Schedule

We will be located all day in the main atrium of the Data and Decision Sciences building for registration and information.

Welcome to the Virginia Tech Teaching with AI: AI On/AI Off Symposium! Dr. Quinn Warnick will start us off with an overview of today's activities.

Presenter: Dr. David Wiley Questions About Using Generative AI to Help All Learners Achieve Their Potential What are the unique affordances of generative AI? What is the relationship between writing, thinking, and AI? When should we turn AI “On” and when should we turn it “Off”? What does psychometrics have to say about the way AI impacts assessment? How can AI help us more closely approximate ideal pedagogy? And how can we help all learners benefit from generative AI? Come discuss these questions and more in this interactive keynote presentation.

If the main auditorium fills to capacity, this room will be open to attend the opening keynote session.

Round 1: Tracking Student LLM Behavior Across Individual and Team Assignments (Sehrish Basir Nizamani, Zannah Ziew, and Saad Nizamani) Round 2: AI Literacy Across the Disciplines: Teaching AI in Non-Technical Courses (Sehrish Basir Nizamani, Hulya Dogan, Can Dogan, Aparna Shah, Saad Nizamani, Arianna Schuller Scott, Sana Illahe, and Traci Gardner) Round 3: DBWorkout: An AI Integrated Framework for SQL Pedagogy (Sehrish Basir Nizamani, Deepikia Devaraj, Tien Nguyen, Jaren Goldberg, Saad Nizamani, and Sally Hamouda) Round 4: AI in the Sales World (Mark Michalisin and Brian Collins) Round 1 Description: When students in CS3654 moved from individual homework to team projects, something unexpected happened - LLM usage dropped from 84% to just 40%. This talk shares findings from a survey-based study of 96 undergraduates, tracking how collaboration reshaped students' AI behaviors across 6 assignments. Results showed significant shifts in role framing, prompting strategies, and output verification - all becoming less deliberate in team settings. These findings raise practical questions for instructors: does team work naturally push AI to the sidelines, and should course design account for that? Attendees will leave with data-backed insights for thinking intentionally about AI in collaborative assignments. Round 2 Description: What does it mean to teach AI literacy in non-AI disciplines? This lightning talk shares early insights from adapting and piloting discipline-specific AI literacy modules across non-technical courses at Virginia Tech and Radford University. The approach follows a four-stage structure: helping students understand how large language models work, guiding them in applying AI tools within their discipline, encouraging critical reflection on their interactions, and evaluating changes in their understanding and behavior. Presenters from Economics, Neuroscience, and Engineering Education will take 3 minutes each to briefly share how this approach was implemented in their courses, highlighting what worked, what did not, and the challenges they encountered. These include translating technical AI concepts for non-technical audiences, varying levels of instructor familiarity with AI, and designing activities that go beyond surface-level engagement. Rather than presenting a polished success story, this talk offers a candid, cross-disciplinary snapshot of early experimentation with AI literacy in the classroom. It also contributes to a broader effort to build reusable, discipline-specific AI literacy modules that can be adopted across academia. Attendees will leave with a practical understanding of how AI literacy can be introduced in non-technical courses and what to expect when doing so. Round 3 Description: This demo presents DBWorkout, a web-based SQL practice platform deployed at Virginia Tech that integrates GPT-4o for instructor-facing task and schema generation within a deliberate human-in-the-loop workflow. DBWorkout embodies the AI On / AI Off balance directly: AI is on for instructors - generating candidate SQL tasks and database schemas from natural language prompts to reduce content creation overhead - but every piece of content is reviewed and approved before students see it. Instructors can also upload course materials such as lecture slides, PDFs, and reading documents, which the LLM parses and formats into structured learning modules displayed to students before a practice session. A session of SQL tasks follows the module, allowing students to immediately apply what they just read. Instructors can also toggle AI assistance on or off for students during live sessions, giving faculty explicit control over the role of AI in each deployment. The demo walks through the full instructor workflow: generating a schema from a natural language description, reviewing LLM-generated tasks, launching a live session, and monitoring student progress in real time. Attendees will see how automated grading provides immediate multi-dimensional feedback across seven correctness dimensions, and how recent feedback enhancements - including value-level cell diff, directional hints for common SQL error patterns, and jargon tooltips - were designed directly in response to student research findings. Attendees leave with a concrete model for intentional AI integration that keeps human judgment central, and a live link to try the platform themselves. Round 4 Description: I will be holding a lightening talk regarding how I have implemented AI into my Principles of Professional Selling and Sales Technology courses. We focus on using AI to develop sales pitches, elevator pitch, cold call scripts, role playing, discovery question, and how to prepare for a sales call anticipating buyer questions and potential objections and how to overcome said objections and how AI can help you work more efficiently in the sales word. AI is reinventing how salespeople go about their day. AI will never replace the human salesperson but it will support and make them become more efficient in their day-to-day activities

Panel Members: Louis Hickman, Dan Dunlap, Dale Pike, Lee Vinsel, Amy Allen, and Joan Watson. Generative AI is rapidly reshaping how students learn, complete assignments, and demonstrate understanding—challenging long-standing assumptions about instruction, assessment, and academic integrity. This panel brings together educators from diverse disciplinary backgrounds alongside administrators involved in developing institutional guidance for adapting instruction in response to Generative AI. Panelists will share practical approaches for adapting instruction in response to these changes, including: designing AI-integrated assignments that promote higher-order thinking; strategically using “non-AI” activities to ensure foundational knowledge and skills are gained; engaging AI itself as a domain of learning and application within disciplines; and emerging assessment strategies that emphasize process, authenticity, and critical engagement. The discussion will also address ongoing concerns about academic integrity, moving beyond detection toward proactive course design that limits the potential negative impacts of generative AI misuse and clear communication of expectations regarding AI use. Attendees will gain insight into how instructors can align their pedagogical choices with course learning objectives and assessment goals. By highlighting both challenges and opportunities, this session aims to equip participants with concrete strategies for fostering meaningful student learning in the age of generative AI.

Come have a more intimate follow-up conversation with our keynote presenter.

