This is a hands-on laptop-recommended workshop that introduces partici-

pants to the PLCC compiler-compiler tool set for building Java-based language

interpreters in an upper-level Programming Languages course. One approach

to teaching this course is to show how programming language features such

as language syntax, variable lifetime, procedure application, parameter pass-

ing, recursion, and object-orientation are implemented "under the hood". This

workshop is designed for CS educators who want to explore how PLCC can be

used to support such an approach.

This hands-on workshop shows how to integrate generative AI into course design, assessment, and feedback—without surrendering pedagogy to tooling. After a quick live poll to select a familiar “Intro to Computing” style topic, we’ll co‑build a starter module: outline learning outcomes, generate draft content, create fresh question banks, and configure auto-graded and rubric‑based assessments. We’ll demonstrate platform‑agnostic patterns (prompt libraries, context packaging, and process‑based grading) and then illustrate the same patterns in an integrated demo environment for speed and scale. Participants will also configure AI assistants that give targeted hints and feedback while preserving academic integrity through transparent constraints and logs. All attendees receive a reproducible checklist and a starter pack they can integrate with their own LMS or adapt to their own platform. Free Instructor Codio accounts are pre‑provisioned to minimize setup time. (No platform purchase is required.)

Introductory programming students routinely struggle with the “invisible” aspects of modular programs: where values come from, how they move between functions, and what it means to “return” a result. Amp Stack is a lightweight, visual methodology that makes data flow and modular structure explicit by having students design programs as top-to-bottom stacks of “amps” (Input, Compute, and Output blocks) wired together by named ports before they write any code. In this 3-hour, hands-on workshop, participants will experience Amp Stack from the student point of view, translate diagrams into runnable Python, and then step back to examine how the approach scaffolds CS1 concepts such as IPO structure, parameter passing, and modular decomposition. Attendees will leave with ready-to-use classroom materials (worksheets, sample exercises, and rubrics) and a plan for integrating Amp Stack into their own introductory courses with minimal disruption to existing syllabi.

Data visualization is a powerful tool for identifying patterns and interpreting complex datasets. However, effective visualizations critically depend on the accuracy of the underlying data marts—typically created through SQL GROUP BY clauses and aggregate functions. Because each visualization type (e.g., line charts, bar charts, pie charts, scatter plots, histograms) requires a correctly structured data mart, even small mistakes in data preparation can lead to misleading graphs and incorrect conclusions.

This study examines several recurring errors made by students when constructing data marts for visualization tasks and illustrates how these mistakes can distort subsequent analyses. Using the Simplemaps US ZIP code demographic dataset—where each record represents a ZIP code—we present practical examples and demonstrate correct and incorrect approaches. For instance, when creating a scatter plot to compare average income and unemployment rate by state, students often incorrectly apply SQL’s AVG() function directly to ZIP code–level data, failing to account for population differences that introduce unintended weighting. Another common error occurs when comparing county-level populations by grouping solely on county_name without including the state, inadvertently merging counties with the same name across different states.

Because users often focus on charts rather than on how the underlying data marts were generated, these errors can remain hidden yet have significant consequences for interpretation and decision-making. This work highlights the importance of teaching students how to construct accurate data marts and provides guidance for avoiding common aggregation pitfalls in SQL-based analytics

Ransomware remains one of the most disruptive cyber threats to critical infrastructure, with healthcare systems particularly vulnerable due to their stringent availability and privacy requirements. This lightning talk presents Ransom Sentinel (RaSe), a multi-layered ransomware resilience framework that integrates offensive attack simulation with cryptographic and systems-level defense mechanisms. Unlike many proprietary or black-box solutions, RaSe is grounded in reproducible adversarial modeling: reconnaissance, unauthorized access, and payload execution were emulated using tools such as Nmap and custom client–server exploits (like Heartbleed) to expose real attack surfaces and failure modes. Building on these insights, RaSe combines Shamir’s Secret Sharing for distributed key management, Reed–Solomon erasure coding for fault-tolerant storage, entropy-based anomaly detection for early encryption discovery, Zero Trust access control, and immutable audit logging to support forensic integrity and regulatory compliance. The framework is evaluated in the context of healthcare workflows, emphasizing the Durability–Accessibility–Loss (DAL) principles and demonstrating how secure recovery and “break-glass” access can coexist with strong cryptographic safeguards. This work highlights how integrating offensive experimentation with defensive architecture can transform ransomware response from reactive recovery to proactive resilience. The lightning talk will summarize the system design, threat model, and educational value of using adversarial simulation to teach and research next-generation cyber defense strategies.

