revision
  • Evaluate a KC model with Additive Factors Model (AFM)
  • Explain the role of cognitive task analysis in scaling instructional impact
  • Discover a cognitive model via search and perform a system redesign in light of a new cognitive model
  • Interpret and generalize the cognitive process

Duration: two weeks, approximately 6 hours/week
Prerequisite: E-Learning Cognitive Task Analysis