Demo 1: Using AI Roleplays to Build Communication Skills and Student Confidence for Challenging Conversations Demo 1 Presenter: Kathy Perkins Demo 2: Supporting Critical AI Literacy in a ‘Post-Truth’ Era Demo 2 Presenters: Katlyn Griffin & Kayla McNabb Demo 1 Description: How do we teach communication skills effectively in asynchronous online courses when students need multiple practice opportunities, a safe place to fail, and authentic resistance to build their confidence? Traditional peer roleplays often fall short: high stakes, limited iteration, and performative interactions that don't mirror real-world difficulty. This session presents an accessible AI-enhanced instructional design that transforms how students practice difficult conversations. The presenter will describe how she used freely available AI chatbots (ChatGPT, Claude) as practice partners to prepare students in the course Leadership and Communications for Environmental Sustainability to engage in difficult conversations. Participants in this session will see the 3-phase structure in action: (1) Individual roleplaying with a programmed AI platform, (2)Practice with scaffolded difficulty levels, (3) Programmed, AI-generated coaching analysis and feedback to students based on previously learned strategies and skills for communicating in polarized environments. The presenter will share design principles for creating role-play chatbots, using faculty-friendly materials that don't require coding. She will describe the anatomy of effective AI roleplay prompts, students' perceptions of the exercise, and their levels of achievement of learning outcomes. This session offers practical strategies for integrating AI as a practice tool rather than a replacement for human interaction in communication-intensive coursework. Demo 2 Description: Amid rapid change and uncertain futures, time-tested digital literacy practices build a solid foundation for fostering student AI literacy. Drawing on insights from a multi-institutional research project coordinated by Ithaka S+R, this session will explore strategies for cultivating student AI literacy, especially as they navigate truth and trust in the current information environment. Attendees will have the opportunity to practice a hands-on learning activity related to critically evaluating generative AI outputs.

Demo 1: Using Offline AI Models with Sensitive/Protected Data Presenter: Caleb Bradberry Demo 2: Gemini Notebook: Your AI Partner for Teaching and ResearchPresenter: Eman A.M. Amer Demo 1 Description: While many people are familiar with frontier AI models such as OpenAI's GPT and Anthropic's Claude, few are aware of the myriad offline and open-weight models that can be run on an individual PC. By running AI models locally, we can ensure regulatory compliance across a variety of sensitive datasets while still taking advantage of recent advances in generative pre-trained transformer technology. This workshop will introduce participants to the landscape of open-weight large language models and demonstrate how to deploy them entirely offline using freely available tools. We will cover practical considerations such as hardware requirements, model selection based on task and resource constraints, and software frameworks that simplify local deployment. Participants will see firsthand how these models can be applied to tasks like summarization, classification, and text generation, all without sending data to an external server. A key focus will be on use cases involving sensitive or protected data, including student records governed by FERPA, health information under HIPAA, and proprietary institutional data. For researchers, educators, and IT professionals working under these constraints, local AI models offer a compelling path forward: the analytical power of modern language models with none of the data-sharing risks inherent in cloud-based services. No prior experience with AI or machine learning is required. Attendees will leave with a clear understanding of what local AI can and cannot do, a working knowledge of how to get started, and concrete guidance on aligning local model use with their institution's data governance policies. Demo 2 Description: Synthesizing and managing large volumes of data remains a significant challenge in higher institutions. For educators, preparing instructional materials and staying current with pedagogical research are ongoing challenges. However, with the right tools, data synthesis and analysis can become more efficient and powerful. This workshop introduces Gemini Notebook, Google’s AI-powered research assistant, designed to help faculty, department chairs, and deans streamline the course design, create personalized study guides, and synthesize complex information into accessible student resources. First, attendees will be introduced to the key features of Gemini Notebook. Second, they will use Gemini Notebook to synthesize and extract information and generate summaries from diverse sources, in various output formats. By the end of this workshop, attendees will know how to use Gemini Notebook to identify gaps and opportunities aligned with their specific, teaching strategies, research interests and professional goals.

Presenter: Ayat Mohammed ARC offers on-prem Large Language Models (LLMs) for research, education, and administration. In this workshop we will show several ways to interact with LLMs on ARC, 1) a web interface to a selection of LLMs run by ARC, 2) API access to a selection of LLMs run by ARC, 3) a custom dedicated LLM via Open OnDemand. We also will explore some advanced tools such as advanced custom deployment via personalized Slurm scripts and the integration of VS Code with ARC’s LLMs using API keys. Outline: https://llm.arc.vt.edu https://llm-api.arc.vt.edu https://ood.arc.vt.edu vLLM GitHub Copilot Chat (integration of VS Code)

Presenter: Brett Jones Generative artificial intelligence (AI) tools are creating new opportunities and challenges for instructors as they design their courses. Much of the discussion about AI in higher education has focused on tool selection, academic integrity, efficiency, and policy. However, instructors also need a practical way to think about how AI-related decisions affect students’ motivation to engage in learning. This interactive workshop introduces the MUSIC Model of Motivation as a research-based framework that faculty can use to design courses that incorporate AI in more human-centered ways. The workshop will explain how the five MUSIC components (i.e., eMpowerment, Usefulness, Success, Interest, and Caring) can guide decisions related to AI policies, assignments, learning activities, feedback, and support structures. Rather than assuming that AI should always be embraced or restricted, the session will help participants decide when AI use is likely to support learning goals and when limits may be more appropriate. Participants will examine examples of AI-related instructional decisions and consider how those decisions can strengthen or weaken students’ motivation. They will then use a guided planning tool to apply the MUSIC framework to their own courses by identifying an AI-related teaching decision and evaluating it through the lens of the five MUSIC components. The session is designed for instructors across disciplines, including those who are “AI on,” “AI off,” or somewhere in between. Participants will leave with a practical framework for making AI-related course design decisions and a draft plan for aligning one aspect of their teaching with students’ motivation, engagement, and learning.

Led by Dale Pike, Associate Vice Provost for Technology-Enhanced Learning "AI Off" is usually said defensively, as if choosing to work without AI were a failure of nerve or a refusal to keep up. It is neither. Deciding where AI does not belong in a course requires more fluency than adopting it uncritically, because it means understanding what a tool does, where it fails, and what a particular kind of learning actually depends on. This session treats that decision as a skill and gives participants a method for making it, activity by activity.We will start with a single distinction: the difference between removing a barrier to learning and removing the learning itself. Cognitive science is clear that some difficulty is the point, and recent studies show that AI assistance which feels helpful in the moment can leave students confident but unable to perform on their own. From there, participants will work with a short diagnostic for identifying which activities in their own teaching should stay human, and why, then discuss concrete ways to preserve them, from in-class writing and oral defense to visible drafts and work anchored in a student's own experience. Participants will leave able to name specific activities they intend to keep analog and to defend that choice on pedagogical grounds. "Intentionally analog" is a design decision made from judgment, not nostalgia.

