Teaching & Learning Summit

Encouraging Refusal of GenAI in Teaching and Learning

Jay Aronson

I would like to take the opportunity to talk to my fellow Carnegie Mellon faculty and instructors on why we should consider refusing engagement with AI altogether in certain foundational courses, what refusal might look like, and why the university should take a stronger stand in supporting those faculty and instructors who choose to take this stance. Carnegie Mellon as a university tends to be a booster of generative AI/LLMs at the institutional level, and at best agnostic about the use of these technologies in the classroom. This policy makes it challenging for those of us who believe AI presents and existential risk to a generation of students who might not want to use these technologies in their education, but feel pressure to do so nonetheless. By taking refusal seriously, we can at least give this subset of students the tools they need to confidently reject generative AI and LLMs, and also help students who might believe that they are inevitable see that a different future is possible.

The Use of Generative AI for Student Interviewing Practice & Feedback

Brandon Bodily

Connell Pensky, Allison, Eberly Center

This presentation reports on research from a Eberly-sponsored GAITAR project. The research question focused on whether generative AI can be used to help students gain confidence in conducting customer discovery interviews. The purpose was not to replace in-person interviews with potential customers. Rather the objective was to see whether gen AI could be used as a feedback and role-play partner for students to practice interviews prior to the actual in-person interviews. Practice interviews were conducted across multiple semesters as part of an in-class assignment with students randomly assigned to a class peer or gen AI partner. The results show that student confidence and self-efficacy on conducting interviews improved for both groups and no adverse effects were observed for those using gen AI. However, each method seems to have strengths and weaknesses that could be considered in the application. Ideas on how our results may inform other instructors will be shared.

Wimmer & Eberly Active Learning Incubator Fellowship Program

Allison Connell Pensky

Come and learn about the Wimmer & Eberly Active Learning Incubator (WEALI) Fellowship program. We will present the goals of the WEALI program, how an instructor can develop an idea and apply to the program, and a few insights we’ve gained through student data.

Promoting Effective Teamwork in Group Projects

Emily Drill

An important part of being a scientist is working well with others, but this isn’t often taught in science courses. My goal was to make effective teamwork a more explicit objective of group projects in lab courses and to introduce a framework to intentionally guide students in how to work together successfully. To do this I designed and incorporated three main interventions into group projects in upper-level biology labs: team contracts at the start; mid-project evaluations, designed to prompt students to consider whether their group is working successfully and take action as needed; and final group evaluations. Responses suggested students used team contracts to resolve disputes, and reflected and made changes to their group work at the midpoint. I plan to use these in the current semester, with modifications to improve student accountability. I will share these documents, which are adaptable to group projects in a variety of disciplines.

Examining Intentional and Scaffolded GenAI Instruction in Professional Writing Courses (GAITAR Study)

Rebekah Fitzsimmons

Eckstein, Lydia, Eberly

In Fall 2023, I conducted a GAITAR study in my Writing for Public Policy course to examine how students’ self-efficacy changed over the course of the class, and how GenAI permission and use impacted their writing performance. All three sections of the course incorporated four scaffolded and pedagogically informed activities on utilizing GenAI in public policy writing. The control section was not permitted to use GenAI on formal assignments, while the intervention section was permitted to use GenAI with citations. Results indicate that students’ self-efficacy regarding both public policy writing and ethical GenAI use significantly increased across all sections. These results suggest that while GenAI may serve as a useful supplemental tool, it does not replace the value of scaffolded instruction and engaged learning, as well as the value of integrating GenAI literacy into field-specific curricula as an emerging foundational skill. For the summit, I propose to present the findings from this study and our soon to be published article, co-authored by Lydia Eckstein at Eberly.

Interdisciplinary Community: Pittsburgh Partnerships in an Environmental Equity Grand Challenge

Barbara George

Nuria Ballesteros Soria

This class description explores interdisciplinary teaching strategies in a Grand Challenge course at Carnegie Mellon University focused on Equity and the Environment. Rooted in environmental justice and the Sustainable Development Goals, the course scaffolded freshmen from diverse disciplines to tackle "wicked" environmental problems through interdisciplinary teamwork. Addressing complex environmental challenges requires cross-disciplinary approaches. However, interdisciplinary research remains rare among university freshmen. The course design integrated guest speaker panels and structured primary and secondary research involving Pittsburgh community stakeholders and policy makers. Students learned to synthesize local community insights into actionable research and translate these experiences into ongoing undergraduate research opportunities. Additionally, this presentation highlights an ongoing study with CMU’s Eberly Center, investigating how freshmen understand "interdisciplinarity" in their research. Instructors will gain practical strategies for course design, student scaffolding, and research methods aimed at fostering undergraduate interdisciplinary collaboration.

