This research paper systematically identifies the perceptions of learning machine learning (ML) topics. To keep up with the ever-increasing need for professionals with ML expertise, for-profit and non-profit organizations conduct a wide range of ML-related courses at undergraduate and graduate levels. Despite the availability of ML-related education materials, there is lack of understanding how students perceive ML-related topics and the dissemination of ML-related topics. A systematic categorization of students' perceptions of these courses can aid educators in understanding the challenges that students face, and use that understanding for better dissemination of ML-related topics in courses. The goal of this paper is to help educators teach machine learning (ML) topics by providing an experience report of students' perceptions related to learning ML. We accomplish our research goal by conducting an empirical study where we deploy a survey with 83 students across five academic institutions. These students are recruited from a mixture of undergraduate and graduate courses. We apply a qualitative analysis technique called open coding to identify challenges that students encounter while studying ML-related topics. Using the same qualitative analysis technique we identify quality aspects do students prioritize ML-related topics. From our survey, we identify 11 challenges that students face when learning about ML topics, amongst which data quality is the most frequent, followed by hardware-related challenges. We observe the majority of the students prefer hands-on projects over theoretical lectures. Furthermore, we find the surveyed students to consider ethics, security, privacy, correctness, and performance as essential considerations while developing ML-based systems. Based on our findings, we recommend educators who teach ML-related courses to (i) incorporate hands-on projects to teach ML-related topics, (ii) dedicate course materials related to data quality, (iii) use lightweight virtualization tools to showcase computationally intensive topics, such as deep neural networks, and (iv) empirical evaluation of how large language models can be used in ML-related education.

Simulations are widely used to teach science in grade schools.These simulations are often augmented with a conversational artificial intelligence (AI) agent to provide real-time scaffolding support for students conducting experiments using the simulations.AI agents are highly tailored for each simulation, with a predesigned set of Instructional Goals (IGs), making it difficult for teachers to adjust IGs as the agent may no longer align with the revised IGs.Additionally, teachers are hesitant to adopt new third-party simulations for the same reasons.In this research, we introduce SimPal, a Large Language Model (LLM) based meta-conversational agent, to solve this misalignment issue between a pre-trained conversational AI agent and the constantly evolving pedagogy of instructors.Through natural conversation with SimPal, teachers first explain their desired IGs, based on which SimPal identifies a set of relevant physical variables and their relationships to create symbolic representations of the desired IGs.The symbolic representations can then be leveraged to design prompts for the original AI agent to yield better alignment with the desired IGs.We empirically evaluated SimPal using two LLMs, ChatGPT-3.5 and PaLM 2, on 63 Physics simulations from PhET and Golabz.Additionally, we examined the impact of different prompting techniques on LLM's performance by utilizing the TELeR taxonomy to identify relevant physical variables for the IGs.Our findings showed that SimPal can do this task with a high degree of accuracy when provided with a well-defined prompt.

Use of technology-enhanced education and online learning systems has become more popular, especially after COVID-19. These systems capture a rich array of data as students interact with them. Predicting student performance is an essential part of technology-enhanced education systems to enable the generation of hints and provide recommendations to students. Typically, this is done through use of data on student interactions with questions without utilizing important data on the temporal ordering of students’ other interaction behavior, (e.g., reading, video watching). In this paper, we hypothesize that to predict students’ question performance, it is necessary to (i) consider other learning activities beyond question-answering and (ii) understand how these activities are related to question-solving behavior. We collected middle school physical science students’ data within a K12 reading platform, Actively Learn. This platform provides reading-support to students and collects trace data on their use of the system. We propose a transformer-based model to predict students' question scores utilizing question interaction and reading-related behaviors. Our findings show that integrating question attempts and reading-related behaviors results in better predictive power compared to using only question attempt features. The interpretable visualization of the transformer’s attention can be helpful for teachers to make tailored interventions in students’ learning.

Defects in infrastructure as code (IaC) scripts can have serious consequences, for example, creating large-scale system outages. A taxonomy of IaC defects can be useful for understanding the nature of defects, and identifying activities needed to fix and prevent defects in IaC scripts. The goal of this paper is to help practitioners improve the quality of infrastructure as code (IaC) scripts by developing a defect taxonomy for IaC scripts through qualitative analysis. We develop a taxonomy of IaC defects by applying qualitative analysis on 1,448 defect-related commits collected from open source software (OSS) repositories of the Openstack organization. We conduct a survey with 66 practitioners to assess if they agree with the identified defect categories included in our taxonomy. We quantify the frequency of identified defect categories by analyzing 80,425 commits collected from 291 OSS repositories spanning across 2005 to 2019.