Round 1: Can I do that? Or is this allowed in class? (Katherine L. Hall) Round 2: Teaching in the Age of AI: Practical Strategies for the AI-Enabled Classroom (Warren Lucero) Round 3: Physics Students’ Use and Perceptions of AI: Before and After Explicit Instruction (Jake O'Brien, Alma Robinson, Margaret Ellis, John Simonetti, and Naren Ramakrishnan) Round 4: HokieLearn: An AI Teaching Assistant That Actually Knows Your Course (Tessema Mengistu) Round 1 Description: In this lightning talk, I will discuss how my students and I are testing the waters with Chat GPT in a first year writing course. Together in my classes, we brainstorm ways to use AI for good. We also work together to build a body of knowledge as to the ethical usage of AI as it applies to writing and research; this co-created understanding directly informs the AI usage policy that I include on my syllabus. This building of knowledge is particularly important in a first year writing course as it is often the only time that students will take an undergraduate writing course at VT. By presenting real life examples of use and student voices on the topic of ethical AI, I hope attendees will gain a snapshot into how to use and explore AI in their own classes. Round 2 Description: Generative AI reached 100 million daily active users in two months — faster than any technology in history. Whether or not faculty assigned it, students are already using it. This session argues that the most effective response is neither blanket prohibition nor uncritical adoption, but intentional, pedagogically grounded integration. Drawing on Teaching with AI: A Practical Guide to a New Era and cross-disciplinary classroom experience, this presentation proposes a tiered policy framework that gives students clear, rationale-backed guidance on when AI use is prohibited, permitted, or required. Tier 1 (AI-Free) preserves human judgment for reflective and high-stakes assessments; Tier 2 (AI-Assisted) treats AI as a starting point requiring critical revision and documentation; Tier 3 (AI-Integrated) assesses the prompting and evaluation process itself. The session addresses academic integrity not through detection software — which carries documented false-positive risks — but through assignment redesign: layered drafts, process portfolios, locally rooted prompts, and tasks that require disciplinary expertise to evaluate AI output. Three concrete examples span the humanities, STEM, and social sciences, each treating AI as a subject of inquiry rather than a shortcut. Attendees will leave with five transferable takeaways, including: how to communicate tiered policies clearly in syllabi; how to build AI literacy without replacing the thinking we aim to develop; and how transparency, rather than policing, builds trust and reduces academic misconduct. Recommended for: faculty across all disciplines seeking practical, immediately implementable strategies. Round 3 Description: With large language models (LLMs) such as ChatGPT from OpenAI becoming widely available to physics students in recent years, understanding how these tools are used and perceived is essential for designing effective physics courses. Students may use LLMs outside of class in a variety of ways, including generating additional conceptual explanations, summarizing course materials, or obtaining extra practice problems. LLMs may also be used to support problem solving and assignment completion by generating solutions, providing guidance or hints, or checking work and reasoning. Students turn to LLMs for many reasons. Some believe these tools improve their understanding of course content, although this may not always be the case. Others may feel more comfortable asking an LLM for help than approaching an instructor, find LLMs more convenient, experience time pressure when completing assignments, or perceive LLM generated explanations as clearer than those provided by instructors. To better understand students’ behaviors and perceptions, we engaged students in a problem solving activity focused on using LLMs for learning and then provided explicit instruction on how LLMs function, along with discussion of existing research on their impact on learning. In this lightning talk, we highlight our classroom experiences and results from anonymous pre and post surveys. Round 4 Description: Imagine every student in your class having a personal teaching assistant, the one available at 11 PM before a project deadline, that never gets tired of the same question, and that answers based on your course, not the entire Internet. That's HokieLearn. HokieLearn is an AI-powered Q&A system that acts as a course-specific teaching assistant. HokieLearn uses Retrieval-Augmented Generation(RAG) and integrates directly with Canvas. It pulls in your syllabus, lecture notes, assignments, and resources to build a course-specific knowledge base. When a student asks a question, whether logistical, "When is project 2 due?", or technical, "How does this algorithm work?", HokieLearn retrieves the relevant course content and generates a personalized, accurate response in seconds. The impact is twofold. Students get instant, reliable, around-the-clock support grounded in their actual course. Instructors and TAs are freed from answering the same repetitive questions, so they can focus on the teaching that matters most.

Round 1: Too Long; Didn't Read (TL;DR): An Assignment That Helps Students Use AI to Create, Not to Cheat (Zhuofan Li) Round 2: Designing Learning Objectives and Assessment Strategies with Custom Gemini Gems (Eman A.M. Amer) Round 3: Guardrails by Design: What Health Education Teaches Us About Intentional AI Integration (Ramandeep Kaur) Round 4: From Foundations to Application: Teaching AI Literacy in Engineering (Hulya Dogan and Mags Blackie) Round 1 Description: How do we design assignments that harness the power of AI tools without letting them do the thinking for students? How do we prepare them for an economy where employers want AI skills but clients won't pay for formulaic AI-generated work? This lightning talk introduces the Too Long; Didn't Read (TL;DR), the capstone assignment of Introductory Sociology (SOC 1004), a Pathway course that attracts students across the University, to address both challenges. As AI makes it easier than ever for anyone to collect and produce information more than our attention span can ever handle, the ability to distill and communicate for a specific audience becomes the key skill that sets people apart. For their TL;DR projects, students choose a sociological monograph based on their own interests and distill its key insights in a format that their target audience actually wants to consume — a podcast episode, a YouTube video, a TikTok clip, an infographic, a blog post, you name it. The assignment also works hand in hand with in-class activities to encourage creative and responsible AI use while safeguarding against fully AI-generated output: students must critically evaluate their use of AI tools, workshop and defend their design choices in groups, and present their final products to peer evaluators in class. In this talk, I will showcase student works and walk through the assignment design, AI policy, as well as what I've learned from two semesters of implementation for people looking for ways to integrate AI into experiential learning. Round 2 Description: While previous research has noted the negative impact of AI in the classroom, AI models are now evolving from conversational agents to cognitive partners. This session will present a customized Gem that transforms GenAI into a Socratic coach. The Gem uses the AMERH framework to help instructors develop learning objectives, assessment strategies, and align with pedagogical frameworks such as Universal Design for Learning (UDL), rather than providing a finished plan. Participants will receive a practical guide to creating subject-specific Gems, enabling them to interact with AI and receive individualized feedback for their custom plans. This approach facilitates the development of objectives and assessments, reducing instructors’ workload and increasing productivity. Round 3 Description: As AI tools proliferate across higher education, faculty face a deceptively simple question: when should AI be allowed in, and when should it be kept out? Health education offers a uniquely instructive answer. In clinical and medical training contexts, the stakes of AI getting it wrong are high enough that "responsible integration" cannot be an afterthought. It has to be built in from the start. This talk draws on firsthand experience designing and launching an AI-powered learning companion for medical students, built on verified clinical sources and deliberately scoped to support learning without replacing clinical reasoning. The design process required making explicit decisions about what the AI should do, what it should refuse to do, and where human judgment must remain sovereign. Those decisions, it turns out, are exactly the right questions for any faculty member designing AI policy for their course. Attendees will come away with a practical framework for thinking about AI integration not as an on/off switch, but as a set of intentional design choices: what the AI is grounded in, what it is permitted to generate, and where the human must remain in the loop. Round 4 Description: ENG 1984: AI Literacy for Engineers is a 1-credit, 15-week introductory course at Virginia Tech’s College of Engineering designed to build foundational and applied AI literacy for first-year engineering students. The course is structured in two phases that progressively develop both understanding and practice. In the first eight weeks, students focus on AI literacy foundations. They explore how large language models work, how to design effective prompts, and how to critically evaluate and verify AI-generated outputs. They also examine ethical considerations and academic integrity in AI-assisted work. To support learning, each week follows a consistent instructional cycle of Prepare, Practice, Reflect, and Checkpoint, ensuring that students build confidence and competence in a structured and iterative way. The second phase shifts from understanding to application. Each student selects one of Virginia Tech’s 14 engineering disciplines and completes a small AI-assisted project within that context. Students use AI tools for tasks such as literature exploration, ideation, analysis, and technical communication. Throughout the process, they maintain a Prompt Journal documenting their interactions with AI and reflecting on decision-making, limitations, and responsible use. This lightning talk introduces the course design, with emphasis on the scaffolding strategies that support novice learners and the integration of reflection-based assessment. The goal is to demonstrate a replicable model for embedding AI literacy into early engineering education while cultivating critical, discipline-specific AI use.