Empowers students to think critically and write intentionally in the AI Era

Suguru Ishizaki

Kaufer, David, English

This presentation introduces myProse, a writing support tool that engages students in critical thinking through structured, multi-layered feedback aligned with their writing assignments. myProse addresses writing at multiple levels, from higher-order concerns such as genre conventions, argumentation, and logical flow to fine-tuning issues including paragraph clarity, sentence clarity, tone, and source usage. By prompting students to reflect on their composing decisions across multiple drafts, myProse builds metacognitive awareness that helps students grasp threshold concepts in writing, transforming not just their writing, but how they think about writing itself.

Designed for use across Foundational Writing Courses and Writing in the Disciplines, myProse supports students in developing transferable writing skills and applying them within genre-specific contexts, while freeing instructors to focus on higher-order teaching goals.

We will present preliminary research findings and sample instructor-developed lesson plans from pilot implementations, illustrating how myProse can be meaningfully integrated across diverse writing contexts.

Building Student Understanding of Interdisciplinarity through Introspection and Collaboration

Kevin Jarbo

Creating and applying interdisciplinary perspectives and approaches to society’s most pressing issues are central to Dietrich College’s mission. While students value this mission, when probing them directly, it isn’t clear how deeply they understand and have internalized any specific meaning of interdisciplinarity toward analyzing and addressing social problems. In my course 88285 Deconstructing and Dismantling Discrimination, I present a definition of interdisciplinarity that we revisit during discussions and through implicit and explicit assignment prompts. In a final reflection, students assess how specific assignments and a team project contributed to deepened understanding of their own disciplinary perspective, independently developing an interdisciplinary perspective, and addressing a social issue with a multidisciplinary team in an interdisciplinary way. With colleagues, I will share my definition, assignment prompts and student artifacts, approach to constructing multidisciplinary teams, and excerpts of students’ final reflections from 2 semesters of the course.

Supporting Student Writing with AI and Teacher Feedback among High-beginner Korean Learners

Sanghee Kang

This showcase presents the design and implementation of writing activities and feedback practices for high-beginner Korean language learners. The writing tasks were based on grammatical features covered in the course and contextualized around Korean cultural events or familiar topics. Each task provided content and language support appropriate for learners’ proficiency level. Students received AI feedback and teacher feedback on each writing task. Teacher feedback selectively addressed language-related issues as well as overall aspects of students’ writing, whereas AI feedback reformulated learners’ texts for linguistic accuracy and naturalness. To encourage active engagement with feedback, students completed a guided feedback processing activity, in which they examined feedback points, indicated whether they agreed with the feedback, and explained the reason for the change. The showcase will present sample writing tasks, examples of teacher and AI feedback, the guided feedback processing form, and the overall instructional sequence from drafting to revision and reflection.

Multilayer Workflows: Parallel Foundations in Beginning Interaction Design

Haeyoung Kim

In Playful Interaction, a School of Design elective open to all CMU students, many are new to interaction design and must imagine interactive experiences before knowing what technologies can do. The usual sequence—concept first, tools second—limits their ability to imagine beyond solutions they already know. My strategy, "parallel foundations," treats hands-on making and technology exploration as equal starting points. From the outset of designing a sensory, interactive object, students alternate between three prototyping workshops (quick physical mockups, material tests, and moving between physical and digital prototypes) and three hands-on technology workshops exploring electronics and sensors, projection mapping, motion tracking, AI-assisted coding, and internet-connected devices. Student reflections, peer discussions, and final projects showed early technology experiments often redirected students' initial ideas, so technical possibilities shaped concepts rather than being added afterward. I will share the workshop sequence, materials, and student project examples, adaptable for any hands-on course.

Join us at the Provost's Institute on Inclusive Teaching

Sophie le Blanc

Patrick Walsh

Want to learn more about inclusive teaching and join others interested in the topic? The Provost's Institute on Inclusive Teaching is for you! It is a week-long faculty program in May run by the Eberly Center. Come learn more about the program (format, sample topics/activities) as well as the affiliated Eberly Student Partners Program that all Institute fellows are eligible to participate in. An Eberly Student Partner will join us to answer faculty questions about engaging in student partnerships. We will also answer questions about the proposal process or any other questions about both programs.

