About

Margaret Ellis is a Professor of Practice in the Department of Computer Science at Virginia Tech, located in Torgersen Hall, RM 2230A, Blacksburg, VA. She holds a Master of Science degree in computer science and a Bachelor of Science degree in mathematics, both from Virginia Tech. Her research interests focus on computer science education. She is involved in various locations including the Blacksburg campus, the Institute for Advanced Computing in Alexandria, VA, and the Virginia Tech Research Center in Arlington, VA. Her contact email is maellis1@vt.edu, and her office phone number is (540) 231-9560.

Research topics

• Computer Science
• Artificial Intelligence
• Psychology
• Engineering
• Mathematics education
• Engineering management
• Pedagogy
• Knowledge management
• Operations management
• Demographic economics
• Software engineering
• Geography
• Marketing
• Business

Selected publications

• VISTA: Virtual Interactive Simulation for Teaching Assistants

  2026-02-13

  articleOpen access
  PublisherDOI

• A Multi-Institutional Study on Peer Instruction: Evaluating Text-Chat with Assigned Group Members vs Verbal Discussion

  2026-02-13

  articleOpen access

  In Peer Instruction (PI) an instructor displays a challenging multiple-choice question during lecture that students answer individually, discuss verbally with nearby peers, answer individually again, and finally, the instructor leads a discussion of the question. Peer Instruction typically increases student learning and motivation over traditional lecture. We added a text-chat mode to improve PI for remote synchronous learning. This feature assigns students to discussion groups to maximize the number of groups that have members with different answers. The tool was pilot tested in Winter 2022 and revised. In Fall 2022 and Winter 2023, it was tested at one institution. In Fall 2024, it was tested at four institutions. We conducted a log file analysis of student data from 1394 students and analyzed surveys with 848 student responses. We found that questions answered using the text-chat had a significantly higher improvement than those using traditional verbal discussion, although the two modes were tested with different questions. Interestingly, most of the students preferred to discuss the question verbally, although some preferred the text-chat discussion. These results inform efforts to improve the effectiveness of Peer Instruction and increase its adoption.

  PublisherDOI

• Increase Student Engagement and Learning using a Free Tool for Peer Instruction

  2026-02-13

  articleOpen access

  Decades of research on Peer Instruction (PI) have shown that it improves student engagement, retention, and learning in many fields, including computing. In PI, as defined by Eric Mazur of Harvard, the instructor displays a difficult multiple-choice question during lecture which students first answer individually, then discuss with nearby peers, and then answer individually again. We created a free tool, Peer+, for Peer Instruction and tested it over several semesters at multiple institutions (University of Michigan, Duke University, Virginia Tech, and Berea College) on over 1,000 students. Peer+ is freely available in the open-source ebook platform Runestone Academy. Instructors can use any of the 90+ free ebooks on that platform or use an empty ebook. There are ebooks for Advanced Placement Computer Science, CS1 (Python, Java, and C++), CS2 in C++ and Java, and Discrete Math. We added hundreds of existing PI questions to these ebooks, which instructors can use or they can create their own multiple-choice questions. In this tutorial you will learn best practices for Peer Instruction and gain hands-on experience creating PI assignments, browsing existing PI questions, writing new PI questions, and grading PI assignments. You will also experience PI sessions using three different modes for peer discussion: traditional verbal discussion with nearby peers, text-chat with assigned group members, and text-chat with an LLM for students answering questions after lecture.

  PublisherDOI

• Demystify, Use, Reflect: Preparing students to be informed LLM-users

  2026-02-13

  articleOpen access

  We updated our post-CS1 course that introduces various subfields of computer science so that it integrates Large Language Models (LLMs). The course now includes explicit instruction on how LLMs work, exposure to current AI tools and ethical issues, and student re- flection on personal use of LLMs and the larger evolving landscape of AI. We demonstrate the use and verification of LLM outputs, guide students in the use of LLMs within a larger problem-solving loop, and require disclosure of the nature and extent of LLM as- sistance. Throughout the course, we discuss risks and benefits of LLMs across CS subfields. In our first iteration of the course, we collected and analyzed data from students' pre and post surveys. Student understanding of how LLMs work became more technical, and their verification and use of LLMs shifted to be more discerning and collaborative. These strategies can be used in other courses to prepare students for the AI-integrated future.

  PublisherDOI

• There is an App for That: Moving a Large Introductory Computer Science Course to an Automated Course Delivery System

  2025-04-22

  articleSenior author

  Computer science as a discipline is well-placed to research and adopt new pedagogical technologies as they are developed. Many popular software platforms that have been adopted across institutions have started as research projects in computer science programs. Additionally, because of the growth in demand for computer science as a program of study, CS instructors have been on the forefront of adopting new approaches and technologies to be able to meet the requirements of teaching at scale. However, being at this intersection creates challenges of its own. There are hundreds of tools and approaches, many with some level of research behind them. These tools come from vendors large and small, with costs both large and small, and with a wide variety of features that may or may not meet the needs of the instructor and the approach they want to pursue. From infinite diversity comes infinite combinations, with the attendant number of questions that instructors must answer when designing their courses. What tools do we use? How do we pick? How can we integrate them? Can we get support? Will my institution approve them? How can I adopt a technique like peer instruction in a class of 600 ? Will the tools I pick support 50,000 assignment submissions per term? How do I manage change? In this paper we will report our experience redesigning our CS 2 course from a mix of disparate tools to a fully-automated, integrated course delivery system. We combined multiple teaching platforms into a unified design, that is delivered to the student as a coherent whole, rather than piecemeal. We will discuss the design decisions we made, and some we didn't, and why. We will discuss our approach, the challenges we faced, and also provide instructors with a set of “lessons learned”, questions they should ask themselves and their colleagues as they embark on any course redesign.