Demo 1 Presenters: Onur Seref & Jim Dickhans Demo 1 Description: Most current approaches to AI in teaching treat student-AI interactions as something to detect, reconstruct after the fact, or prohibit. This demo proposes a different frame: treat the student's AI thread itself as the primary artifact of learning. A thread captures what students actually did, when they pushed back, when they accepted, when they noticed drift, when they missed it, in a resolution no post-hoc reflection can match. We will demonstrate a working four-stage pipeline built as a general framework for AI-centered pedagogy. The pipeline is course-agnostic: block vocabulary, assessment logic, and feedback delivery are designed to be customized for any course that includes AI interaction as part of the learning. The freshman-level AI literacy course offered by the Pamplin College of Business serves as the showcase implementation, in which the framework has been validated across multiple lectures through empirically calibrated assessments. Stage 1 parses a lecture source file into student-facing and instructor-facing HTML with embedded prompt blocks, copy-to-clipboard buttons, and rubric envelopes. Students run the thread on an AI platform and submit it. Stage 2 captures the thread. Stage 3 parses student contributions and generates detailed assessments aligned to pre-specified rubrics via a language model API. Stage 4 delivers feedback to Canvas. The demo will run all four stages live on a sample lecture, with a simulated student thread illustrating how block-level pedagogical design materializes as an assessable interaction. The framework generalizes to any discipline where AI-integrated learning warrants a structured record of student engagement.

Presenters: Dan Yaffe, Carl Harris, and Chris Bateson HokieAI (a branded version of CloudForce's NebulaOne) is Virginia Tech's newest generative AI platform approved for use by faculty, staff, and students. Get started navigating the HokieAI interface in this hands-on session.

Presenters: Amy Allen & David Hicks Generative AI has unsettled long-standing assessment practices in higher education, particularly in disciplines that rely heavily on writing. Faculty across the humanities report widespread student use of AI for essays and written assignments, alongside growing concern about its impact on critical thinking, originality, and academic integrity. Yet emerging research suggests that the problem is not simply AI itself, but what it reveals. Generative AI exposes longstanding weaknesses in assessment design, including overreliance on product-based tasks that are easily replicated by automated systems. At the same time, studies indicate that faculty are often unable to reliably detect AI-generated work, and that commonly proposed solutions such as “authentic” assessments or disclosure policies offer limited protection. This interactive session invites participants to confront this moment directly. We begin by briefly mapping the current landscape of AI and assessment, highlighting key tensions and misconceptions. We then share examples of how we have redesigned our own assessments to emphasize process, disciplinary thinking, and student interaction with AI as a learning partner rather than a shortcut, including the use of AI Awareness Statements. The second half of the session shifts to collaborative design. Participants will consider an existing assignment and work in small groups to rethink it in light of current realities. Facilitators will support this process by offering concrete strategies and feedback. Rather than offering quick fixes, this session positions assessment as a site of struggle and uncertainty that requires critical reflection and creative redesign (or a blank sheet of paper!) in response to AI.

Presenters: Elijah Carter, Mags Blackie, and Estrella Johnson This workshop uses a backward course design framework to promote a thoughtful approach to allowing or restricting student use of generative AI tools. Participants will begin by identifying or refining clear, measurable learning objectives for a course or module. From there, they will examine how different types of assessments align with those objectives and consider where AI use may either support or hinder student learning. A central focus of the workshop is helping faculty move beyond blanket policies and toward intentional, context-specific decisions about AI. Through guided activities, participants will analyze common assignment types (e.g., essays, problem sets, presentations) and evaluate whether AI use in each case reduces unnecessary barriers (such as language proficiency, access to feedback, or writing support access) or undermines essential skill development (such as critical thinking, synthesis, or disciplinary practice). Facilitators will provide practical examples of “AI-supported,” and “AI-restricted” assignment designs, along with sample policy language that can be adapted across disciplines. The workshop assumes no prior expertise with AI tools and emphasizes accessibility and practical application. Participants will leave with a draft alignment map connecting their course objectives, assessments, and AI-use guidelines, as well as strategies for communicating expectations clearly to students. By centering pedagogical goals rather than technology itself, this session equips faculty to make informed, flexible decisions that support student learning in an evolving technological landscape.

