Abstract:Bayesian Knowledge Tracing (BKT) is a probabilistic model of a learner's state of mastery corresponding to a knowledge component. It considers the learner's state of mastery as a "hidden" or latent binary variable and updates this state based on the observed correctness of the learner's response using parameters that represent transition probabilities between states. BKT is often represented as a Hidden Markov Model and the Expectation-Maximization (EM) algorithm is used to infer these parameters. However, this algorithm can suffer from several issues including producing multiple viable sets of parameters, settling into a local minima, producing degenerate parameter values, and a high computational cost during fitting. This paper takes a "from first principles" approach to deriving constraints that can be imposed on the BKT parameter space. Starting from the basic mathematical truths of probability and building up to the behaviors expected of the BKT parameters in real systems, this paper presents a mathematical derivation that results in succinct constraints that can be imposed on the BKT parameter space. Since these constraints are necessary conditions, they can be applied prior to fitting in order to reduce computational cost and the likelihood of issues that can emerge from the EM procedure. In order to see that promise through, the paper further introduces a novel algorithm for estimating BKT parameters subject to the newly defined constraints. While the issue of degenerate parameter values has been reported previously, this paper is the first, to our best knowledge, to derive the constrains from first principles while also presenting an algorithm that respects those constraints.

What problem does this paper attempt to address?

The problems that this paper attempts to solve are several issues encountered in the parameter estimation process of the Bayesian Knowledge Tracing (BKT) model, especially: 1. **Multi - solution problem**: When using the Expectation - Maximization (EM) algorithm for parameter estimation, multiple different parameter sets may be generated. Although these parameter sets can all fit the data well, their interpretability is different. 2. **Degenerate solution problem**: The EM algorithm sometimes converges to unreasonable parameter values, that is, degenerate solutions. Although these parameters can fit the data, they will lead to wrong decisions in practical applications. 3. **High computational cost**: If the EM algorithm converges to a degenerate solution, the algorithm needs to be re - run, which will increase the computational cost. To solve these problems, the paper proposes a method starting from first - principles and derives the constraint conditions that can be imposed on the BKT parameter space. These constraint conditions can be applied before parameter estimation to reduce the computational cost and avoid the possibility of the EM algorithm converging to a degenerate solution. In addition, the paper also introduces a new EM algorithm, which uses the Interior - Point Method to ensure that the parameter estimation satisfies these constraint conditions. ### Specific content summary #### 1. Introduction to the Bayesian Knowledge Tracing model The BKT model is used to describe the change in the learner's mastery state of a knowledge point (Knowledge Component, KC). It assumes that the learner's mastery state is a hidden binary variable (Mastery or Non - Mastery) and updates this state through four parameters: - Initial mastery probability \(P(L_0)\) - Transition probability from non - mastery to mastery \(P(R)\) - Probability of guessing the correct answer \(P(G)\) - Probability of making mistakes in the mastery state \(P(S)\) #### 2. Derivation of parameter constraints Starting from the basic mathematical principles of probability, the paper derives the following constraint conditions: - \(0 < P(G) < 1\) - \(0 < P(S) < 1\) - \(0 < P(R) < 1\) - \(1 - P(S) - P(G)\geq0\) - \(\frac{(1 - P(G))\cdot P(R)}{1 - P(S) - P(G)} < P(L_0) < 1\) These constraint conditions ensure the rationality and non - degeneracy of the parameters. #### 3. New EM algorithm To ensure that the parameter estimation satisfies the above - mentioned constraint conditions, the paper proposes a new EM algorithm, which combines the Interior - Point Method. By introducing a barrier function and using the Newton method for iterative solution, the parameter estimation finally converges to values that satisfy the constraint conditions. #### 4. Experimental verification The paper verifies the effectiveness of the new method through simulated data. The results show that, compared with the traditional Baum - Welch algorithm, the new EM - Interior - Point method can converge to reasonable parameter values in more cases and avoid degenerate solutions. ### Conclusion The paper successfully derives the constraint conditions of the BKT model parameters and proposes a new EM algorithm, which can ensure that the parameter estimation satisfies these constraint conditions, thereby improving the stability and reliability of the model. Future work will focus on how to use this algorithm to identify poorly - performing knowledge points and provide improvement suggestions for learning designers.

Parametric Constraints for Bayesian Knowledge Tracing from First Principles

An Introduction to Bayesian Knowledge Tracing with pyBKT

pyBKT: An Accessible Python Library of Bayesian Knowledge Tracing Models

Equity and Fairness of Bayesian Knowledge Tracing

How deep is knowledge tracing?

Bayesian Optimisation for Constrained Problems

Uncertainty-preserving deep knowledge tracing with state-space models

Learning physical parameters from dynamic scenes

Extending Deep Knowledge Tracing: Inferring Interpretable Knowledge and Predicting Post-System Performance

Parameter Estimation for Hidden Markov Models with Intractable Likelihoods

A Deeper Knowledge Tracking Model Integrating Cognitive Theory and Learning Behavior

Time-and-Concept Enhanced Deep Multidimensional Item Response Theory for interpretable Knowledge Tracing

Towards Practical Bayesian Parameter and State Estimation.

Learning Bayesian Network Parameters from Limited Data by Integrating Entropy and Monotonicity

Marginally Specified Priors for Nonparametric Bayesian Estimation

Interpretable Knowledge Tracing: Simple and Efficient Student Modeling with Causal Relations

A Kernel Approach to Tractable Bayesian Nonparametrics

Efficient Algorithms for Bayesian Network Parameter Learning from Incomplete Data

A probabilistic generative model for tracking multi-knowledge concept mastery probability

Counterfactual Monotonic Knowledge Tracing for Assessing Students' Dynamic Mastery of Knowledge Concepts

Properties and Bayesian fitting of restricted Boltzmann machines