Rich models of students' learning and problem-solving behaviors can support tailored interventions by instructors and scaffolding of complex learning activities. Our goal in this paper is to identify students' reading behaviors as they engage with instructional texts in domain-specific activities. In this work, we apply theory and methodology from the learning sciences to a large-scale middle school dataset within a digital literacy platform, Actively Learn. We compare students' reading behaviors both within and across domains for 12,566 science and 16,240 social studies students. Our findings show that higher-performing students in science engaged in more metacognitively-rich reading activities, such as text annotation; whereas lower-performing students relied more on simple highlighting and took longer to respond to embedded questions. Higher-performing students in social studies, by contrast, engaged more with the vocabulary and took longer to read before attempting question responses. Our finding may be used as recommendations to help both teachers and students engage in and support more effective behaviors.

Scientific software is defined as software that is used to analyze data to investigate unanswered research questions in the scientific community. Developers use programming languages such as Julia to build scientific software. When programming with Julia, developers experience program execution time discrepancy i.e. not obtaining desired program execution time, which hinders them to efficiently complete their tasks. The goal of this paper is to help developers in achieving desired program execution time for Julia by identifying the causes of why program execution time discrepancies happen with an empirical study of Stack Overflow posts. We conduct an empirical study with 263 Julia-related posts collected from Stack Overflow, and apply qualitative analysis on the collected 263 posts. We identify 9 categories of program execution time discrepancies for Julia, which include discrepancies related to data structures usage such as, arrays and dictionaries. We also identify 10 causes that explain why the program execution time discrepancies happen. For example, we identify program execution time discrepancy to happen when developers unnecessarily allocate memory by using array comprehension.

Static analysis tool alerts can help developers detect potential defects in the code early in the development cycle. However, developers are not always able to respond to the alerts with their preferred action and may turn away from using the tool. In this paper, we qualitatively analyze 280 Stack Overflow (SO) questions regarding static analysis tool alerts to identify the challenges developers face in understanding and responding to these alerts. We find that the most prevalent question on SO is how to ignore and filter alerts, followed by validation of false positives. Our findings confirm prior researchers' findings related to notification communication theory as 44.6% of the SO questions that we analyzed indicate developers face communication challenges.

Despite being the most popular question and answer website for software developers, answers posted on Stack Overflow (SO) are susceptible to contain Python-related insecure coding practices. A systematic analysis on how frequently insecure coding practices appear in SO answers can help the SO community assess the prevalence of insecure Python code blocks in SO. An insecure coding practice is recurrent use of insecure coding patterns in Python. We conduct an empirical study using 529,054 code blocks collected from Python-related 44,966 answers posted on SO. We observe 7.1% of the 44,966 Python-related answers to include at least one insecure coding practice. The most frequently occurring insecure coding practice is code injection. We observe 9.8% of the 7,444 accepted answers to include at least one insecure code block. We also find user reputation not to relate with the presence of insecure code blocks, suggesting that both high and low-reputed users are likely to introduce insecure code blocks.

Rule-based classification is a popular approach for solving real world classification problems. Once suitable rules have been obtained, rule-based classifiers are easy to deploy and explain. In this paper, we describe an approach that uses biogeography-based optimization (BBO) to compute rule sets that maximize predictive accuracy. BBO is an evolutionary algorithm inspired by the migration patterns of species between the islands of an archipelago. In our implementation, each species corresponds to a classification rule, each island is occupied by multiple species and corresponds to a classifier, and the fitness of an island is computed as the predictive classification accuracy of the corresponding classifier. The archipelago evolves via mutation, selection, and migration of species between islands. Successful islands have a decreased immigration rate and an increased emigration rate. In general, such islands tend to resist invasion and to colonize less successful islands. This results in an evolving set of habitats that corresponds to a population of classifiers. We demonstrate the effectiveness of our approach by comparing it to several traditional and evolutionary based state-of-the-art classifiers.