Initialize BKT parameters from simulated student data when no human data is available¶

Strategy · All strategies
Evidence · no claims cited

Description¶

The article recommends that "researchers might use data generated by simulated students to initialize knowledge tracing parameters when human data is not available." This avoids hand-picked defaults or costly pilot studies: simulated Random-condition logs are fit with BKT to obtain initial parameter estimates. The approach is supported by a positive correlation of around 0.65 between simulated and human learning-rate parameters, though Guess and Slip showed no notable correlation.

Design Implications¶

Context¶

Requirements¶

• Requires simulated student logs from a computational learning model performing the same task, and a BKT fitting pipeline (here the LearnSphere BKT module)

Constraints¶

• The "Known" parameter was near 0 for all skills in simulated data because agents start without prior knowledge, so it will vary in human contexts; the authors state more research is needed to explore and generalize this idea

Target Learners¶

• K-12 students in intelligent tutoring systems

Target Learning Goals¶

• Accurate knowledge tracing parameter estimation before human data collection

Related Strategies¶

• Fit the combined parameter A and fix P(G) or P(L0) externally as an alternative to Dirichlet priors when fitting the BKT HMM
• Match the input-parameter estimation method (managerial estimates vs. time study) to the accuracy the simulation purpose requires
• Use posterior credible intervals on condition-specific BKT parameter contrasts to compare experimental conditions

Examples¶

-

Key Sources¶

• Qiao Zhang and Christopher MacLellan “Going Online: A simulated student approach for evaluating knowledge tracing in the context of mastery learning”. 2021. In: Proceedings of The 14th International Conference on Educational Data Mining (EDM21). International Educational Data Mining Society, 331-337. https://educationaldatamining.org/edm2021/