Bridging the AI Engagement Gap

Getting students to engage meaningfully with course material has always taken thoughtful, intentional design. AI didn’t create that challenge, but it’s made it harder to ignore.

What’s different now

Compliance-based motivation is less reliable than it once was. When students could fake preparation by skimming, some did. Now AI extends that capacity - more easily and more convincingly. This shift also introduces a motivational logic worth taking seriously: if AI can generate a plausible response, why invest effort in sustained reading? That’s a rational response to a changed environment, and it requires a genuine pedagogical response. Research on academic integrity in the post-ChatGPT classroom suggests the answer lies not in detection or prohibition, but in redesigning what we ask students to do (Cotton et al., 2023).

Surface engagement is also easier to simulate. Discussion contributions, written reflections, and even case analyses can be AI-assisted in ways that appear to be engagement but are not. That makes authentic engagement design more important, not less.

What has not changed

Students still want to learn things that are genuinely meaningful to them. The evidence base for active learning – across hundreds of studies – demonstrates that students in well-designed active learning environments outperform peers in traditional lecture courses on both exams and conceptual understanding (Freeman et al., 2014). Active learning also narrows achievement gaps for underrepresented students (Theobald et al., 2020), a meaningful consideration for CBS’s diverse classrooms.

Curiosity, when activated, remains a strong driver of preparation and participation. Self-determination theory helps explain why: students who experience their learning as autonomous and personally relevant exhibit more intrinsic motivation and deeper engagement (Ryan & Deci, 2000; Ryan & Deci, 2020). That dynamic has not changed – AI has just made it more urgent to design for it.

The design implication

The shift is from “students must prepare to avoid looking bad” toward “students want to prepare because the work serves a purpose they value.” That is a higher bar to reach and it is also more durable.

Framing activities and assignments as “AI-resistant” signals that the priority is detecting misconduct rather than supporting learning. A more productive framing is this: the features that render an assignment genuinely meaningful often also make it difficult to shortcut. The design principles below apply to both in-class work and take-home assignments. They increase meaningful engagement while also reducing reliance on AI—not through clever workarounds, but by requiring sustained thinking.

Framing AI use with students

Before setting restrictions, it helps to establish an explicit conversation with students about why independent thinking and problem-solving matter, not as a policy statement, but as a genuine explanation. The goal is for students to understand that grappling with a concept, testing an approach, revising when it does not work, and putting their reasoning into words are not peripheral tasks. They constitute actual practice. AI can generate an answer quickly; it cannot do the cognitive work that builds judgment.

Apply the same logic to assignments that you use for in-class activities. When students understand that the work builds their capacity, not just gives them something to submit, they approach conversations about AI differently. Be clear about the habits students are being asked to practice, such as analysis, persistence, precision, and synthesis. These are the skills they will rely on when they need to interpret, evaluate, and defend their own thinking in real time.

That framing will not remove every temptation to take shortcuts, but it makes the work's purpose visible, and that is a meaningful place to start.

Designing for in-class performance

The most durable response to AI in the classroom isn’t a better assignment prompt. It’s making the classroom itself the primary site of intellectual work – where the real learning is enacted and where AI can’t go. This classroom shift is the core logic of the flipped classroom: move content delivery out of class time and reserve class time for application, discussion, and sense-making. The assumption is that passive reception of information doesn’t require a room full of people, but working through the implications of that information, arguing about it, testing it against real cases, and defending a position in real time does.

In a CBS context, most faculty are already doing a version of this. Case-method teaching, cold calls, think-pair-share activities, and role-play negotiations are fundamentally in-class performance structures. AI intensifies the rationale for leaning into them. If a student can produce a competent case analysis from their laptop the night before, the question of what class adds becomes urgent. The answer has to be: something that can only happen in the room. When the quality of a student’s in-class contribution is what matters – not a document submitted after the fact – AI has nowhere to go. For a deeper look at flipped classroom design and its evidence base, see this Blended Learning resource from the Columbia Center for Teaching and Learning. They feature faculty experiences with flipped classrooms.

Redesigning assignment questions

For work that happens outside the classroom, the design of the question itself is the primary lever. The goal isn’t to make AI impossible to use – it is to render AI output insufficient. If a student could submit a generic AI response and get full credit, the prompt requires redesigning. The adjustments below are intentionally small. Faculty can apply them to existing assignments without rebuilding from scratch. 

Additional resource:  Engaging Strategies that Still Work. These strategies make preparation purposeful and establish the classroom as the site where student thinking, not AI-generated output, carries weight.

A note on AI detection

Detection tools are unreliable (Weber-Wulff et al., 2023) and systematically misclassify non-native English writers (Liang et al., 2023) - a significant equity concern at CBS, where many students write in a second or third language. A more durable approach centers on designing activities in which AI-generated responses are clearly insufficient and student thinking carries clear value. If an AI-generated response would receive full credit, the prompt requires redesign.

