This work in progress paper investigates the use of AI chatbots by undergraduate computer science students learning English as a second language in Japan. The study employs a mixed-methods approach, combining qualitative surveys and interviews with quantitative clustering analysis. The objectives are to identify the types of AI chatbots used, determine their usage patterns, and explore the benefits and challenges associated with their use in language learning. The qualitative data collection (surveys, n=96) has been completed, while interviews and quantitative analysis are ongoing. The study aims to identify distinct clusters of AI chatbot users and their characteristics, highlight challenges, and contribute to the growing knowledge on AI application in STEM students studying a second language. Future research should explore long-term effects, optimal balance between AI-assisted and human-led instruction, and guidelines for integrating AI chatbots in language learning curricula.

This paper proposes a visualization method for the mixed type of high-dimension and low sample-size (HDLSS) data. The data consists of both numerical and categorical data with respect to quantitative and qualitative variables, and the number of variables is much larger than the number of objects. The proposed method can treat both types of data simultaneously in a lower dimensional space. From this, we can compare the two kinds of features obtained from the two types of data in the lower dimensional space. In addition, since this visualization can be obtained based on the mathematical theory of metric multidimensional scaling, the comparability is mathematically guaranteed. As a numerical example, we show how the devices used by students to perform activities of reading, writing, listening, and speaking English texts of different lengths are related to their English test scores. In this case, the choice of device is obtained as qualitative data, and the test scores are obtained as quantitative data. In order to visually obtain these relationships, the two must be mathematically comparable in the same low-dimensional space. With the proposed method, we have achieved this and shown that it is possible to interpret the relationship between the two.

This work-in-progress (WiP) research explores the role of rubrics in mitigating the negative impact of generative AI, such as ChatGPT, on writing assessment practices in STEM. This approach addresses the growing need for innovative methods of ensuring student academic integrity and authenticity in the rapidly expanding ecosystem of AI tools. A rubric consisting of five criteria is employed to rate the students' deconstruction of written text into language frames for the purpose of differentiating between human-written and AI-generated content. The language frames are common English language sentence patterns used for expressing five academic written functions: compare/contrast, cause/effect, classification, chronological order, and spatial order. By evaluating the performance of student deconstruction of one paragraph written by the student and a second paragraph produced by ChatGPT, it is anticipated that the rubric will enable the instructor to differentiate between the two by capturing any gaps in knowledge required to identify, deconstruct, and reproduce previously learned sentence frames. This assumes that the student will be more familiar with a self-written text than an unfamiliar AI produced one. This difference may then be employed by educators to aid in the identification of AI-generated plagiarism submitted by students. The key insights from this pilot study include: The need for a rubric threshold level of between 70% and 80% to differentiate between human and AI texts. Students appear to score higher at the identification of language frames than the production of the same frames. They were equal or better at identifying sentence frames from the AI generated text. Also, students scored very low on critical thinking questions that required the selection of alternative sentence frames. This WiP paper details the rubric design, research methodology, and preliminary insights from a small pilot study, which informs the evolution towards a larger future implementation.

This paper proposes a fuzzy cluster-scaled principal component analysis (fuzzy cluster-scaled PCA) for mixed data, which consists of both numerical and categorical data with respect to quantitative and qualitative variables. The fuzzy cluster-scaled PCA has been proposed for high-dimension, low-sample size (HDLSS) data in which the number of variables (or dimensions) is much larger than the number of objects (or samples). In this case, the target data is only for the numerical data. The proposed fuzzy cluster-scaled PCA in this paper can be applied to the mixed type of HDLSS data by utilizing the feature of the fuzzy cluster-scaled correlation, which is decomposed into two parts: the first part is the correlation of classification structures between variables and the second part is the correlation between variables. Then, the first part can be adapted to the categorical data, and the second part can be used for the numerical data through the same objects. Several numerical examples used data from a survey concerned with the relationship between students’ choice of devices and scores of examination/class marks to measure the educational effect show a better performance of this proposed method, and insightful, important information on educational effectiveness is clarified.

This work in progress research will interest educational stakeholders in the STEM area dealing with Emergency Remote Teaching (ERT) and researchers interested in the affordance of Information and Communications Technologies (ICT) for education and especially providing personalized learning opportunities. The first question of this research seeks to identify students’ general attitudes toward educational ICT. The second question is to identify any common usage patterns. And the third question is to identify affordances of the technology from the perspective of the students. The preliminary findings suggest that recent graduates in the United States have a sophisticated understanding of what educational technology is and how it can benefit their education. This is reassuring when considering the need for a sudden move to off-site teaching necessitated by an ERT. Several concerns were identified, including information quality and distractions from online entertainment. In addition, technical issues are a concern for many respondents. The qualitative questionnaire and coding have provided some insight into the perceived view of educational technology held by recent engineering graduates in the United States. This is an initial phase of this research which is ongoing and will be expanded to include a broader range of analysis techniques.

This paper proposes two methods. One is the cluster identification method for 3-way dissimilarity data among objects over times (or subjects) and the other is the cluster scaling method for dissimilarity data among objects. Both methods are based on the comparative quantification model which can obtain the quantitative amount of relationship between a pair of clusters or relationship between a cluster and a basis which spans a subspace constructed a scale. The merits of these methods are that we can obtain “comparability” of obtained clusters over times (or subjects) and supply an “adaptable scale” for observed dissimilarity between a pair of objects, in order to reduce the number of dimensions of the observed data and explain the dissimilarity relationships among objects in the lower dimensional subspace. Numerical examples to investigate the educational effectiveness by using the cognitive 3-way dissimilarity data of students demonstrate a better performance for the proposed methods.

This paper presents a visualization of a result of fuzzy clustering. The feature of fuzzy clustering is to obtain the degree of belongingness of objects to fuzzy clusters so the result will be more commensurate with reality. In addition, the number of clusters requires less and the solution of the result will be more robust when compared with conventional hard clustering. In contrast, the fuzzy clustering result interpretation tends to be more complicated. Therefore, measuring the similarity (or dissimilarity) between a pair of fuzzy classification status of objects is important. In order to measure the similarity (or dissimilarity) mathematically, it is necessary to introduce a scale to the fuzzy clustering result. That is, the obtained solutions as a fuzzy clustering result must be in a metric space. In order to implement this, we have proposed multidimensional joint scale and cluster analysis. In this analysis, we exploit a scale obtained by multidimensional scaling. This paper clarifies that the multidimensional joint scale and cluster analysis introduces scale to the fuzzy clustering result and then the visualization of the fuzzy clustering result in the metric vector space has a theoretical mathematical meaning through the Euclidean distance structure. In this paper, this is shown by using several numerical comparisons with ordinary visualizations of the fuzzy clustering result.