skip to main content

Title: Knowledge Tracing for Complex Problem Solving: Granular Rank-Based Tensor Factorization
Knowledge Tracing (KT), which aims to model student knowledge level and predict their performance, is one of the most important applications of user modeling. Modern KT approaches model and maintain an up-to-date state of student knowledge over a set of course concepts according to students’ historical performance in attempting the problems. However, KT approaches were designed to model knowledge by observing relatively small problem-solving steps in Intelligent Tutoring Systems. While these approaches were applied successfully to model student knowledge by observing student solutions for simple problems, such as multiple-choice questions, they do not perform well for modeling complex problem solving in students. Most importantly, current models assume that all problem attempts are equally valuable in quantifying current student knowledge. However, for complex problems that involve many concepts at the same time, this assumption is deficient. It results in inaccurate knowledge states and unnecessary fluctuations in estimated student knowledge, especially if students guess the correct answer to a problem that they have not mastered all of its concepts or slip in answering the problem that they have already mastered all of its concepts. In this paper, we argue that not all attempts are equivalently important in discovering students’ knowledge state, and more » some attempts can be summarized together to better represent student performance. We propose a novel student knowledge tracing approach, Granular RAnk based TEnsor factorization (GRATE), that dynamically selects student attempts that can be aggregated while predicting students’ performance in problems and discovering the concepts presented in them. Our experiments on three real-world datasets demonstrate the improved performance of GRATE, compared to the state-of-the-art baselines, in the task of student performance prediction. Our further analysis shows that attempt aggregation eliminates the unnecessary fluctuations from students’ discovered knowledge states and helps in discovering complex latent concepts in the problems. « less
; ; ; ;
Award ID(s):
Publication Date:
Journal Name:
Proceedings of the 29th ACM Conference on User Modeling, Adaptation and Personalization
Page Range or eLocation-ID:
179 to 188
Sponsoring Org:
National Science Foundation
More Like this
  1. The state of the art knowledge tracing approaches mostly model student knowledge using their performance in assessed learning resource types, such as quizzes, assignments, and exercises, and ignore the non-assessed learning resources. However, many student activities are non-assessed, such as watching video lectures, participating in a discussion forum, and reading a section of a textbook, all of which potentially contributing to the students' knowledge growth. In this paper, we propose the  first novel deep learning based knowledge tracing model (DMKT) that explicitly model student's knowledge transitions over both assessed and non-assessed learning activities. With DMKT we can discover the underlying latent concepts of each non-assessed and assessed learning material and better predict the student performance in future assessed learning resources. We compare our proposed method with various state of the art knowledge tracing methods on four real-world datasets and show its effectiveness in predicting student performance, representing student knowledge, and discovering the underlying domain model.
  2. Introduction: The work reported here subscribes to the idea that the best way to learn - and thus, improve student educational outcomes - is through solving problems, yet recognizes that engineering students are generally provided insufficient opportunities to engage problems as they will be engaged in practice. Attempts to incorporate more open-ended, ill-structured experiences have increased but are challenging for faculty to implement because there are no systematic methods or approaches that support the educator in designing these learning experiences. Instead, faculty often start from the anchor of domain-specific concepts, an anchoring that is further reinforced by available textbook problems that are rarely open in nature. Open-ended problems are then created in ad-hoc ways, and in doing so, the problem-solving experience is often not realized as the instructor intended. Approach: The focus in this work is the development and preliminary implementation of a reflective approach to support instructors in examining the design intent of problem experiences. The reflective method combines concept mapping as developed by Joseph Novak with the work of David Jonassen and his characterization of problems and the forms of knowledge required to solve them. Results: We report on the development of a standard approach – a templatemore »-- for concept mapping of problems. As a demonstration, we applied the approach to a relatively simple, well-structured problem used in an introductory aerospace engineering course. Educator-created concept maps provided a visual medium for examining the connectivity of problem elements and forms of knowledge. Educator reflection after looking at and discussing the concept map revealed ways in which the problem engagement may differ from the perceived design intent. Implications: We consider the potential for the proposed method to support design and facilitation activities in problem-based learning (PBL) environments. We explore broader implications of the approach as it relates to 1) facilitating a priori faculty insights regarding student navigation of problem solving, 2) instructor reflection on problem design and facilitation, and 3) supporting problem design and facilitation. Additionally, we highlight important issues to be further investigated toward quantifying the value and limitations of the proposed approach.« less