Student-Led Journal Clubs to Improve Analytical and Presentation Skills in Immunology Education

Yasser Majeed

Lydia Eckstein, Yoko Nakamura, Harley Chang

Journal clubs represent an active-learning approach aimed at developing key student competencies. Traditional lectures do not always fully engage students, which can limit their professional development. This study evaluated student-led journal clubs to help undergraduates improve key competencies, including analytical, critical thinking, and presentation skills. The research was conducted within 03390 - Molecular and Cellular Immunology. The study used a comprehensive pre- and post-survey design to evaluate educational outcomes. Students completed a self-efficacy assessment and a journal club perceptions survey to measure personal growth and collect journal club feedback. The data suggest a robust increase in self-efficacy criteria linked directly to the learning objectives. Students also reported improvements in analyzing and presenting scientific literature. While the data suggests journal clubs are valuable teaching tools, the small class size limited the conclusions. Future iterations of this research will apply the framework to a larger student cohort(s) to further evaluate these educational benefits.

Teaching Students to Use AI Effectively for Writing Tasks

Sarah Mansfield

Wolfe, Joanna, English

Students are sometimes tempted to use AI for writing assignments. While there are multiple reasons for this problem, one is that students, who lack expertise in professional and academic writing, may genuinely see AI writing as better than their own. I will walk participants through an activity I developed for a First-Year Writing course, Writing About Data, in which students generate and compare quantitative data about the language of human-written academic research articles to AI-generated counterparts. I will show participants how I engage students in noticing and articulating the potential downfalls of AI-generated writing in terms of the purposes or expectations it does not serve. For example, when doing this activity, students often notice that AI writing is more certain, less readable, and varies less based on context. Participants will leave with a short, easily adapted activity handout that helps make the flaws of AI-generated writing more visible to students.

See One, Do One, Teach AI: Supporting Mastery and Self-Efficacy Through AI-as-a-Tutee

Chris McComb

Giovanni Natasha, Anshika Agarwal, Lydia Eckstein, Michael Melville, H. Elisabeth Ellington, Zach Mineroff

The use of generative AI tools is increasingly common in engineering education, both formally and informally. Embracing these tools presents an opportunity to enhance student learning through novel pedagogical approaches, and many efforts have explored the use of AI agents to deliver personalized content and information for students. In contrast, this project investigates the impact of using a Large Language Model (LLM) as a “tutee” in a sophomore-level mechanical engineering course. Building on the “see one, do one, teach one” paradigm of mastery, this study explores whether teaching technical skills to an AI tutee can improve student mastery of core mechanical engineering concepts. This work evaluates performance on a well-established concept inventory across a group of students who repeatedly interact with an AI tutee, compared to a group with no such interaction. For students engaging with the AI tutee, we also capture and analyze self-efficacy, chat transcripts, and free response reflection questions. Preliminary results indicate the potential value of utilizing GenAI as a tutee. Ultimately, the aim of this research is to contribute scalable teaching strategies for large engineering classes and address disparities in student preparedness and backgrounds by providing a consistent and accessible mastery opportunity.

How to Best Leverage Industry Mentors?: Assessing Student Perceptions and Learning Outcomes in an Undergraduate Civil and Environmental Engineering Project Course

Joe Moore

Struzak, Maya, CEE; Eckstein, Lydia, Eberly

How does structured engagement with industry mentors affect students’ self-efficacy, perceptions of professional practice, and performance in a project-based engineering course? We examine this question in a required junior-level project course in which teams complete two open-ended engineering projects, each paired with a different professional mentor. We use a mixed-methods design combining pre-, mid-, and post-course surveys measuring self-efficacy, expectations and perceptions of mentoring, and qualitative responses about goal setting, problem-solving, teamwork, and professional growth. We also analyze project performance and its relationships with mentor engagement, self-efficacy, and mentoring satisfaction. Preliminary results suggest that students value mentors for connecting theoretical concepts to professional practice and clarifying expectations for engineering work. Students also report greater motivation when paired with preferred mentors. We will present changes in self-efficacy and perceptions across the three surveys, qualitative themes from mentoring experiences, and relationships among mentor engagement, student performance, self-efficacy, and satisfaction.