Round 1: Balancing AI On and Off: Equity and Literacy for AI-Enabled Futures (Shalaka Khot) Round 2: Using AI-generated Images to Produce Visual Scenarios to Support Learning (Philip Romero-Masters) Round 3: AI as Design Partner (Brad Whitney) Round 4: Lessons learned from practice-based AI incorporated teaching experiments: : Addressing foundational questions about AI use in the classroom (Ghazal Saeidfar) Round 1 Description: As generative AI (GenAI) becomes increasingly embedded in teaching and learning, instructors face a central challenge: how to balance the benefits of AI with the need to preserve authentic student learning. Even within a single course, students bring varying levels of technological familiarity, language confidence, and ability to effectively engage with GenAI. This presentation introduces a three-layer framework of student inequity: gaps in technological literacy, language proficiency, and AI literacy. This includes technological literacy (ability to navigate and use digital tools), language literacy (ability to understand and express ideas in academic contexts), and AI literacy (ability to iteratively use and critically evaluate GenAI outputs). Drawing on instructional practice and observations from an online undergraduate International Business course, I demonstrate how students engage with GenAI unevenly—some using it productively to support learning, while others struggle or rely on it uncritically. To address this, I present a practical AI On / Off framework. AI is “on” during learning and preparation, supported through guided prompting and iterative drafting. AI is intentionally “off” during assessment moments that require independent thinking, such as presentations and reflective tasks, while preserving creativity and authentic student voice. This approach highlights a key insight: GenAI does not automatically reduce inequities—intentional instructional design does. Attendees will leave with concrete strategies for structuring AI-supported learning and designing assessments that preserve authentic student thinking. Round 2 Description: Experiential learning is key to undergraduate development, but it isn’t logically possible to have experimental learning in every class. AI-generated visual scenarios represent a value scalable solution for courses without experiential leaning components. Using commercial cloud-based AI tools and locally hosted models to produce custom images. These images were then combined into synthetic scenarios to depict dynamic interactions within fictional information systems to provide students with opportunities to analyze system components, actors, and interactions. The process is not entirely automated by generative AI models, but images produced using the models can be manually combined to produce a customized scenario according to the instructor’s needs. My scenarios depict use an anime style set of characters to play the roles of actors using an information system. Backgrounds of the business the system I used in are generated by local LLM models while commercial cloud tool are used to generate the characters and modify their expressions to suite the actions of the scenes depicted. By combining these AI generated resources with manually created user interface mock-ups I give the students a synthetic observation experience that goes beyond a text description. This practice provides valuable practice gathering information system requirements from a scenario where the instructor can control the complexity and design intentional scenarios. This practice is adaptable across disciplines and allows instructors to support situated learning in context by simulating authentic tasks. This can lead to improved cognitive engagement and observation skills. Round 3 Description: My proposed session presents a Fall 2026 course that I am developing. The course is built around optimizing AI as a collaborator rather than a tool in order to deepen student design cognition. Prior to the August symposium, I will be conducting field research in Japan - using AI as a collaborator - to study urban and interior systems through a daily diagramming and drawing practice. My field research with AI as a collaborator will further inform my fall course and also feed several scholarship opportunities over the upcoming year. Round 4 Description: Since 2023, I have experimented with multiple approaches to AI integration across in-person first-year writing courses and asynchronous online technical writing courses at Virginia Tech. Through classroom observations and student feedback, I moved away from viewing AI use as a simple “AI-on” or “AI-off” choice. Instead, my teaching evolved toward a more balanced, human-centered approach that considers course context, student feedback, and learning goals. In my talk, I want to share my experiences, how I started and how I changed AI integration, what assignments survived (such as process-based projects, multimodal and fieldwork assignments), and what assignments no longer worked. Using examples from my courses, I aim to discuss some ongoing questions: 1. How much structure around AI use is helpful? To explore this question, I compare highly-structured and lightly-structured AI integration in writing instruction. 2. When does transparency about AI use turn into overwhelming labor for students? To address this question, I share my students’ feedback collected through reflective discussions and anonymous surveys. 3. How do AI policies influence writing authenticity, and ethical use? To answer these questions, I compare my two course’s policies in different modalities (online versus in-person), while considering the differences and similarities of the learning goals and students needs in these two courses. 4. How does AI influence our pedagogy and teaching philosophy? To discuss this, I expose my pedagogy shift from policing to trust-based course design, and will also analyze my observations of students’ fears and my own uncertainties during these experiments with AI.

Presenters: Jon Catherwood-Ginn and Kim Loeffert The increasing ubiquity of AI deepens educators’ responsibility to cultivate students’ AI literacy. Perceptions of AI vary widely, necessitating that educators challenge students to actively test and evaluate AI’s capabilities and limitations within diverse contexts. Educators can sharpen students’ critical faculties when utilizing AI, support students’ identification of the technology’s relevance to their goals, and develop the buy-in and skillsets necessary for students to be adept and mindful in their use of AI. The panel examines two implementations of AI in performing arts courses. In an undergraduate music theory course, students engage with AI as a tutor and test its capabilities. By exposing the AI’s shortcomings relative to a discipline-specific task, the activity reinforces students’ developing expertise and self-confidence. By design, many students leave constructively skeptical of AI as a reliable resource. A graduate-level, theatre-based communicating science course positions AI as a dialogic partner. Students define an individualized “difficult” audience and generate a persona with a chatbot. The students explain their research to the AI in clear and relatable terms, to which the persona responds with questions and “push-back.” This exercise highlights the value and shortcomings of simulated critique with an AI, which includes stereotyping and reduced conversational depth. These examples point to how, paradoxically, educators can integrate AI into coursework as a partner and foil for student learning. We encourage attendees to consider how assignments can guide students to critically evaluate AI and develop informed, context-sensitive approaches to its role in their learning.

Presenter: Ivonne Wallace, Center for Excellence in Teaching and Learning Most approaches to AI in academic work treat these tools as productivity upgrades: better prompting, faster drafts, smarter search. In this session, I start instead by designing a working cognitive architecture built specifically for academic and intellectual workflows. Agentic tools like Claude Cowork accumulate context by default, but I argue that explicit initial design decisions can better organize those context layers to support our work. What we put in different context layers and how we design our emerging infrastructure can make the difference between a tool that supports the kind of intellectual work we want to do or one that contributes to dangerous habits that erode our intellectual autonomy. I walk through the cognitive architecture I have built in Claude Cowork, including a collaboration model as explicit intellectual partnership theory, guardrails engineered as counter-architecture against the tool's own default tendencies, and a convention for tracking intellectual provenance across sustained collaborative work.