  3. Practice plays a critical role in learning engineering dynamics. Typical practice in a dynamics course involves solving textbook problems. These problems can impose great cognitive load on underprepared students because they have not mastered constituent knowledge and skills required for solving whole problems. For these students, learning can be improved by being engaged in deliberate practice. Deliberate practice refers to a type of practice aimed at improving specific constituent knowledge or skills. Compared to solving whole problems requiring the simultaneous use of multiple constituent skills, deliberate practice is usually focused on one component skill at a time, which results in less cognitive load and more specificity. Contemporary theories of expertise development have highlighted the influence of deliberate practice (DP) on achieving exceptional performance in sports, music, and various professional fields. Concurrently, there is an emerging method for improving learning efficiency of novices by combining deliberate practice with cognitive load theory (CLT), a cognitive-architecture-based theory for instructional design. Mechanics is a foundation for most branches of engineering. It serves to develop problem-solving skills and consolidate understanding of other subjects, such as applied mathematics and physics. Mechanics has been a challenging subject. Students need to understand governing principles to gain conceptual knowledgemore »and acquire procedural knowledge to apply these principles to solve problems. Due to the difficulty in developing conceptual and procedural knowledge, mechanics courses are among those that receive high DFW rates (percentage of students receiving a grade of D or F or Withdrawing from a course), and students are more likely to leave engineering after taking mechanics courses. Deliberate practice can help novices develop good representations of the knowledge needed to produce superior problem solving performance. The goal of the present study is to develop deliberate practice techniques to improve learning effectiveness and to reduce cognitive load. Our pilot study results revealed that the student mental effort scores were negatively correlated with their knowledge test scores with r = -.29 (p < .05) after using deliberate practice strategies. This supports the claim that deliberate practice can improve student learning while reducing cognitive load. In addition, the higher the students’ knowledge test scores, the lower their mental effort was when taking the tests. In other words, the students who used deliberate practice strategies had better learning results with less cognitive load. To design deliberate practice, we often need to analyze students’ persistent problems caused by faulty mental models, also referred to as an intuitive mental model, and misconceptions. In this study, we continue to conduct an in-depth diagnostic process to identify students’ common mistakes and associated intuitive mental models. We then use the results to develop deliberate practice problems aimed at changing students’ cognitive strategies and mental models.« less
  4. Students acquire knowledge as they interact with a variety of learning materials, such as video lectures, problems, and discussions. Modeling student knowledge at each point during their learning period and understanding the contribution of each learning material to student knowledge are essential for detecting students’ knowledge gaps and recommending learning materials to them. Current student knowledge modeling techniques mostly rely on one type of learning material, mainly problems, to model student knowledge growth. These approaches ignore the fact that students also learn from other types of material. In this paper, we propose a student knowledge model that can capture knowledge growth as a result of learning from a diverse set of learning resource types while unveiling the association between the learning materials of different types. Our multi-view knowledge model (MVKM) incorporates a flexible knowledge increase objective on top of a multi-view tensor factorization to capture occasional forgetting while representing student knowledge and learning material concepts in a lower-dimensional latent space. We evaluate our model in different experiments to show that it can accurately predict students’ future performance, differentiate between knowledge gain in different student groups and concepts, and unveil hidden similarities across learning materials of different types.
  5. The knowledge tracing (KT) task consists of predicting students’ future performance on instructional activities given their past performance. Recently, deep learning models used to solve this task yielded relative excellent prediction results relative to prior approaches. Despite this success, the majority of these models ignore relevant information that can be used to enhance the knowledge tracing performance. To overcome these limitations, we propose a generic framework that also accounts for the engagement level of students, the difficulty level of the instructional activities, and the natural language processing embeddings of the text of each concept. Furthermore, to capture the fact that students’ knowledge states evolve over time we employ a LSTM-based model. Then, we pass such sequences of knowledge states to a Temporal Convolutional Network to predict future performance. Several empirical experiments have been conducted to evaluate the effectiveness of our proposed framework for KT using Cognitive Tutor datasets. Experimental results showed the superior performance of our proposed model over many existing deep KT models. And AUC of 96.57% has been achieved on the Algebra 2006-2007 dataset.