In recent years, the rise of advanced artificial intelligence technologies has had a profound impact on many fields, including education and research. One such technology is ChatGPT, a powerful large language model developed by OpenAI. This technology offers exciting opportunities for students and educators, including personalized feedback, increased accessibility, interactive conversations, lesson preparation, evaluation, and new ways to teach complex concepts. However, ChatGPT poses different threats to the traditional education and research system, including the possibility of cheating on online exams, human-like text generation, diminished critical thinking skills, and difficulties in evaluating information generated by ChatGPT. This study explores the potential opportunities and threats that ChatGPT poses to overall education from the perspective of students and educators. Furthermore, for programming learning, we explore how ChatGPT helps students improve their programming skills. To demonstrate this, we conducted different coding-related experiments with ChatGPT, including code generation from problem descriptions, pseudocode generation of algorithms from texts, and code correction. The generated codes are validated with an online judge system to evaluate their accuracy. In addition, we conducted several surveys with students and teachers to find out how ChatGPT supports programming learning and teaching. Finally, we present the survey results and analysis.

Programming is a vital skill in computer science and engineering-related disciplines. However, developing source code is an error-prone task. Logical errors in code are particularly hard to identify for both students and professionals, and a single error is unexpected to end-users. At present, conventional compilers have difficulty identifying many of the errors (especially logical errors) that can occur in code. To mitigate this problem, we propose a language model for evaluating source codes using a bidirectional long short-term memory (BiLSTM) neural network. We trained the BiLSTM model with a large number of source codes with tuning various hyperparameters. We then used the model to evaluate incorrect code and assessed the model’s performance in three principal areas: source code error detection, suggestions for incorrect code repair, and erroneous code classification. Experimental results showed that the proposed BiLSTM model achieved 50.88% correctness in identifying errors and providing suggestions. Moreover, the model achieved an F-score of approximately 97%, outperforming other state-of-the-art models (recurrent neural networks (RNNs) and long short-term memory (LSTM)).

With the rapid advancement of technology and its easy accessibility, online activity has become an integral part of everyday human life. Expressing opinions, providing feedback, and sharing feelings by commenting on various platforms, including social media, education, business, entertainment, and sports, has become a common phenomenon. Effectively analyzing these comments to uncover latent intentions holds immense value in making strategic decisions across various domains. However, several challenges hinder the process of sentiment analysis including the lexical diversity exhibited in comments, the presence of long dependencies within the text, encountering unknown symbols and words, and dealing with imbalanced datasets. Moreover, existing sentiment analysis tasks mostly leveraged sequential models to encode the long dependent texts and it requires longer execution time as it processes the text sequentially. In contrast, the Transformer requires less execution time due to its parallel processing nature. In this work, we introduce a novel hybrid deep learning model, RoBERTa-BiLSTM, which combines the Robustly Optimized BERT Pretraining Approach (RoBERTa) with Bidirectional Long Short-Term Memory (BiLSTM) networks. RoBERTa is utilized to generate meaningful word embedding vectors, while BiLSTM effectively captures the contextual semantics of long-dependent texts. The RoBERTa-BiLSTM hybrid model leverages the strengths of both sequential and Transformer models to enhance performance in sentiment analysis. We conducted experiments using datasets from IMDb, Twitter US Airline, and Sentiment140 to evaluate the proposed model against existing state-of-the-art methods. Our experimental findings demonstrate that the RoBERTa-BiLSTM model surpasses baseline models (e.g., BERT, RoBERTa-base, RoBERTa-GRU, and RoBERTa-LSTM), achieving accuracies of 80.74%, 92.36%, and 82.25% on the Twitter US Airline, IMDb, and Sentiment140 datasets, respectively. Additionally, the model achieves F1-scores of 80.73%, 92.35%, and 82.25% on the same datasets, respectively.