Demo 1: Scaffolding Toward Simulation: Using AI to Build Preparatory Assignments for a Hands-on Environmental Conflict Negotiation Presenter: Kathy Perkins Demo 2: Empowering Extension Education: Aligning 4-H Curriculum with AI and the Beyond Ready Framework Presenters: Chad Proudfoot, Tonya Price, Alyssa Walden, and Ashley Craun Demo 1 Description: This presentation outlines how the instructor used AI to assist with the development of an instructional workflow that prepared students for a simulated conflict resolution process. The presenter will show how the workflow built conflict literacy, stakeholder analysis skills, collaborative problem-soving and finally role-based negotiation skills. It will highlight the progression from concept learning, to applied analysis to role-based practice and reflection. The presenter will also discuss the student engagement and learning outcomes resulting from this process. Demo 2 Description: As university faculty navigate AI in traditional classrooms, land-grant institutions like Virginia Tech must also address AI's role in their broader educational and engagement mandate: Cooperative Extension. Following the recent Executive Order prioritizing AI education within the Cooperative Extension System, 4-H educators face a critical turning point. Simultaneously, the national 4-H Beyond Ready framework challenges 4-H faculty and volunteers to prepare youth for a rapidly evolving 2030 landscape. How do we integrate "AI On" workflows while preserving the hands-on, non-formal, experiential learning, relationship-based core of Positive Youth Development? This workshop will bridge the gap between campus classrooms and community education, demonstrating how non-traditional educators are actively balancing AI integration with human-centered learning. We will explore the intersection of this national mandate and the Beyond Ready framework, illustrating how AI can handle heavy instructional design lifting so educators can focus on creating learning environments specific to their local needs while still maintaining high quality curricular and educational standards. One example is that participants will engage in a hands-on demonstration of the Utah State University (USU) 4-H Curriculum Developer and Beyond Ready Curriculum Checker. Built using ChatGPT, this agentic workflow tool streamlines the creation and validation of educational materials, ensuring strict alignment with readiness standards without replacing human pedagogical expertise. Attendees will leave with a broader understanding of the university teaching mission that includes Extension, practical experience using an AI-driven curriculum checker, and actionable strategies for leveraging AI to enhance, rather than replace, human connection in experiential learning environments.

Presenters: Kylee Shiekh, Benjamin Chaback, Mitch Gerhardt, and Andrew Katz Artificial intelligence (AI) is increasingly shaping teaching and learning decisions in higher education. When faculty use or teach about AI, their actions reflect assumptions about how AI works, what AI can and cannot reliably do, where AI can be used, and how students should engage with AI. As a result, these assumptions play a critical role in shaping course policies, assignment design, and assessment strategies. This interactive workshop invites faculty to consider their assumptions about AI and engage with other colleagues across disciplines. Rather than promoting a single “right” approach, this session creates space for guided reflection on the evolving landscape of AI in higher education, including its functions, uses, and implications for teaching and learning. Together, participants will examine how their perspectives influence pedagogical choices and student engagement with AI. Informed by our research group’s ongoing project interviewing nearly 170 engineering instructors across 18 U.S. institutions, we use real-world scenarios to surface points of uncertainty in AI in higher education.This session is designed as a collaborative and reflective space where participants can engage with colleagues across disciplines, compare perspectives, and make sense of emerging practices together. Attendees will map their own AI conceptions, including their beliefs about its capabilities, uses, and limitations, engage in peer discussions to collectively think through and reflect on dilemmas related to AI in teaching and learning. The workshop concludes by helping participants connect these insights to their pedagogical goals and identify next steps for their approach to AI in teaching.

Presenters: Margaret Ellis, Sehrish Basir Nizamani, and Naren Ramakrishnan In this workshop, we introduce DURA (Demystify, Use, Reflect, Assess), a framework for incorporating AI into courses to prepare students to be effective and ethical AI users. The components of DURA deliberately support the process of learning as well as the need for reliable evidence of student learning. The foundational component of DURA is the demystification of Large Language Models (LLMs), which underpin modern AI. When students understand how the technology works, they are more likely to be effective AI users who won’t overly trust the models or attribute thoughts and feelings to them! Students can also benefit from instruction and practice to use LLMs responsibly and productively. Beyond use, students are expected to reflect on their experiences with LLMs, supporting metacognition and self-regulation. Given the rapidly changing landscape and increasing student access to LLMs, instructors must also consider authentic approaches for assessment that align with these realities. We developed the DURA framework when redesigning CS2104 Problem Solving in Computer Science (CS) and have since applied it in other courses, both within and beyond CS. In this workshop, we will provide concrete guidance and adaptable materials to help instructors of all fields integrate DURA into their own courses, including explanations of how LLMs work and practical suggestions for LLM use, student reflection, and assessment.

Presenters: Stephen Edwards and Bob Edmison In the current higher education landscape, Generative AI is frequently viewed as either a threat to academic integrity or a mere shortcut to a finished product. However, when leveraged intentionally, these tools can serve as powerful "Socratic coaches" that scaffold the very self-regulated learning (SRL) skills—metacognition, critical reflection, and strategic planning—that students need to navigate their academic careers. This interactive workshop invites faculty from all disciplines to move beyond the "answer-seeking" paradigm. We will begin by identifying SRL objectives specific to your curriculum, such as mastering complex synthesis, debugging logic, or evaluating source credibility. Participants will then walk through the iterative process of developing and refining "Socratic coach" prompts. Unlike standard queries that provide direct answers, these prompts are designed to challenge students to perform the cognitive heavy lifting while the AI provides targeted feedback, guidance, and correction. The prompts will cover skills including studying, test prep, and reviewing. Through a series of live demonstrations and hands-on "prompt lab" sessions, attendees will build a personalized toolkit of scaffolds ready for immediate implementation. By the end of this session, you will be equipped to transform AI from a passive assistant into an active thinking partner, ensuring that your students remain the primary architects of their own learning. Join us to explore how we can empower students to not just use AI, but to think critically through it.