DARE to Evaluate Differently?: Using GenAI for Educational Assessment

Nimer Murshid

Mineroff, Zach, Eberly Center; Eckstein, Lydia, Eberly Center; Melville, Michael, Eberly Center; Ellington, H. Elisabeth, Eberly Center

This project explores using generative AI (GenAI) as a partner rather than a tool for quick answers in Chemical Building Blocks, an elective chemistry course for non-majors. The study investigated how ChatGPT use affects student learning outcomes and self-efficacy, and how students perceive GenAI as a learning tool. Students engaged in scaffolded activities combining AI-assisted exploration, collaborative evaluation of responses, question generation, and independent application without AI. Learning outcomes for Spring 2025 (GenAI, n=16) were compared with Spring 2024 (no GenAI, n=16), alongside pre/post self-efficacy surveys and student reflections. Students using GenAI performed significantly better on selected assessments, including Exam 2, and showed significant gains in chemistry self-efficacy, although overall GenAI self-efficacy did not increase. Student perceptions were mixed, highlighting both GenAI’s value for clarification and concerns about accuracy and prescribed interactions.

Table of Exam Specifications (TOES): A Reflective Tool for Students and Instructors

Nimer Murshid

Eckstein, Lydia, Eberly Center; Pottmeyer, Laura O., Eberly Center; Nakamura, Yoko, Eberly Center

A Table of Exam Specifications (TOES) is traditionally used to align assessment content, learning outcomes, cognitive levels, and question types. This presentation explores how TOES can move beyond an exam-planning tool to become a reflective framework for both learning and teaching. Students were progressively introduced to TOES, moving from instructor modeling to constructing their own tables as part of exam preparation. TOES provides students with a metacognitive “mirror” to examine how their preparation aligns with expected outcomes and cognitive demands, while allowing instructors to reflect on assessment alignment and instructional effectiveness. Preliminary findings showed a significant increase in overall study-habit self-efficacy and positive student reflections on TOES as a tool for identifying what to focus on and what types of questions to expect. The approach illustrates how transparent assessment design can promote reflection, intentional learning, and continuous improvement in both learning and teaching.

Drawn to the Crown: Practicing Productive Failure

Raelin Musuraca

Hammer, Jessica, Human-Computer Interaction Institute; Decossard, Regine, Human-Computer Interaction Institute; Teffera, Leah, Human-Computer Interaction Institute; Credit, Kimberly, Human-Computer Interaction Institute

University students entering design or innovation programs sometimes struggle to embrace the ambiguous, experimental nature of project-based work because their prior academic success often came from correct answers and polish. This tension creates a challenge across project-based classrooms: students may overinvest in initial ideas, resist feedback, avoid creative risks, or simply stall out. We developed Drawn to the Crown, a rapid, small-group drawing game that helps students practice responding to changing constraints, building on imperfect work, and separating an idea's value from polished presentation. Playful, peer-led judging lowers the stakes, while a guided debrief helps students connect their gameplay to brainstorming, prototyping, critique, teamwork, and other project practices. Our design suggests that time pressure, visible iteration, shifting criteria, and structured reflection are central to the experience. We invite instructors from any project-based discipline to implement and evaluate the game. Participants at the summit can explore printed game materials, facilitation guidance, adaptation suggestions, and reflection questions. Those interested in playtesting the game will later receive game sets and support.

An Experiential Learning Course Design: From Genetic Clues to Personalized Care

Adviti Naik

Nakamura, Yoko, Eberly Center; Melville, Mike, Eberly Center; Chawla, Avi, Eberly Center

Precision medicine is a growing field in healthcare and biological sciences research, yet it has traditionally been taught at the postgraduate level. To introduce the field to undergraduates, Precision Medicine, an advanced Biological Sciences elective, was designed and first offered to eight students at CMU-Qatar in Fall 2024. Experiential learning was used to connect classroom concepts with authentic practice. Unidentified genetic disorder cases were investigated by student teams using patient information and databases. Evidence was evaluated, likely diagnoses were identified, and personalized management plans were proposed. Learning was reinforced through primary-literature seminars, guest speakers, ethical discussions, and attendance at a regional precision medicine and genomics conference. Structured reflections were also used to connect conference presentations, posters, and networking experiences with key course concepts. Increased self-reported understanding across all assessed topics was demonstrated by pre- and post-course surveys. The case-study workflow, conference reflection prompts, and assessment rubrics will be shared.