As AI tools continue to improve automated code generation, CS educators need to ensure students develop problem-solving skills of their own. This work presents collaborative whiteboard-based activities used in an Introduction to Data Structures course. Through these hands-on activities, students work together to reveal and address their conceptual gaps while building their problem-solving abilities.

Zviel-Girshin [4] showed students appear to progress while lacking genuine comprehension while using AI tools. Güner and Er [2] identified student interaction profiles with ChatGPT, showing that many relied on AI for direct solutions without engaging in the problem-solving process. Meanwhile, research demonstrates the effectiveness of using traditional whiteboards. Chapin and Bowen [1] showed that whiteboard activities improve mental models and computational thinking by making abstract concepts visible and manipulable. Wong et al. [3] found that students prefer traditional whiteboards over digital tools for in-person learning.

In this work, we describe whiteboard activities used in our Data Structures classroom, using the example of teaching linked lists. Our four-step approach as described below ensures that students actively engage with the problem-solving process, while the visible nature of whiteboard work allows peers and the instructor to collectively identify misconceptions and facilitate improvements. We show how we use this whiteboarding approach in teaching linked lists and how students learn from each of these steps with concrete examples.

• Step 1. Each group works independently to discuss and develop their own solutions on paper.

• Step 2. All groups write their solution on the whiteboard, making their thinking process and solution visible to the entire class.

• Step 3. Students engage in peer critique by analyzing and discussing each other’s approaches, moderated by the instructor.

• Step 4. Each group revises and improves their work on the whiteboard.

Student feedback to this approach was very positive. They particularly appreciated the peer to peer interactions and the hands-on collaborative nature of the activities. They also reported increased confidence in their ability to work through problems.

We thus argue that traditional teaching methods such as our whiteboarding activities are more valuable than ever in the AI era. By having students externalize their thinking on a whiteboard, they can identify gaps in their understanding and improve through working through solutions by hand, all in a group setting. We plan to conduct a more rigorous evaluation of how these whiteboarding activities affect students’ learning compared to AI coding assistants in our Data Structures course.

References

[1] Chapin, J., & Bowen, B. (2023). Whiteboarding: A tool to improve CS1 student self-efficacy. In Proceedings of the 2023 ACM Conference on Global Computing Education Vol 2 (CompEd 2023) (pp. 161-167). ACM. https://doi.org/10.1145/3576882.3617925

[2] Güner, H., & Er, E. (2025). AI in the classroom: Exploring students’ interaction with ChatGPT in programming learning. Education and Information Technologies, 30, 12681–12707. https://doi.org/10.1007/s10639-025-13337-7

[3] Wong, S. S., Wong, S. F., Mahmud, M. M., & Yong, W. K. (2023). Learning in the margins: Student choice for digital and traditional whiteboards in in-person learning at university. In Proceedings of the 2023 6th International Conference on Educational Technology Management (ICETM 2023) (pp.162-168). ACM. https://doi.org/10.1145/3637907.3637969

[4] Zviel-Girshin, R. (2024). The good and bad of AI tools in novice programming education. Education Sciences, 14(10), 1089. https://doi.org/10.3390/educsci14101089

As artificial intelligence (AI) reshapes industries and academic practice, higher education faces a widening literacy gap between rapidly advancing technologies and the readiness of students. This paper introduces a new undergraduate course designed to build foundational AI literacy through hands-on exploration, ethical reasoning, and creative inquiry. Grounded in reputable frameworks and guidelines, the course equips learners to understand core technical AI concepts, build and apply generative AI (GenAI) tools responsibly, evaluate the societal impacts, and study interdisciplinary applications. Through active learning activities ranging from prompt engineering and vibecoding to bias analysis and environmental sustainability, students develop technical, critical, and ethical fluency necessary for participation in an AI-enhanced world.

