Deep Knowledge Tracing

By C. Piech, J. Bassen, Jonathan Huang, Surya Ganguli, Mehran Sahami, Leonidas J. Guibas, Jascha Narain Sohl-Dickstein. 2015

1,613 citations recorded by the source corpus.

Knowledge tracing—where a machine models the knowledge of a student as they interact with coursework—is a well established problem in computer supported education. Though effectively modeling student knowledge would have high educational impact, the task has many inherent challenges. In this paper we explore the utility of using Recurrent Neural Networks (RNNs) to model student learning. The RNN family of models have important advantages over previous methods in that they do not require the explicit encoding of human domain knowledge, and can capture more complex representations of student knowledge. Using neural networks results in substantial improvements in prediction performance on a range of knowledge tracing datasets. Moreover the learned model can be used for intelligent curriculum design and allows straightforward interpretation and discovery of structure in student tasks. These results suggest a promising new line of research for knowledge tracing and an exemplary application task for RNNs.

Subfields: Artificial Intelligence, Computer Science Applications

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Science Applications

Citations in the corpus

Deep Learning for IoT Big Data and Streaming Analytics: A Survey (2017, methodology): “In addition, Piech et al. applied RNN and LSTM networks to model the prediction of educator answers to exercises and quizzes, based on their past activities and interactions in MOOCs [118].”
On the Expressive Power of Deep Neural Networks (2016, background): “Deep neural networks have proved astoundingly effective at a wide range of empirical tasks, from image classiﬁcation (Krizhevsky et al., 2012) to playing Go (Silver et al., 2016), and even modeling human learning (Piech et al., 2015).”
Exponential expressivity in deep neural networks through transient chaos (2016, background): “Deep feedforward neural networks, with multiple hidden layers, have achieved remarkable performance across many domains [1–4].”
Capacity and Trainability in Recurrent Neural Networks (2016, background): “…application domains in deep learning (speech recognition (Amodei et al., 2015), seq2seq (Sutskever et al., 2014), neural machine translation (Bahdanau et al., 2014), the DRAW model (Gregor et al., 2015), educational applications (Piech et al., 2015), and scientiﬁc discovery (Mante et al., 2013)).”
Biologically inspired protection of deep networks from adversarial attacks (2017, background): “Deep Neural Networks (DNNs) have demonstrated success in many machine learning tasks, including image recognition (Krizhevsky et al., 2012), speech recognition (Hinton et al., 2012), and even modelling mathematical learning (Piech et al., 2015), among many other domains.”
Resurrecting the sigmoid in deep learning through dynamical isometry: theory and practice (2017, background): “1 Introduction Deep learning has achieved state-of-the-art performance in many domains, including computer vision [1], machine translation [2], human games [3], education [4], and neurobiological modelling [5, 6].”
Generative Models of Visually Grounded Imagination (2017, background): “We draw inspiration from the ﬁeld of education, which similarly faces the challenge of assessing whether a student has successfully “understood” a concept ( c.f. , (Piech et al., 2015)).”
Enhancing the prediction of student performance based on the machine learning XGBoost algorithm (2021, methodology): “Regarding the technical improvements that were resorted to, we cite the update made by Piech Chris et al. (Piech et al., 2015), Which has shown how it can achieve very high results by simply making some changes to the BKT approach through the adoption of Deep Learning (DL) algorithms.”
Behavior-Based Grade Prediction for MOOCs Via Time Series Neural Networks (2017, methodology): “[2], [17], [18] and what their final grades will be [19], [20], typically using their past assessment scores.”
MathBERT: A Pre-trained Language Model for General NLP Tasks in Mathematics Education (2021, methodology): “Finally, both AUC and ACC are used to measure 𝑇 𝑘𝑡 because historically both metrics were used for evaluation [14, 18, 23, 31].”
Context-Aware Attentive Knowledge Tracing (2020, methodology): “e sophisticated and flexible models to fully exploit the information contained arXiv:2007.12324v1 [cs.LG] 24 Jul 2020 in large-scale learner response datasets. The deep knowledge tracing (DKT) method [23] was the first method to explore the use of (possibly deep) neural networks for KT by using long”
Deep Knowledge Tracing On Programming Exercises (2017, background): “Other models that are designed to take low dimensional inputs, such as IRT and modiﬁcations of Bayesian Knowledge Tracing [13] [7], sometimes outperform the initial version of Deep Knowledge Tracing (DKT) [12] [5].”
Teaching Categories to Human Learners with Visual Explanations (2018, background): “To date, a variety of different approaches have been explored for modeling the teaching of students from assuming perfect learners [12, 40, 21], heuristic-based approaches [2], Bayesian models [7, 11], recurrent neural networks [23], and reinforcement learning based methods [25, 1, 37].”
Graph-based knowledge tracing: Modeling student proficiency using graph neural networks (2019, methodology): “[17] reported that a method called deep knowledge tracing (DKT), which leverages recurrent neural networks (RNNs) [24], performs significantly better than any previous method.”
A Self Attentive model for Knowledge Tracing (2019, background): “The DKT model faces the issue of its parameters being non-interpretable [4].”
Mining Big Data in Education: Affordances and Challenges (2020, methodology): “Three popular methods are Bayesian knowledge tracing (BKT; Corbett & Anderson, 1995), performance factors analysis (PFA; Pavlik et al., 2009), and deep knowledge tracing (DKT; Khajah et al., 2016).”
Addressing two problems in deep knowledge tracing via prediction-consistent regularization (2018, methodology): “rbolic tangent tanh() and the sigmoid function ˙() are applied in an element-wise manner. The model is parameterized by a weight matrix W and a bias vector b with appropriate dimensions. Piech et al. [12] adopts an RNN variant with long short-term memory (LSTM) cells. An LSTM cell incorporates three”