Four-Phase Taxonomy of Knowledge Tracing Variants

Theory · All theories
Evidence · 7 claims (6 for, 1 mixed) · 3 studies (2 design, 1 review), q1–q2 · 0 of 3 report an effect size · 7 claims rest on one study

Description

The survey's second organizing axis groups extensions of the fundamental models by the learning phase they add: "we classify and review current variants of fundamental KT models into four categories" — individualization before learning, engagement during learning, forgetting after learning, and side information across learning. Examples include student-node BKT and DKT-DSC clustering, KAT and EEG-BKT, BKT-Forget and DKT-forget, and FAST and EKT.

Design Implications

Context

Requirements

• Data beyond exercises and responses: student-specific parameters or clusters, engagement signals (EEG or behavioral indicators such as quick guess and bottom-out hints), time gaps and trial counts, or side information such as response time, tutor intervention, exercise text and code submissions

Constraints

• The survey motivates the variants by stating the fundamental KT models, while straightforward, may have reduced performance in real-world learning scenarios

Target Learners

• Students whose exercise-answering interactions are recorded by online learning systems such as ASSISTments, Junyi Academy, Eedi and EdNet's Santa tutor

Target Learning Objectives

• Monitoring students' evolving knowledge states on knowledge concepts and predicting performance on future exercises

Claims

• Individualized Bkt Reduces Questions Needed For Mastery [+W]
• Engagement Covariates Improve Deep Knowledge Tracing [+W]
• Bkt Overestimates Accuracy After A Day Elapses [+W]
• Language Proficiency Side Information Improves Kt Models [+W]
• Customizing knowledge-node priors from pretest scores improves model accuracy, but tailoring SE-tendency priors yields no significant improvement [~W]
• Standard knowledge tracing models in use do not allow individual learning rates or individualized initial knowledge [+W]

The evidence by question

• Does extending a knowledge-tracing model with individual parameters, engagement signals, forgetting or other side information improve how well it predicts learners' answers? The survey reports, citing Pavlik et al., that no single knowledge tracing model was always the best, and that a better model must consider multiple student features and the learning context. [+W]

Related Theories

• Knowledge Tracing Model Taxonomy
• Engagement
• Intervention-BKT: a BKT extension that models the effect of instructional interventions on student knowledge states

Examples

-

Key Sources

• Shuanghong Shen, Qi Liu, Zhenya Huang, Yonghe Zheng, Minghao Yin, Minjuan Wang, and Enhong Chen. (2021). A Survey of Knowledge Tracing: Models, Variants, and Applications. https://arxiv.org/abs/2105.15106

Page metadata

Field	Value
type	theory
title	Four-Phase Taxonomy of Knowledge Tracing Variants
description	The survey's second organizing axis groups extensions of the fundamental models by the learning phase they add: "we classify and review current variants of fundamental KT models into four categories" — individualizati...
status	review
canonical	True
generated	by claude/unspecified on 2026-09-24
trust	unverified — no one has confirmed this page's content
sources	1 — shuanghong-shen-2021

Frontmatter is the page's machine-readable record — see the Schema & Guide.