In a knowledge tracing task, there is a sequence of interactions for each student: $$[I_1, \cdots, I_T]$$, and each of these interactions consists of several features: $$\{f_t^1, \cdots, f_t^n\}$$. In knowledge tracing, the goal is to predict feature $$f$$ at time $$t$$, which is formulated as: $$p(f_t^k | \{ I_1, \cdots, I_{t-1}, I_t, I_{t+1}, \cdots, I_T \} \setminus \Gamma(f_t^k))$$ where $$\Gamma(f)$$ is a set of features that should be masked. While knowledge tracing typically predicts the correctness of a response, assessment modeling aims to predict the assessment given the sequence of interactions: $$p(a_t^k | \{ I_1, \cdots, I_{t-1}, I_t, I_{t+1}, \cdots, I_T \} \setminus \Gamma(a_t^k))$$

Formal Definition of Assessment Modeling

Assessment modeling frames prediction over student interaction sequences as a pre-training task. The model predicts an assessment feature after masking information that should not be visible for that prediction.

Assessment Modeling as a Pre-training Task

Assessment modeling is a class of pre-training tasks for interactive educational systems. It predicts a masked assessment feature from the remaining interactions in a student's interaction sequence. The approach described in Assessment Modeling: Fundamental Pre-training Tasks for Interactive Educational Systems uses a deep bidirectional transformer encoder.

Google

San Diego State University

1. Introduction
2. Related Works
3. Assessment Modeling
4. Experiments

Assessment Modeling: Fundamental Pre-training Tasks for Interactive Educational Systems

The main goal of interactive educational systems is to design optimized, personalized learning paths for students. A primary limitation of these systems is the lack of a final outcome, such as a final exam grade, in student interaction datasets. To address this issue, assessment modeling is introduced as a class of pre-training tasks. In this approach, a deep bidirectional transformer encoder is utilized as the assessment model to evaluate and predict performance.

Introduction (Assessment Modeling: Fundamental Pre-training Tasks for Interactive Educational Systems)

The related works section of the paper *Assessment Modeling: Fundamental Pre-training Tasks for Interactive Educational Systems* focuses on two main research areas:

1. **Artificial Intelligence in Education (AIEd):** AIEd applies artificial intelligence and deep learning (DL) to improve educational outcomes. Key applications include knowledge tracing, question analysis, and student grade prediction.

2. **Pre-training Methods in Education:** While pre-training is widely used in natural language processing (NLP), computer vision, and speech recognition, its application in AIEd often focuses on predicting student exam scores (e.g., the SAT) or graduation likelihoods. Early educational pre-training methods were primarily NLP tasks where data was produced from learning materials. Notable prior models include the Test-aware Attention-based Convolutional Neural Network (TACNN), which predicts exam question difficulty, and QuesNet, an embedding model trained on question data that demonstrated superior performance over previous models.

Learn Before

Related

Learn After