This lightning talk will briefly introduce and solicit feedback on “Essential Computing Concepts”, a draft set of proposed guidelines developed as an alternative to CS2023 for curriculum design and assessment. The publication of CS2023 resurfaced a long-known tension between curriculum guidance through large and detailed lists of topics, and the curricular priorities and constraints of some programs. For example, smaller institutions, liberal arts colleges, and computing programs looking to innovate around CS+X, would benefit from a smaller list of what is truly essential to computing majors, which can then be expanded based on that program’s focus to fully develop the major. Following several sessions at SIGCSE TS 2025 on this topic sponsored by the SIGCSE Computing Education in Liberal Arts Colleges committee, a dozen interested faculty from a variety of institutions formed a working group committed to developing this smaller core that would support flexibility, exploration, and interdisciplinary work within a CS curriculum. The group designed a draft set of guidelines, structured around 12 learning goals that are organized into four categories: Read, Understand, and Analyze Models; Design and Create Using Models; Use Math, Data and Analytics for Computer Science; and Be a Responsible Computing Practitioner. The guidelines are presented in a way that avoids presuming a particular mapping of concepts to courses and encourages programs to think about how high-level learning outcomes are achieved across an entire major. This lightning talk introduces this draft of Essential Computing Concepts and associated learning goals, and invites attendees to provide feedback on this work in progress.

Course-Based Undergraduate Research Experiences (CUREs) engage students in authentic, open-ended research in which learners collaboratively investigate meaningful questions, iterate on methods, and contribute research products beyond the classroom. While well established in the life sciences, CUREs remain less common—and less clearly articulated—in computer science. This lightning talk will introduce a proposed CURE-informed training framework being developed through two coordinated Vermont Biomedical Research Network (VBRN) Pilots at Landmark College, a college specializing in the education of neuro-divergent learners. The Pilots are designed to work in tandem, providing an opportunity to make research workflows, expectations, and collaboration practices explicit—features expected to support neuro-divergent students while benefiting a broad range of learners. The project will consist of a Computer Science Pilot and a complementary Biology Pilot. The Pilots will interface through shared research questions, coordinated training activities, and aligned workflows, together forming an interdisciplinary CURE model that supports collaboration rather than treating disciplinary preparation in isolation. The Pilots will center on research-team-based CUREs situated outside formal coursework, preserving core CURE elements—authentic research, discovery, collaboration, and iteration—while reducing barriers associated with grading pressure and rigid pacing. Within this structure, students in the Computer Science Pilot will collaborate with Biology students to design reproducible Python-based workflows for analyzing open biomedical imaging datasets. Although developed within a neuro-divergent-serving institutional context, the training framework emerging from the Pilots is intentionally designed to be transferable and neuro-diversity inclusive. By sharing this work at the design stage, we aim to incorporate practitioner insight into the Pilot design itself, strengthening both the framework and its relevance for computer science educators across institutional contexts.

Assessments featuring open-ended and coding questions are staples of the Computer Science curriculum, yet automatically validating these answers remains a significant challenge. While Regular Expressions (regex) are the industry standard for text pattern matching, their cryptic syntax makes them notoriously difficult to write, read, and maintain. Existing tools typically focus on testing regex or inserting common patterns, but fail to make the editing process intuitive or the results readable. This work introduces ReRead, a human-readable text pattern editor and JavaScript library that generates standard regular expressions compatible with major Learning Management Systems (LMS) and auto-graders. We demonstrate ReRead in two distinct scenarios: (1) creating flexible answer keys for open-ended questions in LMS exams, and (2) generating unit tests for student code within a custom Python auto-grading platform.

