Currently an increasing number of head mounted displays (HMD) for virtual and augmented reality (VR/AR) are equipped with integrated eye trackers. Use cases of these integrated eye trackers include ...rendering optimization and gaze-based user interaction. In addition, visual attention in VR and AR is interesting for applied research based on eye tracking in cognitive or educational sciences for example. While some research toolkits for VR already exist, only a few target AR scenarios. In this work, we present an open-source eye tracking toolkit for reliable gaze data acquisition in AR based on Unity 3D and the Microsoft HoloLens 2, as well as an R package for seamless data analysis. Furthermore, we evaluate the spatial accuracy and precision of the integrated eye tracker for fixation targets with different distances and angles to the user (n=21). On average, we found that gaze estimates are reported with an angular accuracy of 0.83 degrees and a precision of 0.27 degrees while the user is resting, which is on par with state-of-the-art mobile eye trackers.
Conducting experiments fosters conceptual understanding in science education. In various studies, combinations of real (hands-on) and virtual (computer-simulated) experiments have been shown to be ...especially helpful for gaining conceptual understanding. The present systematic review, based on 42 experimental studies, focuses on the following: (1) What is the relative effectiveness of combining real and virtual experiments compared with a single type of experimentation? (2) Which sequence of real and virtual experiments is most effective? The results indicate that (1) in most cases combinations of real and virtual experiments promote conceptual understanding better than a single type of experimentation, and (2) there is no evidence for the superiority of a particular sequence. We conclude that for combining real and virtual experiments, apart from the individual affordances and the learning objectives of the different experiment types, especially their specific function for the learning task must be considered.
Mobile devices (smartphones or tablets) as experimental tools (METs) offer inspiring possibilities for science education, but until now, there has been little research studying this approach. ...Previous research indicated that METs have positive effects on students’ interest and curiosity. The present investigation focuses on potential cognitive effects of METs using video analyses on tablets to investigate pendulum movements and an instruction that has been used before to study effects of smartphones’ acceleration sensors. In a quasi-experimental repeated-measurement design, a treatment group uses METs (TG,
N
TG
= 23) and a control group works with traditional experimental tools (CG,
N
CG
= 28) to study the effects on interest, curiosity, and learning achievement. Moreover, various control variables were taken into account. We suppose that pupils in the TG have a lower extraneous cognitive load and higher learning achievement than those in the CG working with traditional experimental tools. ANCOVAs showed significantly higher levels of learning achievement in the TG (medium effect size). No differences were found for interest, curiosity, or cognitive load. This might be due to a smaller material context provided by tablets, in comparison to smartphones, as more pupils possess and are familiar with smartphones than with tablets. Another reason for the unchanged interest might be the composition of the sample: While previous research showed that especially originally less-interested students profited most from using METs, the current sample contained only specialized courses, i.e., students with a high original interest, for whom the effect of METs on their interest is presumably smaller.
Remote eye tracking has become an important tool for the online analysis of learning processes. Mobile eye trackers can even extend the range of opportunities (in comparison to stationary eye ...trackers) to real settings, such as classrooms or experimental lab courses. However, the complex and sometimes manual analysis of mobile eye-tracking data often hinders the realization of extensive studies, as this is a very time-consuming process and usually not feasible for real-world situations in which participants move or manipulate objects. In this work, we explore the opportunities to use object recognition models to assign mobile eye-tracking data for real objects during an authentic students’ lab course. In a comparison of three different Convolutional Neural Networks (CNN), a Faster Region-Based-CNN, you only look once (YOLO) v3, and YOLO v4, we found that YOLO v4, together with an optical flow estimation, provides the fastest results with the highest accuracy for object detection in this setting. The automatic assignment of the gaze data to real objects simplifies the time-consuming analysis of mobile eye-tracking data and offers an opportunity for real-time system responses to the user’s gaze. Additionally, we identify and discuss several problems in using object detection for mobile eye-tracking data that need to be considered.
Classical statistical analysis of data can be complemented or replaced with data analysis based on machine learning. However, in certain disciplines, such as education research, studies are ...frequently limited to small datasets, which raises several questions regarding biases and coincidentally positive results. In this study, we present a refined approach for evaluating the performance of a binary classification based on machine learning for small datasets. The approach includes a non-parametric permutation test as a method to quantify the probability of the results generalising to new data. Furthermore, we found that a repeated nested cross-validation is almost free of biases and yields reliable results that are only slightly dependent on chance. Considering the advantages of several evaluation metrics, we suggest a combination of more than one metric to train and evaluate machine learning classifiers. In the specific case that both classes are equally important, the Matthews correlation coefficient exhibits the lowest bias and chance for coincidentally good results. The results indicate that it is essential to avoid several biases when analysing small datasets using machine learning.