As the educational landscape evolves, understanding and fostering student adaptability has become increasingly critical. This study presents a comparative analysis of XAI techniques to interpret machine learning models aimed at classifying student adaptability levels. Leveraging a robust dataset of 1205 instances, we employed several machine learning algorithms with a particular focus on Random Forest, which demonstrated highest accuracy at 91%. The models’ precision, recall and F1-score were also evaluated, with Random Forest achieving a precision of 0.93, a recall of 0.94, and an F1-score of 0.94. Our study utilizes SHAP, LIME, Anchors, ALE, and Counterfactual explanations to reveal the specific contributions of various features impacting adaptability predictions. SHAP values highlighted ‘Class Duration’ significance (mean SHAP value: 0.175); LIME explained socio-economic and institutional factors’ intricate influence. Anchors provided high-confidence rule-based explanations (confidence: 97.32%), emphasizing demographic characteristics. ALE analysis underscored the importance of ‘Financial Condition’ with a positive slope, while Counterfactual scenarios highlighted the impact of slight feature variations of 0.5 change in ‘Class Duration’. Consistently, ‘Class Duration’ and ‘Financial Condition’ emerge as key factors, while the study also underscores the subtle effects of ‘Institution Type’ and ‘Load-shedding’. This multi-faceted interpretability approach bridges the gap between machine learning performance and educational relevance, presenting a model that not only predicts but also explains the dynamic factors influencing student adaptability. The synthesized insights advocate for educational policies accommodating socioeconomic factors, instructional time, and infrastructure stability to enhance student adaptability. The implications extend to informed and personalized educational interventions, fostering an adaptable learning environment. This methodical research contributes to responsible AI application in education, promoting predictive and interpretable models for equitable and effective educational strategies.

The rate of software development has increased dramatically. Conventional compilers cannot assess and detect all source code errors. Software may thus contain errors, negatively affecting end-users. It is also difficult to assess and detect source code logic errors using traditional compilers, resulting in software that contains errors. A method that utilizes artificial intelligence for assessing and detecting errors and classifying source code as correct (error-free) or incorrect is thus required. Here, we propose a sequential language model that uses an attention-mechanism-based long short-term memory (LSTM) neural network to assess and classify source code based on the estimated error probability. The attentive mechanism enhances the accuracy of the proposed language model for error assessment and classification. We trained the proposed model using correct source code and then evaluated its performance. The experimental results show that the proposed model has logic and syntax error detection accuracies of 92.2% and 94.8%, respectively, outperforming state-of-the-art models. We also applied the proposed model to the classification of source code with logic and syntax errors. The average precision, recall, and F-measure values for such classification are much better than those of benchmark models. To strengthen the proposed model, we combined the attention mechanism with LSTM to enhance the results of error assessment and detection as well as source code classification. Finally, our proposed model can be effective in programming education and software engineering by improving code writing, debugging, error-correction, and reasoning.

Computer programming has attracted a lot of attention in the development of information and communication technologies in the real world. Meeting the growing demand for highly skilled programmers in the ICT industry is one of the major challenges. In this point, online judge (OJ) systems enhance programming learning and practice opportunities in addition to classroom-based learning. Consequently, OJ systems have created a large number of problem-solving data (solution codes, logs, and scores) archives that can be valuable raw materials for programming education research. In this paper, we propose an educational data mining framework to support programming learning using unsupervised algorithms. The framework includes the following sequence of steps: ( $i$ ) problem-solving data collection (logs and scores are collected from the OJ) and preprocessing; ( $ii$ ) MK-means clustering algorithm is used for data clustering in Euclidean space; ( $iii$ ) statistical features are extracted from each cluster; ( $iv$ ) frequent pattern (FP)-growth algorithm is applied to each cluster to mine data patterns and association rules; ( $v$ ) a set of suggestions are provided on the basis of the extracted features, data patterns, and rules. Different parameters are adjusted to achieve the best results for clustering and association rule mining algorithms. For the experiment, approximately 70,000 real-world problem-solving data from 537 students of a programming course (Algorithm and Data Structures) were used. In addition, synthetic data have leveraged for experiments to demonstrate the performance of MK-means algorithm. The experimental results show that the proposed framework effectively extracts useful features, patterns, and rules from problem-solving data. Moreover, these extracted features, patterns, and rules highlight the weaknesses and the scope of possible improvements in programming learning.