This paper presents ConnectPlus, a full-stack project management

application designed to unify multiple collaboration features within a single

platform, including a dashboard, project workspace, messaging, task

tracking, alerts, and user settings. The application addresses a common

challenge in group work: communication often becomes scattered across

multiple tools, leading to missed updates and fragmented information. The

proposed application, ConnectPlus provides a centralized environment

where students, instructors, and coworkers can manage tasks, share

updates, and maintain all project information in one dependable location.

The proposed application is implemented using MERN stack, MongoDB,

Express.js, React.js, and Node.js, with supporting middleware such as

CORS, JSON parsing, and environment configuration to ensure reliable

server behavior. A key feature is the integrated alerts system, which enables

users to send immediate notifications about needs, locations, or urgent

updates, helping prevent important information from being lost. The

proposed application serves both as a practical collaboration application and

as a demonstration of full-stack development, backend design, and scalable

communication features.

As part of a computer science curriculum, it is typical to separate individual fields into separate courses. Specifically for security, this is difficult due to the breadth of the field and the amount of knowledge from other fields necessary for a well-rounded understanding. While many larger institutions are moving towards programs in this field, many small liberal arts colleges do not have the resources to support this. In this work, we examine the difficulties of teaching security in this setting and compare techniques for teaching this difficult topic.

Students often struggle to understand data structures due to the challenges of developing mental models of concepts like abstract data representations, memory organization, and the ways these structures evolve during program execution. This tutorial introduces a suite of hands-on activities designed to make abstract data structures tangible by manipulating concrete physical models.

Drawing on insights from active learning in STEM education, we will present ready-to-use activities that allow students to explore common data structures and fundamental algorithms by manipulating physical objects. By initially focusing on the development of accurate mental models, these activities help students scaffold their understanding towards formal representations or code-based implementations.

The tutorial is aimed at instructors interested in active learning approaches. Participants will engage with several of these activities firsthand and discuss strategies for integrating them into CS1, CS2, and early algorithms courses, leaving with classroom-ready activities, implementation guidance, and materials that can be easily adapted to their own institutional contexts.

Students enter CS1 courses with widely varying programming experience, mathematical maturity, and learning preferences. We propose a framework called PathMX that addresses this challenge through a ``curriculum as code'' methodology that treats curriculum materials as structured, version-controlled repositories editable by both instructors and LLM-based coding agents. When curriculum is authored as code following clear conventions, agents can generate personalized content on demand, enabling scalable differentiated instruction that was previously impractical. Our CS1 case study demonstrates that this approach makes true differentiated instruction achievable within the constraints of real classroom teaching.

In today’s digital world, understanding how websites function is

essential, yet the inner workings of web applications remain unclear to many

learners. While web technologies are widely used, practical opportunities to

explore client-side scripting, dynamic content, and user interface design in a

cohesive project are often limited. There is a need for interactive tools that

both engage users and illustrate the integration of core web development

concepts. This paper presents the proposed tool, Apache Sudoku, a webbased game designed to demonstrate and teach web development principles.

The application integrates multiple programming languages, including

HTML, CSS, and JavaScript, to create an interactive and engaging user

experience. Apache Sudoku illustrates fundamental concepts such as DOM

manipulation, event handling, and dynamic content updates, while

providing a complete, playable game. Beyond its entertainment value, the

tool serves as a practical platform for learning how different web

technologies interact to produce a functional application. By combining

educational purpose with interactive design, Apache Sudoku provides a

hands-on example for students to explore and understand modern web

development techniques.

The rise of generative AI necessitates faculty development to equip diverse student populations with essential AI skills and knowledge. This paper details the findings of an AI Across the Curriculum (AIAC) initiative at an urban community college, a Hispanic Serving Institution (HSI). The project trained faculty in both STEM and non-STEM disciplines on designing and implementing formal AI learning activities. Qualitative data was collected from 13 focus groups, involving 157 students in Spring 2025 courses taught by AIAC faculty. Students consistently reported the activities were highly instructive, moving them beyond superficial tool use to a nuanced understanding of generative AI, particularly concerning AI ethics and societal implications. These learning experiences fostered increased comfort and confidence with AI technology, enhancing their ability to evaluate information reliability and apply AI in future studies and careers. The results demonstrate the potential of this interdisciplinary approach to increase AI literacy and reduce intimidation barriers across a diverse student body. The AIAC model serves as a practical, adaptable template for Computer Science faculty to collaborate with colleagues, successfully expanding AI learning experiences for all students.