Subject-method barriers and cognitive load (CL) of students have a particular importance in the complex learning process of scientific inquiry. In this work, we investigate the valid measurement of ...CL as well as different scaffolds to reduce it during experimentation. Specifically, we examine the validity of a subjective measurement instrument to assess CL in extraneous cognitive load (ECL), intrinsic cognitive load, and germane cognitive load (GCL) during the use of multimedia scaffolds in the
planning
phase of the scientific inquiry process based on a theoretical framework of the CL theory. The validity is analyzed by investigating possible relationships between causal (e.g., cognitive abilities) and assessment (e.g., eye-tracking metrics) factors in relation to the obtained test scores of the adapted subjective measurement instrument. The study aims to elucidate possible relationships of causal factors that have not yet been adequately investigated in relation to CL. Furthermore, a possible, still inconclusive convergence between subjective test scores on CL and objectively measured indicators will be tested using different eye-tracking metrics. In two studies (
n
=250), 9th and 11th grade students experimentally investigated a biological phenomenon. At the beginning of the
planning
phase, students selected one of four multimedia scaffolds using a tablet (Study I:
n
=181) or a computer with a stationary eye-tracking device (Study II:
n
=69). The subjective cognitive load was measured
via
self-reports using a standardized questionnaire. Additionally, we recorded students’ gaze data during learning with the scaffolds as objective measurements. Besides the causal factors of cognitive-visual and verbal abilities, reading skills and spatial abilities were quantified using established test instruments and the learners indicated their representation preference by selecting the scaffolds. The results show that CL decreases substantially with higher grade level. Regarding the causal factors, we observed that cognitive-visual and verbal abilities have a significant influence on the ECL and GCL in contrast to reading skills. Additionally, there is a correlation between the representation preference and different types of CL. Concerning the objective measurement data, we found that the absolute fixation number is predictive for the ECL. The results are discussed in the context of the overall methodological research goal and the theoretical framework of CL.
With the recent increase in the use of augmented reality (AR) in educational laboratory settings, there is a need for new intelligent sensor systems capturing all aspects of the real environment. We ...present a smart sensor system meeting these requirements for STEM (science, technology, engineering, and mathematics) experiments in electrical circuits. The system consists of custom experiment boxes and cables combined with an application for the Microsoft HoloLens 2, which creates an AR experiment environment. The boxes combine sensors for measuring the electrical voltage and current at the integrated electrical components as well as a reconstruction of the currently constructed electrical circuit and the position of the sensor box on a table. Combing these data, the AR application visualizes the measurement data spatially and temporally coherent to the real experiment boxes, thus fulfilling demands derived from traditional multimedia learning theory. Following an evaluation of the accuracy and precision of the presented sensors, the usability of the system was evaluated with n=20 pupils in a German high school. In this evaluation, the usability of the system was rated with a system usability score of 94 out of 100.
Generative AI technologies such as large language models show novel potential to enhance educational research. For example, generative large language models were shown to be capable of solving ...quantitative reasoning tasks in physics and concept tests such as the Force Concept Inventory (FCI). Given the importance of such concept inventories for physics education research, and the challenges in developing them such as field testing with representative populations, this study seeks to examine to what extent a generative large language model could be utilized to generate a synthetic dataset for the FCI that exhibits content-related variability in responses. We use the recently introduced ChatGPT based on the GPT 4 generative large language model and investigate to what extent ChatGPT could solve the FCI accurately (RQ1) and could be prompted to solve the FCI as if it were a student belonging to a different cohort (RQ2). Furthermore, we study, to what extent ChatGPT could be prompted to solve the FCI as if it were a student having a different force- and mechanics-related preconception (RQ3). In alignment with other research, we found that ChatGPT could accurately solve the FCI. We furthermore found that prompting ChatGPT to respond to the inventory as if it belonged to a different cohort yielded no variance in responses, however, responding as if it had a certain preconception introduced much variance in responses that approximate real human responses on the FCI in some regards.
Domain-specific understanding of digitally represented graphs is necessary for successful learning within and across domains in higher education. Two recent studies conducted a cross-sectional ...analysis of graph understanding in different contexts (physics and finance), task concepts, and question types among students of physics, psychology, and economics. However, neither changes in graph processing nor changes in test scores over the course of one semester have been sufficiently researched so far. This eye-tracking replication study with a pretest–posttest design examines and contrasts changes in physics and economics students’ understanding of linear physics and finance graphs. It analyzes the relations between changes in students’ gaze behavior regarding relevant graph areas, scores, and self-reported task-related confidence. The results indicate domain-specific, context- and concept-related differences in the development of graph understanding over the first semester, as well as its successful transferability across the different contexts and concepts. Specifically, we discovered a tendency of physics students to develop a task-independent overconfidence in the graph understanding during the first semester.