  PublisherDOI

• MARCO: Multi-Agent Code Optimization with Real-Time Knowledge Integration for High-Performance Computing

  ArXiv.org · 2025-05-06

  preprintOpen access

  Large language models (LLMs) have transformed software development through code generation capabilities, yet their effectiveness for high-performance computing (HPC) remains limited. HPC code requires specialized optimizations for parallelism, memory efficiency, and architecture-specific considerations that general-purpose LLMs often overlook. We present MARCO (Multi-Agent Reactive Code Optimizer), a novel framework that enhances LLM-generated code for HPC through a specialized multi-agent architecture. MARCO employs separate agents for code generation and performance evaluation, connected by a feedback loop that progressively refines optimizations. A key innovation is MARCO's web-search component that retrieves real-time optimization techniques from recent conference proceedings and research publications, bridging the knowledge gap in pre-trained LLMs. Our extensive evaluation on the LeetCode 75 problem set demonstrates that MARCO achieves a 14.6\% average runtime reduction compared to Claude 3.5 Sonnet alone, while the integration of the web-search component yields a 30.9\% performance improvement over the base MARCO system. These results highlight the potential of multi-agent systems to address the specialized requirements of high-performance code generation, offering a cost-effective alternative to domain-specific model fine-tuning.

  PublisherOA PDFDOI

• Demystify, Use, Reflect: Preparing students to be informed LLM-users

  ArXiv.org · 2025-11-14

  preprintOpen access

  We transitioned our post-CS1 course that introduces various subfields of computer science so that it integrates Large Language Models (LLMs) in a structured, critical, and practical manner. It aims to help students develop the skills needed to engage meaningfully and responsibly with AI. The course now includes explicit instruction on how LLMs work, exposure to current tools, ethical issues, and activities that encourage student reflection on personal use of LLMs as well as the larger evolving landscape of AI-assisted programming. In class, we demonstrate the use and verification of LLM outputs, guide students in the use of LLMs as an ingredient in a larger problem-solving loop, and require students to disclose and acknowledge the nature and extent of LLM assistance. Throughout the course, we discuss risks and benefits of LLMs across CS subfields. In our first iteration of the course, we collected and analyzed data from students pre and post surveys. Student understanding of how LLMs work became more technical, and their verification and use of LLMs shifted to be more discerning and collaborative. These strategies can be used in other courses to prepare students for the AI-integrated future.

  PublisherOA PDFDOI

• Assessing the motivational climate in a university course: Considerations for motivating all students

  Journal of University Teaching and Learning Practice · 2025-09-16

  articleOpen access

  The motivational climate in a course can affect students’ motivation, engagement, and performance in the course. To examine if the motivational climate in a computer science course varied by gender or students’ high school experiences, we assessed motivational climate across 4 time points and compared students’ responses by gender and experience in computer science. The purpose of this article is to present the results of this mixed-methods survey study and to provide an example of how instructors can assess motivational climate to investigate potential differences among groups of students. We measured motivational climate using the MUSIC Model of Motivation Inventory, which measures five motivational climate perceptions: eMpowerment, Usefulness, Success, Interest, and Caring (MUSIC is an acronym). All of these MUSIC perceptions decreased significantly across the semester, except for interest. Empowerment and success decreased across the semester for all students. Success expectancies were significantly lower for women and for those without high school computer science experience. On average, perceptions of usefulness and caring decreased across the semester for women. These findings indicate that the course design may have had a more negative effect on women than men. The instructor could use these results to address these course perceptions in ways that foster a motivational climate that is more equitable and inclusive of all students.

  PublisherDOI

• Embedded Ethics in CS: Experiences with Integrating Ethics Assignments in Sophomore, Junior, and Senior Level Courses

  2025-06-13

  articleOpen accessSenior author

  Technical and ethical aspects of Computer Science (CS) are interdependent. Many CS departments teach ethical and social implications of technology in separate standalone courses. However, prior research shows that ethical issues are better taught in tandem with their related technical content as an integral required skill in CS curricula. In this experience report, we share our experience with embedding ethics assignments in 3 CS courses at different levels: a CS2 course in software design and data structures, a CS3 course in data structures and algorithms, and a Software Engineering capstone course, all taught at Virginia Tech (a large public R1 institution) in Spring 2024. Students from the 3 courses were surveyed at the beginning and end of Spring 2024. By comparing results from the pre and post surveys, we found that the embedded assignments for the CS2 and CS3 courses improved students' confidence in their knowledge about how ethical issues may come into play in their career, their confidence in their ability to address ethical issues arising from applying technology in real contexts, and their confidence in communicating and defending their positions on how to address these issues. For all 3 courses, students gave positive feedback on how the assignments were engaging and relevant to the course, and how it improved their ability in raising, and reasoning about, ethical implications of technology. We believe that the practices and results of our experience will be helpful to other CS instructors thinking of injecting ethical content into their technical courses.

  PublisherDOI

• Connections between students’ perceptions of the motivational climate in a course and their motivational beliefs and academic and career goals

  European Journal of Engineering Education · 2024-12-28 · 1 citations

  article
  PublisherDOI

Frequent coauthors

• Clifford A. Shaffer

  Virginia Tech

  13 shared

• Brett D. Jones

  Centre for Addiction and Mental Health

  13 shared

• Derek Haqq

  Virginia Tech

  12 shared

• Molly Domino

  Virginia Tech

  11 shared

• Mohammed F. Farghally

  Virginia Tech

  10 shared

• Hamdy F. F. Mahmoud

  Virginia Tech

  7 shared

• Godmar Back

  Virginia Tech

  6 shared

• Stephen H. Edwards

  Virginia Tech

  6 shared