skip to main content


Title: How Does Augmented Observation Facilitate Multimodal Representational Thinking? Applying Deep Learning to Decode Complex Student Construct
In this paper, we demonstrate how machine learning could be used to quickly assess a student’s multimodal representational thinking. Multimodal representational thinking is the complex construct that encodes how students form conceptual, perceptual, graphical, or mathematical symbols in their mind. The augmented reality (AR) technology is adopted to diversify student’s representations. The AR technology utilized a low-cost, high-resolution thermal camera attached to a smartphone which allows students to explore the unseen world of thermodynamics. Ninth-grade students (N= 314) engaged in a prediction–observation–explanation (POE) inquiry cycle scaffolded to leverage the augmented observation provided by the aforementioned device. The objective is to investigate how machine learning could expedite the automated assessment of multimodal representational thinking of heat energy. Two automated text classification methods were adopted to decode different mental representations students used to explain their haptic perception, thermal imaging, and graph data collected in the lab. Since current automated assessment in science education rarely considers multilabel classification, we resorted to the help of the state-of-the-art deep learning technique—bidirectional encoder representations from transformers (BERT). The BERT model classified open-ended responses into appropriate categories with higher precision than the traditional machine learning method. The satisfactory accuracy of deep learning in assigning multiple labels is revolutionary in processing qualitative data. The complex student construct, such as multimodal representational thinking, is rarely mutually exclusive. The study avails a convenient technique to analyze qualitative data that does not satisfy the mutual-exclusiveness assumption. Implications and future studies are discussed.  more » « less
Award ID(s):
1712676 1626228
NSF-PAR ID:
10192388
Author(s) / Creator(s):
; ; ; ; ; ;
Date Published:
Journal Name:
Journal of Science Education and Technology
ISSN:
1059-0145
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. Abstract

    Tracking students’ learning states to provide tailored learner support is a critical element of an adaptive learning system. This study explores how an automatic assessment is capable of tracking learners’ cognitive and emotional states during virtual reality (VR)‐based representational‐flexibility training. This VR‐based training program aims to promote the flexibility of adolescents with autism spectrum disorder (ASD) in interpreting, selecting and creating multimodal representations during STEM‐related design problem solving. For the automatic assessment, we used both natural language processing (NLP) and machine‐learning techniques to develop a multi‐label classification model. We then trained the model with the data from a total of audio‐ and video‐recorded 66 training sessions of four adolescents with ASD. To validate the model, we implemented both k‐fold cross‐validations and the manual evaluations by expert reviewers. The study finding suggests the feasibility of implementing the NLP and machine‐learning driven automatic assessment to track and assess the cognitive and emotional states of individuals with ASD during VR‐based flexibility training. The study finding also denotes the importance and viability of providing adaptive supports to maintain learners’ cognitive and affective engagement in a highly interactive digital learning environment.