Algorithmic Bias is the result of algorithms that have systematic errors that produce biased (unfair or discriminatory) results. Understanding Algorithmic Bias is of vital importance to both Computer Science majors and non-majors. Computer Science majors are poised to become the developers of systems that can be susceptible to bias, while non-majors (as well as majors) live in a world where many of their day to day interactions are mediated by machine learning applications that can exhibit bias. Aspects of our lives that are impacted by these algorithms include social media, marketing campaigns, government interactions, credit ratings, college admissions, and many more. Understanding the causes of algorithmic bias and the preventative measures necessary to produce systems that are equitable and fair, is an important ethical issue for future programming professionals.

Although teaching ethical computing practices is essential for developing responsible technologists, ethics is still too frequently overlooked in computing education. As a consequence, students may graduate not only without the skills to identify and address ethical dilemmas, but also without an understanding of why ethical decision-making is essential, in the first place. To overcome this gap, we have created a set of easily adoptable, experiential ethics-focused [hidden] designed to systematically introduce students to core ethical concepts in computing and to emphasize the real-world consequences and importance of ethical computing practices. In the following paper, we report on instructor observations regarding the inclusion of these ethics-focused computing labs at a diverse set of categorically distinct partner institutions. The complete project materials are openly available on our website: [hidden]

In this study, we analyze the effectiveness of generative AI in diagnosing student misconceptions in an undergraduate operating systems class. Notably, we ask students to assess whether an LLM's feedback correctly diagnosed what mistaken belief was the proximate cause of their errors. We tested 3 models: ChatGPT, Claude, and Gemini. Gemini was the most consistent in terms of perceived student correctness. ChatGPT received the highest student ratings. In a qualitative assessment, the LLMs were correct 42 percent of the time in diagnosing the students' misconceptions. In a quantitative assessment, 63.5 percent of LLM responses were perceived as exactly or almost exactly correct. We conclude that LLMs can provide valuable formative feedback but are not yet ready to be the sole source of insights.

As data-driven decision making becomes increasingly vital across industries, skills in online data visualization and analysis are in high demand. While many colleges now offer coursework in data analytics and visualization, providing earlier access to these concepts can better prepare motivated high school students for advanced study and early internship opportunities in a competitive job market.

This study presents the design and implementation of a four-week summer curriculum in data visualization and analysis tailored for high school students. The program meets four days per week for six hours per day, combining three hours of lecture with three hours of hands-on lab work. The curriculum focuses on three-tier architecture for data visualization design, development, and analysis. No prior programming experience is required; however, students are expected to have completed Algebra and to demonstrate strong interest in computing.

The learning outcomes of the curriculum are for students to:

• understand the purpose and impact of data visualization and analysis;

• develop online data visualization and analysis applications;

• create interactive visualizations that support user engagement; and

• apply storytelling techniques to communicate insights and patterns.

Students learn the fundamentals of Structured Query Language (SQL) to query and aggregate data within a MySQL database; PHP and MySQL functions for application-layer development; and JavaScript, event handling, and the Google Charts Library for presentation-layer design. Instruction covers core chart types—including line, pie, bar, histogram, and scatter plots—and guides students through retrieving data from a database and visualizing it effectively. Analytical components introduce students to interpreting distributions, identifying correlations and outliers, and validating findings with AI-assisted tools such as ChatGPT.

Overall, this accelerated curriculum provides high-achieving high school students with foundational skills in computer science, data science, and AI-related fields, positioning them for early success in college-level coursework, internships, and future career pathways.