Increasing engagement and performance with consistent active learning strategies

Alexandra Newby

Pincus, Jacqui, Eberly Center; Connell Pensky, Allison, Eberly Center; Ni, Xiaolin, Eberly Center

In 06-262 Mathematical Methods of Chemical Engineering, I redesigned class sessions to replace portions of traditional lecture with frequent, structured opportunities for active learning. Each class includes short warmups, think-pair-share activities, and exit tickets. Think-pair-share activities vary from solving problems and creating concept maps to teaching and questioning, discussion, and computational activities. These activities give students repeated opportunities to make decisions, explain their reasoning, and receive feedback during class. I will share examples of these activities and practical strategies for incorporating active learning into a mathematically intensive course. Following the redesign, I observed increased student engagement, more positive ECF qualitative feedback, and increased FCE scores. Preliminary results also show an increase in student performance on exams. I will discuss what worked, what I would change, and lessons learned about using small, repeatable changes to shift students from primarily listening to actively participating in their learning.

Whose Knowledge Counts? Building Belonging in First-Year Writing by Examining Privilege and Access

Julie Pal-Agrawal

Research has shown that two early predictors of students' entry into their disciplines' research communities are confidence in research skills and a sense of belonging. First-year writing courses have predominantly focused on skills. They rarely invite students to examine their position relative to the research they are assigned. As a test case, sections of First-Year writing invited students to examine both their privilege in accessing research at CMU and their exclusion from research communities. This study asks whether naming that double position increases belonging. Across Fall 2024 and Spring 2025, we added several reflective prompts to the existing curriculum, and pre- and post-surveys measured self-efficacy and belonging. Self-efficacy rose from 67 to 84.6 percent in Fall 2024 and from 69.4 to 84.1 percent in Spring 2025. Belonging averaged 5.95 out of 7, and students identified interactions with the instructor and peers as its strongest contributor.

Assessing Reasoning with Sequential Images: Possibilities and Limitations

Robert Rapoport

David Lin (HCI)

How far can sequential images teach critical reasoning? As image and video generation change rapidly, what role should these tools play in the classroom? From a critical design perspective, I ask whether students bring latent literacies to juxtaposing images, and whether these might count as multimodal reasoning. Research suggests some AI models can reason through image sequences without text, at least on spatial tasks. For several years I have developed Critical Montage: exercises in which students build essay-like arguments from ordered images, then defend each juxtaposition in annotations and live critique. VideoMap (Lin, 2024), an interface developed at CMU, may help reveal how students engage critically with the latent spaces in which AI models organize images. I propose exercises that ask whether some modes of interpretation arise only when a room reads images together. I will share two Critical Montage exercises and discuss how they might transfer across disciplines.

Stop Rushing the Debrief

Julie Saunders

Active learning activities are a staple of the modern classroom, yet the debrief—where deep conceptual synthesis occurs—is often rushed due to time constraints, fear of "dead air," or loss of control over discussion.
In this showcase, I share how streamlining the activity in my Interaction Design Fundamentals studio class created space for an expanded debrief. These discussions reduced reliance on traditional lecturing, allowing students to "experience" concepts during the activity and unpack them together afterward.
Beyond content delivery, evaluating existing products serves as a crucial scaffolding step for design critique, building confidence before students evaluate peer work.
I will share practical strategies for implementing longer debriefs—including establishing classroom routines, using concept prompts, and managing open-ended dialogue.

Context to Coordination: Tacit Knowledge and Situated Assistance in Creative Technology Practice

Alice Tang

Nikolas Martelaro, Daragh Byrne

Computational systems increasingly offer assistance within creative practice, which relies on tacit knowledge that may be difficult to articulate, capture, or interpret from observable activity. Emerging multimodal and generative AI systems create new possibilities for context-aware assistance by making more of practitioners’ ongoing activity available to computational systems. We investigate how tacit knowledge becomes consequential for assistance through a multimodal Wizard-of-Oz study, in which experienced creative-technology practitioners and instructors assisted practitioners engaged in self-directed work. Paired retrospective accounts showed that tacit knowledge was often enacted through observable activity without making practitioners’ intentions, rationales, or trajectories directly legible. Appropriate assistance depended on interpreting what mattered within the practitioner’s unfolding trajectory, iteratively constructing and repairing understandings of the practitioner and their work, and negotiating whether and how to intervene. These findings suggest that supporting tacit practice requires coordinating around partial understandings, beyond capturing more context or requiring practitioners to explicate tacit knowledge.