     
    more » « less
  2. Wang, N. (Ed.)
    In education, intelligent learning environments allow students to choose how to tackle open-ended tasks while monitoring performance and behavior, allowing for the creation of adaptive support to help students overcome challenges. Timely feedback is critical to aid students’ progression toward learning and improved problem-solving. Feedback on text-based student responses can be delayed when teachers are overloaded with work. Automated evaluation can provide quick student feedback while easing the manual evaluation burden for teachers in areas with a high teacher-to-student ratio. Current methods of evaluating student essay responses to questions have included transformer-based natural language processing models with varying degrees of success. One main challenge in training these models is the scarcity of data for student-generated data. Larger volumes of training data are needed to create models that perform at a sufficient level of accuracy. Some studies have vast data, but large quantities are difficult to obtain when educational studies involve student-generated text. To overcome this data scarcity issue, text augmentation techniques have been employed to balance and expand the data set so that models can be trained with higher accuracy, leading to more reliable evaluation and categorization of student answers to aid teachers in the student’s learning progression. This paper examines the text-generating AI model, GPT-3.5, to determine if prompt-based text-generation methods are viable for generating additional text to supplement small sets of student responses for machine learning model training. We augmented student responses across two domains using GPT-3.5 completions and used that data to train a multilingual BERT model. Our results show that text generation can improve model performance on small data sets over simple self-augmentation. 
    more » « less
  3. Often, the first step in managing bug reports is related to triaging a bug to the appropriate developer who is best suited to understand, localize, and fix the target bug. Additionally, assigning a given bug to a particular part of a software project can help to expedite the fixing process. However, despite the importance of these activities, they are quite challenging, where days can be spent on the manual triaging process. Past studies have attempted to leverage the limited textual data of bug reports to train text classification models that automate this process -- to varying degrees of success. However, the textual representations and machine learning models used in prior work are limited by their expressiveness, often failing to capture nuanced textual patterns that might otherwise aid in the triaging process. Recently, large, transformer-based, pre-trained neural text representation techniques such as BERT have achieved greater performance in several natural language processing tasks. However, the potential for using these techniques to improve upon prior approaches for automated bug triaging is not well studied or understood. Therefore, in this paper we offer one of the first investigations that fine-tunes transformer-based language models for the task of bug triaging on four open source datasets, spanning a collective 53 years of development history with over 400 developers and over 150 software project components. Our study includes both a quantitative and qualitative analysis of effectiveness. Our findings illustrate that DeBERTa is the most effective technique across the triaging tasks of developer and component assignment, and the measured performance delta is statistically significant compared to other techniques. However, through our qualitative analysis, we also observe that each technique possesses unique abilities best suited to certain types of bug reports. 
    more » « less
  4. Modern 3D printing technology makes it relatively easy and affordable to produce physical models that offer learners concrete representations of otherwise abstract concepts and representations. We hypothesize that integrating hands-on learning with these models into traditionally lecture-dominant courses may help learners develop representational competence, the ability to interpret, switch between, and appropriately use multiple representations of a concept as appropriate for learning, communication and analysis. This approach also offers potential to mitigate difficulties that learners with lower spatial abilities may encounter in STEM courses. Spatial thinking connects to representational competence in that internal mental representations (i.e. visualizations) facilitate work using multiple external representations. A growing body of research indicates well-developed spatial skills are important to student success in many STEM majors, and that students can improve these skills through targeted training. This NSF-IUSE exploration and design project began in fall 2018 and features cross-disciplinary collaboration between engineering, math, and psychology faculty to develop learning activities with 3D-printed models, build the theoretical basis for how they support learning, and assess their effectiveness in the classroom. We are exploring how such models can support learners’ development of conceptual understanding and representational competence in calculus and engineering statics. We are also exploring how to leverage the model-based activities to embed spatial skills training into these courses. The project is addressing these questions through parallel work piloting model-based learning activities in the classroom and by investigating specific attributes of the activities in lab studies and focus groups. To date we have developed and piloted a mature suite of activities covering a variety of topics for both calculus and statics. Class observations and complementary studies in the psychology lab are helping us develop a theoretical framework for using the models in instruction. Close observation of how students use the models to solve problems and as communication tools helps identify effective design elements. We are administering two spatial skills assessments as pre/post instruments: the Purdue Spatial Visualizations Test: Rotations (PSVT:R) in calculus; and the Mental Cutting Test (MCT) in statics. We are also developing strategies and refining approaches for assessing representational competence in both subject areas. Moving forward we will be using these assessments in intervention and control sections of both courses to assess the effectiveness of the models for all learners and subgroups of learners. 
    more » « less
  5. Mechanics instructors frequently employ hands-on learning with goals such as demonstrating physical phenomena, aiding visualization, addressing misconceptions, exposing students to “real-world” problems, and promoting an engaging classroom environment. This paper presents results from a study exploring the importance of the “hands-on” aspect of a hands-on modeling curriculum we have been developing that spans several topics in statics. The curriculum integrates deep conceptual exploration with analysis procedure tutorials and aims to scaffold students’ development of representational competence, the ability to use multiple representations of a concept as appropriate for learning, problem solving, and communication. We conducted this study over two subsequent terms in an online statics course taught in the context of remote learning amidst the COVID-19 pandemic. The intervention section used a take-home adaptation of the original classroom curriculum. This adaptation consisted of eight activity worksheets with a supplied kit of manipulatives and model-building supplies students could use to construct and explore concrete representations of figures and diagrams used in the worksheets. In contrast, the control section used activity worksheets nearly identical to those used in the hands-on curriculum, but without the associated modeling parts kit. We only made minor revisions to the worksheets to remove reference to the models. The control and intervention sections were otherwise identical in how they were taught by the same instructor. We compare learning outcomes between the two sections as measured via pre-post administration of a test of 3D vector concepts and representations called the Test of Representational Competence with Vectors (TRCV). We also compare end of course scores on the Concept Assessment Test in Statics (CATS) and final exam scores. In addition, we analyze student responses on two “multiple choice plus explain” concept questions paired with each of five activities covering the topics of 3D moments, 3D particle equilibrium, rigid body equilibrium (2D and 3D), and frame analysis (2D). The mean pre/post gain across all ten questions was higher for the intervention section, with the largest differences observed on questions relating to 3D rigid body equilibrium. Students in the intervention section also made larger gains on the TRCV and scored better on the final exam compared to the control section, but these results are not statistically significant perhaps due to the small study population. There were no appreciable differences in end-of-course CATS scores. We also present student feedback on the activity worksheets that was slightly more positive for the versions with the models. 
    more » « less