BKT+IRT: Bayesian Knowledge Tracing augmented with multidimensional generalizable student abilities and problem effects¶

Theory · All theories
Evidence · no claims cited

Description¶

BKT+IRT merges BKT's temporal hidden-state dynamics with IRT-style student ability and problem difficulty effects: student ability and problem-specific offsets modulate BKT's guessing, slipping, learning, and forgetting probabilities via logit offsets. Student abilities are multidimensional (learning, not-forgetting, guessing, not-slipping dimensions), and new students' abilities are inferred by a sequential Bayesian update over discretized ability values or learnable student prototypes. The article states this "merges BKT temporal dynamics with problem and student effects from IRT models", retaining interpretable parameters unlike black-box NN models.

Design Implications¶

Context¶

Requirements¶

• Requires an answer sequence split by KC and V BKT layers (one per ability level) combined via a sequential Bayesian update layer

Constraints¶

• The model assumes new students can be categorized into roughly the same prototypes learned during training, which the authors judge reasonable for large student cohorts

Target Learners¶

• students practicing exercises in intelligent tutoring systems

Target Learning Objectives¶

• prediction of student performance to enable adaptive exercise selection

Claims¶

• Multidimensional Abilities Improve Unidimensional Some Datasets [+M]

Related Theories¶

• Adaptive Learning
• Bayesian Knowledge Tracing: a two-state Hidden Markov Model inferring skill mastery from response histories
• Bayesian Knowledge Tracing: a four-parameter student-learning model in two forms (HMM and Knowledge Tracing Algorithm)
• Bayesian Knowledge Tracing as a model of changing skill mastery during game-based assessment
• Skill discovery BKT: end-to-end stochastic learning of the problem-KC assignment matrix with Gumbel-Softmax

Examples¶

-

Key Sources¶

• Khajah, M. M. (2024). Supercharging BKT with Multidimensional Generalizable IRT and Skill Discovery. Journal of Educational Data Mining, Volume 16, No 1, 2024. https://jedm.educationaldatamining.org/index.php/JEDM/article/view/16-1