As part of a NSF S-STEM grant, Elms College has piloted requiring students who are STEM scholars to take two cohort-based interdisciplinary Data Science courses to strengthen student skills and student engagement across STEM disciplines. First-year STEM scholars across the majors of Biology, Biotechnology, Chemistry, Computer Science, Comp. Info. Tech. and Security, and Data Science & AI, enroll as a cohort in a newly developed course, Computational Statistics, which introduces statistics, data analysis, and programming using R with example data across the sciences and meets a math core requirement. And transfer STEM scholars and scholars in their junior year are required to take a Data Analytics & Visualization course in Python in Computer Science which includes a final project where students choose to work with real-world data in their disciplines. These cohort-based courses prepare students for advanced research in STEM fields that require students to apply data analysis tools to real-world scientific questions. Embedding data science across disciplines has helped students develop transferable skills in quantitative reasoning, computational skills, and data analysis. The shared learning experiences promote peer support and sustained engagement, particularly for students entering with diverse academic backgrounds. Since 2022, 47 STEM scholars have participated in these cohort-based data science courses, with 26 currently active students, providing multiple cohorts for observing patterns of engagement and skill development. In addition to positive student feedback and increased participation in research, internships, or advanced courses, the initiative has contributed to growth in the institution’s Data Analytics minor, Computer Science, and Data Science and AI major. This work highlights how interdisciplinary, cohort-based data science instruction can serve as a powerful mechanism for enhancing student learning and engagement across STEM programs.

While teaching discrete mathematics using open education materials, I observed that many students needed additional practice with immediate, targeted feedback. Standard non-interactive practice problem sets suffer from the limitation that they must either be graded by the instructor or distributed with solutions so that the student can check their own work. Instructor-graded problems necessarily limit the time that a student can usefully spend practicing. Problem sets distributed with solutions present a temptation to the student to read the answers rather than attempt the problems independently and limit the amount of feedback a student can receive per problem, since they are typically not able to detect a mistake at any step without viewing the complete solution.

I developed a web application that provides scaffolded practice in constructing truth tables for arbitrary logical statements, and optionally for using the resulting truth table to answer questions on equivalence and argument-form validity. I also created a series of procedural question generators, allowing students to encounter a wide range of distinct problems.

When using the application, a student is first asked to indicate all logical operators in a given statement. Once this task is completed, the application highlights each operator in turn, asking the student to indicate all symbols (whether logical operators or statement variables) belonging to sub-statements of the indicated operator. After successfully processing all operators, the student is asked to choose an order in which to evaluate the sub-statements when constructing the truth table. If a valid order is specified, the student is provided with a blank truth table to fill in, with the columns in their chosen order. For logical equivalence and argument-form validity questions, the student is further asked to identify the columns and, for argument questions, rows needed to answer the final question. This structure allows feedback to be immediately provided at each intermediate stage of reasoning, rather than only on final answers.

For each student, the application asks a series of questions of escalating complexity before switching to a free-practice mode once the student has successfully completed their question series. Because questions are generated procedurally, each student will continue to receive fresh questions during free-practice, though since questions are of finite size duplicates may be encountered as well during free-practice.

In this poster, I present my experiences designing the software and using it during the Fall 2025 semester and the first part of the Spring 2026 semester, including observations about how the application supported more flexible and sustained student engagement with discrete mathematics practice.

Our [hidden] initiative helps participants learn how to design ethical software while introducing them to essential concepts in Artificial Intelligence and Machine Learning (AI/ML). Several of these developed labs focus on foundational ethics-focused topics. These experiential interactive modules highlight why ethics in AI/ML matters and provide practical experiences that reveal the diverse ways ethics-focused topics can influence modern systems. The tutorial is suitable for a broad audience within the software engineering community—from students to seasoned professionals—who wish to better understand ethical implications across domains and ensure that the software they develop is ethical and fair. Complete project materials are openly available on our website: [hidden]

This paper examines how Generative AI coding tools impact computer science education and describes instructional experiments conducted across several courses that inform the design of a new Special Topics course, Software Development in the GenAI Era. The newly designed course introduces students to the evolving landscape of AI-assisted development through five modules: The Stochastic Engine, Prompt & Context Engineering, Vibe Coding, Agentic Coding, and Specification-Driven Development. Drawing from classroom experiences, the paper highlights how students respond to AI tools, where they benefit most, and where they struggle with the verification and documentation practices required for responsible AI use. The goal is to offer a practical, evidence-based framework for preparing students to become architects and supervisors of AI-driven software systems, equipped with the foundational CS knowledge and modern AI-native skills needed in today’s rapidly evolving development environment.