Round 1: Designing for Honest AI Use: What Happens When Students Must Disclose? (Eric Kaufman) Round 2: Vibe Coding in the Humanities: Designing Pedagogies for AI Experimentation (William Taggart) Round 3: An Interdisciplinary Approach to AI Education: Inside the New AI Minor (Margaret Ellis, Joan Watson, Stephen Edwards, Christine Julien, Brian Mayer, and Naren Ramakrishnan) Round 4: From Foundations to Application: Teaching AI Literacy in Engineering (Hulya Dogan and Mags Blackie) Round 1 Description: Asking students to disclose their AI use seems straightforward—until you try it. This lightning talk reflects on efforts to require student disclosure of generative AI use across four graduate-level courses, including three asynchronous online offerings. In response to growing ambiguity around academic integrity, I introduced a combination of explicit syllabus language, ongoing communication, and an AI transparency checklist that asked students to document how AI tools were used across stages of their work (e.g., ideation, drafting, editing). The goal was to move away from a policing model of academic integrity toward a transparency-based approach that treats AI use as permissible but accountable. Students were encouraged to view AI as a partner in their learning process while maintaining responsibility for accuracy, attribution, and original thinking. What emerged was more complex than anticipated. While some students engaged with the transparency framework as intended, others continued to use AI in ways that were undisclosed or inconsistent with course expectations, including cases referred for academic misconduct. These mixed outcomes suggest that requiring disclosure alone does not resolve the challenges faculty face. Instead, transparency appears to surface a deeper issue: many students lack clear frameworks for deciding when and how AI use supports—or undermines—their learning. This session will share practical materials (including a reusable transparency checklist) and reflect on implications for teaching. I argue that disclosure requirements must be paired with intentional efforts to build student judgment and AI literacy, reframing integrity as a function of visible process, not just final products. Round 2 Description: “Vibe coding,” a term popularized by Andrej Karpathy, describes a mode of software development in which users build functional programs primarily through natural language prompting rather than formal coding expertise. This presentation introduces approaches to teaching vibecoding in FL 2744: AI and Global Languages, a new undergraduate course that introduces students, many of whom have no programming background, to the design of AI-enabled language tools. Within the course, students treat large language models not just as objects of analysis but as experimental partners, iteratively prompting, testing, and remixing code to build and refine cross-language AI systems. While thecurriculum includes translating at scale, examining ethical questions surrounding AI, and developing a working understanding of model architectures, this presentation focuses specifically on the pedagogical challenges and opportunities of teaching students to build with AI. The process is exploratory and nonlinear: students begin with open-ended concepts and progressively refine them through iterative cycles of prompting, debugging, and revision. In doing so, theyencounter both the limitations and possibilities of AI coding tools, leaving room for surprises, creative failures, and solutions that open up new possibilities for experimentation. This presentation argues that vibe coding offers a powerful entry point into computationalcreativity for humanities students, while also raising important questions about authorship, epistemology, and the boundaries between users and developers. Round 3 Description: The new Artificial Intelligence (AI) Minor provides a coherent, interdisciplinary framework through which students from any major can develop a foundational understanding of how AI systems function—and why their design and use matter. Anchored in Computer Science fundamentals and complemented by integrative modules, the minor advances both technical fluency and contextual literacy. Students complete a set of core coursework in AI fundamentals and explore applied AI in a particular disciplinary context by layering coursework selected from one of a set of augmenting modules: Computational, Arts, Social Science and Human Impact. Across the minor, learning outcomes progress from introductory fluency to advanced synthesis, uniting computational reasoning with ethical and creative insight. Graduates will emerge as adaptive thinkers and responsible innovators, embodying Virginia Tech’s mission of Ut Prosim—that I may serve—through interdisciplinary perspectives on the creation and use of AI. Round 4 Description: ENG 1984: AI Literacy for Engineers is a 1-credit, 15-week introductory course at Virginia Tech’s College of Engineering designed to build foundational and applied AI literacy for first-year engineering students. The course is structured in two phases that progressively develop both understanding and practice. In the first eight weeks, students focus on AI literacy foundations. They explore how large language models work, how to design effective prompts, and how to critically evaluate and verify AI-generated outputs. They also examine ethical considerations and academic integrity in AI-assisted work. To support learning, each week follows a consistent instructional cycle of Prepare, Practice, Reflect, and Checkpoint, ensuring that students build confidence and competence in a structured and iterative way. The second phase shifts from understanding to application. Each student selects one of Virginia Tech’s 14 engineering disciplines and completes a small AI-assisted project within that context. Students use AI tools for tasks such as literature exploration, ideation, analysis, and technical communication. Throughout the process, they maintain a Prompt Journal documenting their interactions with AI and reflecting on decision-making, limitations, and responsible use.

Round 1: Student-Authored AI in a First-Year Professional Development Course (Paige Normand) Round 2: Lost in Translation: AI’s Evolving Role in International Student Learning--Insights from the VT Language & Culture Institute (Rebecca Etzler) Round 3: AI²: Fixing the System, Not Just the Symptoms (Tiffany Willis) Round 4: Co-Constructing an AI Aware and Empowered Art History Classroom (James Jewitt) Round 1 Description: This teaching demo introduces a student co-built AI career assistant developed in SPES 2014: Professional Development at Virginia Tech. Rather than presenting AI as a tool to simply use, this course has students generate the knowledge base the AI draws from. Throughout the semester, student note-takers document peer feedback, internship insights, field-specific strategies, and networking takeaways in a shared folder. That documentation becomes the foundation for a Retrieval-Augmented Generation (RAG) system, deployed through Microsoft Copilot Studio, that students can use for tailored professional development guidance long after the class is over. The approach addresses two challenges at once: it gives students a reason to engage deeply with career development content, and it produces a field-specific resource for our seven majors that generic platforms cannot replicate. Students are not just users of AI; they are its authors. This demo will walk participants through the course structure, the documentation workflow, the Copilot Studio build, and live interaction with the deployed assistant. No coding required, though participants should expect a meaningful setup investment. Round 2 Description: Two instructors in the Virginia Tech Language and Culture Institute’s (VTLCI) will share their insights into international student use of AI to circumnavigate reading, writing, listening, and even speaking tasks. Students in our pathways program, Advantage VT (AVT), lack the required minimum language scores to directly matriculate into VT undergraduate and graduate programs. However, entrance into our highest level is very close to the undergraduate international admissions requirement, and our higher level students take credit-bearing courses in addition to our classes. We have been seeing a tsunami of effects on student comprehension, output, learning strategies, and overall motivation for language learning. In short, AI-powered translation technologies are revolutionizing student relationships with the English language and our role as their educators. We will share our unique perspective on working with Virginia Tech’s second language learners. We will discuss discrepancies between minimum TOEFL scores and students’ actual abilities, highlight anecdotal information about international student AI use, and share suggestions for instruction and assessment. Our goals are to raise awareness about potential range of language abilities of the international students in your classes, to improve professors’ abilities to identify the difference between true proficiency and technologically enhanced skills, and to help ensure that students receive the educations they deserve in spite of growing linguistic gaps being filled by these technologies. Round 3 Description: The Office for Undergraduate Academic Integrity (OAUI) has seen an uptick in cases reported for generative artificial intelligence since its inception in 2022. OUAI received 60 reports for AI-related Honor Code violations in 2023-2024, 137 reports in 2024-2025, and approximately 263 reports to date. In an effort to address this 303% growth, Virginia Tech convened an AI2 working group to address the capabilities of AI, the educational needs and the intellectual property rights of faculty and students, meeting the preponderance of evidence standard, and the potential for bias in reporting. This light talk will explore the plan of action for academic integrity and artificial intelligence through Virginia Tech’s framework for implementing appropriate measures to prevent and address violation of the Virginia Tech Honor Code/Policy 6000. In this session, participants will gain an actionable approach for implementation in four areas: policy, faculty resources, student resources, and AI tools/detection. Round 4 Description: Over the past several years, art history faculty at Virginia Tech have observed exponential increases in the use of generative AI (GenAI) by undergraduates. It is a major “disruptor” in our field that faculty must not only contend with but embrace. Yet cheating and plagiarism as a result have reached critical mass in the program’s courses. At present, the relatively unregulated misuse of AI across VT, despite new guiding principles published recently, have created significant pedagogical challenges. In the art history program, course policies concerning its use are left up to individual faculty, often lacking requisite training or expertise. This lightning talk presentation examines the Art History Program’s efforts to craft a discipline-specific AI policy. It focuses on three aspects of this process: 1) drafting an intentional and sustainable AI policy that is specific to the visual arts and related GenAI apps, software, and tools; 2) designing and developing AI policy and literacy training modules for students embedded in Canvas that includes: learning outcomes, short video clips, documents, knowledge assessments, and proficiency exercises; 3) and finally, an attendant written policy on cheating and plagiarism involving misuse of GenAI specific to the visual arts. Art History’s intention was to create a policy that is, on one hand, tailored to the disciplinary specific intersections of AI and art history, while on the other hand, customizable enough to potentially scale to integrate with the curricula of the additional three undergraduate majors within the School of Visual Arts—Graphic Design, Studio Art, and Creative Technologies.