The same tools students use to shortcut their work can also support faculty in designing more effective activities, more efficiently. A national survey of over 2,600 instructors found that nearly three-quarters had experimented with AI for teaching (Blankstein & Love, 2024), most often to generate content, formative questions, and variations in course materials. The key is to position AI as a thought partner and draft generator—not a substitute for pedagogical judgment (Kasneci et al., 2023). The sections below map each phase of ReCAP to concrete uses of AI.

CBS tools to use

Gemini Gem (CBS ReCAP Gem): The fastest starting point. The ReCAP Gem comes pre-configured for CBS faculty. Enter the course topic, session goal, and student level, and it generates draft activities for each ReCAP phase in under a minute. Use it as a springboard, not a finished product.

All Day TA: CBS’s AI-powered course assistant, currently in pilot. It supports student questions outside class time and frees up class sessions for higher-order discussion. Reach out to the Samberg team to explore use cases for a specific course.

Other AI tools faculty are using

ChatGPT / Claude:  Useful for generating scenario variations, discussion questions, and rubric drafts. These tools work best with a specific prompt that includes course level, topic, and what students should do, not just what they should know.

NotebookLM (Google):  Particularly helpful for generating questions and summaries from course documents. Upload a case or article and ask it to create discussion questions tied to that text. It stays grounded in the materials provided.

A note on AI-generated activities: Every draft needs review for accuracy, appropriateness to students’ level, and alignment with specific learning goals. The strongest AI-assisted activities combine AI’s speed with course knowledge and pedagogical judgment. The tool produces a useful draft quickly; faculty shape it into something that fits the course.

Blankstein, M., & Love, S. (2024). Generative AI and postsecondary instructional practices: Findings from a national survey of instructors. Ithaka S’R. https://doi.org/10.18665/sr.320892

Carini, R. M., Kuh, G. D., & Klein, S. P. (2006). Student engagement and student learning: Testing the linkages. Research in Higher Education, 47, 1–32. https://doi.org/10.1007/s11162-005-8150-9

Cotton, D. R. E., Cotton, P. A., & Shipway, J. R. (2023). Chatting and cheating: Ensuring academic integrity in the era of ChatGPT. Innovations in Education and Teaching International, 61(6), 1–12. https://doi.org/10.1080/14703297.2023.2190148

Eaton, S. E. (2024). How do we maintain academic integrity in the ChatGPT era? Liberal Education, 110(1). Association of American Colleges & Universities. https://www.aacu.org/liberaleducation/articles/how-do-we-maintain-academic-integrity-in-the-chatgpt-era

Freeman, S., Eddy, S. L., McDonough, M., Smith, M. K., Okoroafor, N., Jordt, H., & Wenderoth, M. P. (2014). Active learning increases student performance in science, engineering, and mathematics. Proceedings of the National Academy of Sciences, 111(23), 8410–8415. https://doi.org/10.1073/pnas.1319030111

Kasneci, E., Sesler, K., Kuchemann, S., Bannert, M., et al. (2023). ChatGPT for good? On opportunities and challenges of large language models for education. Learning and Individual Differences, 103, Article 102274. https://doi.org/10.1016/j.lindif.2023.102274

Liang, W., Yuksekgonul, M., Mao, Y., Wu, E., & Zou, J. (2023). GPT detectors are biased against non-native English writers. Patterns, 4(7), Article 100779. https://doi.org/10.1016/j.patter.2023.100779

Lyman, F. T. (1981). The responsive classroom discussion: The inclusion of all students. In A. S. Anderson (Ed.), Mainstreaming digest (pp. 109–113). University of Maryland, College of Education.

Ryan, R. M., & Deci, E. L. (2000). Self-determination theory and the facilitation of intrinsic motivation, social development, and well-being. American Psychologist, 55(1), 68–78. https://doi.org/10.1037/0003-066X.55.1.68

Ryan, R. M., & Deci, E. L. (2020). Intrinsic and extrinsic motivation from a self-determination theory perspective. Contemporary Educational Psychology, 61, Article 101860. https://doi.org/10.1016/j.cedpsych.2020.101860

Theobald, E. J., Hill, M. J., Tran, E., et al. (2020). Active learning narrows achievement gaps for underrepresented students in undergraduate science, technology, engineering, and math. Proceedings of the National Academy of Sciences, 117(12), 6476–6483. https://doi.org/10.1073/pnas.1916903117

Weber-Wulff, D., Anohina-Naumeca, A., Bjelobaba, S., Foltynek, T., Guerrero-Dib, J., Popoola, O., Sigut, P., & Waddington, L. (2023). Testing of detection tools for AI-generated text. International Journal for Educational Integrity, 19(1), Article 26. https://doi.org/10.1007/s40979-023-00146-z