The era of GenAI began with the release of ChatGPT 3.5 in November 2022. GenAI systems that can generate passable code for undergraduate-level programming-focused courses require a shift in faculty mindset in what is taught in programming-focused courses and in how student learning is assessed. Rather than trying to create assignments that the current GenAI systems cannot easily generate code for, this paper suggests approaching the GenAI era by updating assessments of student learning and incorporating GenAI into the coursework. Small class sizes at small colleges provide an excellent environment for experimenting with new pedagogical approaches.

This paper presents a digital learning framework, coLearn-AI (hereafter referred to as the system), that combines guided inquiry pedagogy with intelligent, adaptive feedback mechanisms to support collaborative learning in introductory programming courses. The system enables instructors to author structured inquiry activities, manage student groups, and provide embedded AI guidance that scaffolds learning in real time. Students work in groups and engage with progressively complex problems through exploration, collaboration, and reflection. The system integrates synchronous collaborative workspaces, turn-taking mechanisms, and analytics dashboards that help instructors monitor student progress at scale. A key innovation is the integration of AI-driven formative feedback, where instructors embed AI guidance directives within activity files. These directives, visible only to instructors, enable the AI module to provide context-aware feedback that aligns with pedagogical goals. By bridging structured human facilitation with adaptive, intelligent tutoring capabilities, the system promotes engagement, accountability, and deeper conceptual understanding. This paper describes the pedagogical foundations, system architecture, activity model, AI feedback pipeline, and group-learning mechanisms and outlines directions for formal efficiency evaluation and refinement.

As artificial intelligence(AI) becomes increasingly visible in everyday life, many students are experiencing growing anxiety about what AI is and how it affects them. This signals the urgent need for meaningful AI literacy education that helps learners understand how AI systems work and empowers them to use these tools confidently and effectively. Recognizing the importance of ensuring that all learners can develop this understanding, this paper introduces a learner-centered AI literacy lesson designed for 6-8 grade students. The lesson envisions AI literacy through a learner-centered lens, prioritizing accessibility, agency, and inclusivity. Grounded in the AI4K12’s “Big Five Ideas in AI” and informed by learning science frameworks such as Universal Design for Learning (UDL), Funds of Knowledge, and Depth of Knowledge, the proposed lesson supports students to learn AI in ways that connect to their experiences, questions, and capabilities.

Eye-Assist is an AI visual navigation system that converts camera input into concise, context-aware audio feedback for people with visual impairments. It fills gaps in current tools by combining real-time object detection, distance measurement, and an advanced priority algorithm so users can move safely, read text, and understand their surroundings. Eye-Assist uses a YOLOv8-based TensorFlow Lite model trained on custom day and night datasets, an Intel RealSense D435i depth camera for distance and motion cues, and a scoring algorithm that ranks nearby objects before generating spoken guidance. The demo will showcase real-time navigation alerts, a Read Mode for signs and documents using on-device OCR, and an Explain Surroundings feature triggered by voice commands, running on Android and in a Raspberry Pi 5 hybrid setup.

In classroom and online learning environments, identifying which students need help at any moment is challenging. Students often enter a state of ``spinning,'' continuing to work without making progress, and would benefit from timely intervention. We are developing a real-time system to detect spinning using behavioral patterns from students' programming editors. We collected fine-grained, often keystroke-level data from a Massively Open Online Course (MOOC) programming environment. In the first phase, we focus on assignments with known correct solutions, developing tools to measure students' distance from the goal using Levenshtein and AST edit distances, revealing proximity to or struggle toward the correct answer. By segmenting work into active sessions, we map progress over time across 28,000 students and 70 exercises, revealing improvement, divergence, and sustained effort without progress. We find that spinning often involves ups and downs rather than stagnation. Behavioral features extracted from these episodes will train a machine learning model in Phase II to detect spinning when solutions are unknown, enabling smarter, more responsive learning tools for online and classroom orchestration. We present our analytical approach, findings on student behavior patterns, and hypotheses for future work.