Presenters: Junghwan Kim, Jennifer Mooney, Ruichuan Zhang, and Julia Feerrar This panel discussion will highlight the lived experiences of instructors teaching AI across diverse disciplines at Virginia Tech. The panel brings together perspectives from the library, humanities, social sciences, and engineering. Julia Feerrar will share her experience leading AI literacy initiatives at Virginia Tech, including reflections on the concept of AI literacy and its practical applications. Jennifer Monney will discuss her approach to integrating AI into writing instruction. Ruichuan Zhang will provide insights from an engineering perspective, focusing on teaching AI to engineering students. Junghwan Kim, who developed and taught the first-of-its-kind course, Generative AI Applications in Social Science, at Virginia Tech, will share his lessons learned and serve as the session moderator. The session is designed to emphasize practical, experience-based insights rather than theoretical discussions.

Demo 1: AI in the Design Studio: Student Perceptions and AI-Integrated Rendering Workflows Presenters: Alp Tural, Michelle Huh, and Elif Tural Demo 2: AI-Supported Canvas Discussions: Scaling Feedback and Engagement Presenters: Dan Dunlop, Sanjay Shankar, Jay Foti, and Bob Edmison Demo 1 Description: Design studio pedagogy is rooted in problem-based learning and reflection-in-action, in which students construct knowledge through an iterative process. This unique pedagogical position of the design studio makes it a significant environment for examining whether or how generative AI is adopted considering creativity, ownership, and design and analytical thinking. This session bridges architecture and design students’ perceptions of generative AI use in the studio with live tool demonstration, offering insights and exposure to AI use in design education. The session will open with findings from an ongoing survey of design students, focusing on: (1) how AI use shifts as students move from lower-level to upper-level studios, (2) whether students differentiate between visual and language-based AI tools, and if that distinction changes as studio work becomes more research-driven or technically focused, (3) the role instructor attitudes, peer norms, and policy clarity play in students' decisions to adopt or avoid these tools, and (4) whether students develop their own internal criteria for appropriate use, even when formal studio guidance is ambiguous or absent. The demonstration component covers AI-integrated rendering workflows students tested in junior and senior interior design studios, including Chaos Veras, ChatGPT and Nano Banana, showing what these tools produce, where they intervene in the design process, and what they demand from the designer versus what they replace. Attendees will leave with an understanding of the current AI rendering tools in design education, informed by both design students’ perceptions and hands-on tool exposure. Demo 2 Description: Threaded discussions have long supported student engagement, written reflection, and peer learning, yet their design and use have changed little over time. In large-enrollment courses, they are especially difficult to monitor and sustain as meaningful instructional spaces. Instructors face a persistent challenge: how to track participation, motivate substantive contributions, and provide timely, targeted feedback across many posts. Instructors and students in CS3604 Professionalism in Computing have engaged these challenges over many years. Large language models (LLMs) offer new possibilities for analyzing discussion data at scale. This session demonstrates an AI-powered tool developed through a senior computer science capstone project at Virginia Tech that integrates with Canvas via LTI and the Canvas API. Leveraging LLMs within the VT ARC environment, the tool analyzes discussion activity to help instructors identify emergent themes, common misconceptions, and high-quality contributions aligned with instructional goals and assessment practices. The demo will showcase the interface, core features, and integration within Canvas, including how the tool supports instructor workflows and potential interaction with student posts. Emphasis is placed on both practical capabilities and the broader questions these tools raise about the role of AI in asynchronous discussion. Attendees will see a live demonstration, along with brief insights into design decisions, limitations, and classroom applications, with time for questions. This session is intended for instructors seeking practical, scalable approaches to discussion-based learning in the age of generative AI.

Presenter: Scott Mutchler Large language models offer powerful new possibilities for personalized instruction, but realizing that potential requires thoughtful design grounded in pedagogical intent. This hands-on workshop introduces Virginia Tech educators to two practical workflows using ARC's locally hosted LLMs. First, participants will learn to build custom GPTs that function as course-specific tutors — models grounded in their own syllabi, readings, and problem sets, and guided by constitutional prompts that enforce disciplinary norms, academic integrity, and instructional boundaries. Second, participants will explore OpenCode, an agentic AI coding tool, as a means of rapidly generating and iterating on course materials such as lesson plans, assessments, rubrics, and interactive learning activities. No prior AI or coding experience is required. Participants will leave with a working prototype of a custom tutor and a repeatable workflow for AI-assisted course development, along with a framework for evaluating when and how these tools genuinely serve student learning.

Presenters: Jonathan Briganti and Anne Brown Artificial intelligence is rapidly changing how STEM students engage with data, enabling those with little to no programming training to analyze complex datasets but not necessarily to understand or evaluate the underlying methods. This demand introduces new challenges where molecular biosciences students may rely on AI-generated code for data analysis without understanding the appropriateness of the code or their actions. This creates a need for instructional approaches that emphasize intentional use of AI alongside foundational computational thinking. To address this, we propose a set of training materials and guided activities centered on “Motif Maps”, which are structured reference tools that help students recognize and select common coding patterns. These motifs represent foundational building blocks that students can map to specific problem types that they are trying to solve with their data. This approach provides a practical model for designing AI-integrated assignments that maintain student engagement with core computational thinking skill. Participants will engage in an interactive activity where they identify required motifs, prompt an AI tool to generate code for each component, and assemble these into a functional pipeline. The emphasis is not on independently writing code, but on understanding, evaluating, and correctly applying AI-generated outputs for analysis. The session will focus on building student confidence while maintaining critical evaluation skills. The session will also address code validation, troubleshooting, and the limitations of AI tools. This work presents a discipline-specific framework for integrating AI into coding instruction through structured, motif-based activities that preserve critical thinking and meaningful student engagement.

In this session, we will reflect and answer final questions, as well as discuss possible next steps.