Enhancing Active Learning Through Structured Peer Collaboration

Xiaoying Tu

Pensky, Allison, Eberly Center; Pincus, Jacqui, Eberly Center

This presentation shares a research-informed redesign of active learning in a graduate database course serving primarily students from non-technical backgrounds. The course already incorporated active learning through in-class polling, hands-on SQL exercises, and exam review activities, but peer collaboration was largely optional and informal. Rather than adding new activities, I redesigned these existing activities to make collaboration intentional and consistent through structured peer discussion, shared materials, and pair programming. I evaluated the redesign by comparing student learning across two course offerings and found higher performance after collaborative learning was introduced. The presentation will focus on the practical strategies behind the redesign: how I established collaborative norms, structured peer interaction, and adapted existing activities with relatively modest changes.

Helping Musicians Build Research-Creation Capabilities and Curiosity

Alexa Woloshyn

Undergraduate music majors tend to think of themselves primarily as creators who perform and/or compose. The idea that research is or could be part of their creative practice has been met with skepticism and even hostility in the past.

Through my PITF course revision, I targeted “research” as a core value in the required Music History 3. In this informal conversation, I will focus on two interventions: a scaffolded set of research skills assignments and a research-creation modalities framework for our case studies and special guests.

This conversation is primarily targeted to those who teach in practice-oriented disciplines, such as the arts and languages, where students may have narrow conceptions of “research, “ and, therefore, question its relevancy.

During the discussion, I will also share some of the pre-/post-course survey results that underline the efficacy of my efforts to build research skills and deepen the role they envision for research in their studies.

(re)Membering First Year Studio in Radical Empathy
Transparency, Universal Design for Learning, and Plural Voices

Tommy Yang

This paper examines, “Radical Empathy in Architecture”, the first year spring undergraduate studio that reframes architectural fundamentals as a socially engaged, relational practice grounded in creative citizenship, care, and radical empathy (Michelle Caswell and Cifor, 2016). The studio foregrounds the entangled nature of the design process, asking first year designers to move recursively between precedent (searching widely, imagining openly), narrative, representation and compositional explorations so that design emerges as an ongoing inquiry of making worlds (Federico Campagna, 2021).
Across four years, the studio cultivated a culture that privileged plural voices and peer-to-peer learning as central conditions of architectural education. Studio “family” agreements, structured reflection, and rotating roles in giving feedback position students as co-authors of studio norms and evaluative frameworks. This challenges hierarchical and opaque studio traditions. The pedagogy adapts Transparency in Learning and Teaching (TILT) to clarify purposes, tasks, and criteria for the studio, and incorporates Universal Design for learning (IDL) to diversify modes of participation, representation and expression.

Understanding motivation for taking an elective circuits class with EVT

Tom Zajdel

Allison Connell Pensky

I teach an elective introductory electronics course for non-engineering majors. I would like to better understand how is motivation to pursue electronics study has been impacted by their experience in the class. Expectancy-value theory (EVT) posits that motivation for a student to pursue a task is the combination of the expectancy of a positive outcome and the value they place on that success, and this has been the guiding framework for this study. Students taking this course in Fall 2025 and Spring 2026 (n = 27) completed a survey measuring self-efficacy (0-100 scale) and three subjective task value components: intrinsic, attainment, and utility value (7-point Likert scales) at multiple time points. Results showed self-efficacy increased significantly with a large effect size. Value ratings remained stable between midterm and end of course after a self-driven final project, suggesting that a utility value-building intervention may be added to make this experience more effective at building student motivation.

Beyond “Just Use ChatGPT”: Three Strategies for Effective AI-Supported Learning 

Bo Zhan

This project grew from a challenge in my first-semester Japanese course: how could students receive meaningful speaking practice beyond limited classroom time? I developed “Sato,” a customized ChatGPT speaking partner aligned with students’ proficiency level, course content, and learning goals. However, classroom implementation revealed that simply providing students with an AI tool was not enough. Students needed guidance in how to communicate with AI, evaluate its responses, and respond when interactions did not work as expected. Based on these experiences and findings from my classroom research, this presentation shares three transferable strategies for AI-supported teaching: design AI activities around specific learning objectives, explicitly teach students how to interact critically and effectively with AI, and incorporate structured reflection on their AI experiences. These strategies shift the focus from adopting AI technology itself to designing thoughtful learning experiences and can be adapted by instructors across disciplines who are exploring generative AI in their courses.