Students frequently choose correct answers for incorrect reasons, leaving traditional formative assessments unable to surface the misconceptions that matter most for learning. We introduce Structured Post-Evaluation Interviews and Remediation (SPEIR), a workflow that extends two-tier Justified Multiple-Choice Questions (JMCQs) with guided discussions and targeted recovery opportunities to surface and address those hidden misunderstandings. Implemented across ten course sections and compared with a traditional MCQ control, SPEIR showed that correctness alone substantially overestimates understanding, while per-question analyses revealed higher rates of fully correct reasoning in SPEIR sections. Students who completed recovery quizzes demonstrated notable gains, and instructors reported that SPEIR enabled efficient, focused feedback. These results suggest that SPEIR is a scalable approach for integrating diagnostic assessment with timely remediation.

This paper is a case-study on utilizing learning analytics to evaluate the success of integrating Smart Learning Content (SLC) into a project-based undergraduate Python programming course delivered through an online learning platform. We showcase how integration of the logging capabilities of SLC with the platform enabled us to study students' engagement with the SLC activities, and examine connection between the SLC usage and student learning outcomes. There are large publicly available repositories of SLC materials, yet their integration into a specific course and evaluation of its success are often not straightforward. By providing access to several SLC types, such as program construction examples, code animations, and parsons puzzles, we aimed to bridge the gap between static conceptual reading and programming projects already present in the course. To evaluate the integration of SLC, we released the augmented course in the Spring 2025 semester to 252 students at 20 community colleges across the US and analyzed collected interaction logs.

Foundational STEM areas—including artificial intelligence, cybersecurity, and inclusive software development—are often missing from 9–12 computing curricula. A major reason for this lack of inclusion is the shortage of high-quality, easy-to-use teaching materials. This gap is especially challenging for smaller or under-resourced schools that disproportionately serve underrepresented students. As a result, many 9-12 classrooms are unable to cover these essential topics, contributing to a broader ``diversity crisis'' in the U.S. STEM pipeline.

The [hidden] are a collection of 15 experiential, computing-focused educational modules originally developed and validated for undergraduate learners. Designed to promote hands-on engagement with computing-focused topics such as machine learning, artificial intelligence, inclusive software development and ethics-focused topics, these modules have demonstrated strong effectiveness in building foundational technical skills and increasing student confidence. This paper presents the adaptation of the [hidden] modules for high school students in grades 9–12, highlighting modifications to content, scaffolding, and delivery to ensure developmental appropriateness and curricular alignment with secondary education standards. We describe the pedagogical framework guiding the redesign, pilot implementations strategies, and provide recommendations for integrating the modules into existing computer science pathways. By offering flexible, accessible, and engaging learning experiences, the adapted [hidden] modules aim to expand early exposure to inclusive computing practices and support a more diverse pipeline of future technologists. Complete project materials are openly available on our website: [hidden]

This experience report describes a six-week Collaborative Online In-

ternational Learning (COIL) module integrated into an undergraduate

Software Engineering course, connecting students from the United States

and Panama. The primary aim was to enhance student engagement and

motivation through cross-cultural, project-based collaboration.

Students collaborated to complete software projects, including a stand-

out game featuring campus-native animals symbolizing the partnership.

Survey data from U.S. participants indicate increased motivation, im-

proved project completion rates, and heightened interest in global col-

laboration.

Challenges such as time zone coordination and language barriers

emerged but were effectively mitigated through the universal language

of programming. The experience also fostered the development of soft

skills, including enhanced communication and intercultural competence.

This report offers insights into the pedagogical benefits, technological

tools employed, and lessons learned.