What AI is doing to teaching, assessment, and hiring, and the cross-institutional conversations we need to keep pace.
See the full agenda Eleven talks · two keynotesThe program sits at the intersection of how we teach with AI, how we assess what students have learned, what the labor market is asking of graduates, and the peer networks that make any of this tractable across institutions.
The talks are arranged against three of those threads. Pedagogy runs through the morning, from AI that gives feedback rather than answers, to proactive tutors, to instructor-built companions and a structured framework for introductory programming. Assessment is the question underneath it: what an applied trade assignment, an applied calculus course, or a programming course should ask for once AI can produce the artifact. Workforce readiness anchors the afternoon, with Michael Horrigan on AI exposure and the future of work and Cassandra Merritt on how students and managers are adjusting, before Justin Wolfers closes the day.
Friday, October 16, 2026 in Brown Hall, Room 2028. Registration opens at 8:45 a.m. and the conference closes at 4:55 p.m. All times Eastern.
Students learn most when they get timely, specific feedback on their own thinking. That has always been the bottleneck: good feedback is slow and expensive, and in large classes most students rarely get enough of it. Generative AI changes the cost of feedback, makes it easily available instantly and 24/7, but does not change its purpose. Feedback should therefore be the organizing principle for using AI in teaching, and I draw lessons from several experiments at the LSE Department of Economics.
The first lesson is that AI helps when it gives feedback/feedforward, not answers. A tutor that asks questions and pushes back keeps students doing the thinking, while one that hands out solutions replaces it. The second is that what we give feedback on has to change. When AI can produce the work, the skills worth practicing are reading, organising and judging that work, so assessment and feedback should target those skills. The third is that AI feedback works best as a hybrid. It adds speed and detail, while people keep judgment and know the student. The fourth is that if AI can perform a task in few minutes, completing that task is not a metric of learning, and cannot be a learning outcome either.
The implications for redesigning Economics teaching and learning practices are straightforward: teach how to think and test reasoning skills, in a setting that calls for knowledge-based decision-making and judgement, where AI is a complement and the timeline for learning and assessing is compressed.
Generative AI is changing what our students need to learn and how they learn it. The same tools that can replace a student’s thinking can also deepen it, and which one happens depends largely on design. In this talk I share how I moved from assigning commercial AI tools to building an AI companion with my students: one that is free for every student, protects their privacy, and teaches the way the instructor decides. I discuss the pedagogy behind it, what I have learned from using it in my courses, and what it suggests about equity, ethics, and trust when we bring AI into the classroom. My goal is to show how instructors can shape these tools instead of only adopting them.
In this interactive activity, students will work in teams to solve a variety of challenges across subjects like math, science, grammar, pop culture, and more, all while competing to earn the highest score. At first, teams will rely solely on human effort and collaboration. But as the rounds progress, they’ll face a choice: should they “hire” AI to help complete tasks more efficiently, even if it comes at a cost?
AI can play a meaningful role in introductory programming when it is used not simply to generate code, but to guide students through the process of thinking, problem-solving, and evaluating their own work. To support this goal, a five-stage framework was developed that integrates an AI-powered learning assistant into problem breakdown, algorithm design, code implementation, self-evaluation and debugging, and reflection and feedback. Implemented in an introductory Python course, the framework connects learning and assessment while promoting accountability and academic integrity. Student feedback highlighted its value in supporting reflection, problem-solving, critical thinking, and confidence. This talk will share how the framework was implemented, lessons learned from the experience, and practical considerations for educators interested in adapting structured AI-supported learning and assessment approaches within their own disciplines.
LLMs can now produce a polished international trade paper, but does it show that a student understands the economic reasoning behind it? This presentation describes the redesign of a major applied assignment in an International Trade course at Lawrence Technological University, intended to teach students to leverage LLMs while requiring them to conduct and defend their own analysis.
Students act as analysts advising a U.S. manufacturer considering international expansion. They evaluate a foreign market using public data, build a financial model of exporter and distributor economics, and analyze the effects of tariffs, currency devaluation, and local assembly. Because no single country choice is correct, assessment focuses on evidence, reasoning, modeling, and recommendations.
To make student thinking visible, the redesign adds a data log of every quantitative input, a live Excel model with formulas, intermediate checkpoints, AI-use disclosure, a brief oral defense, and personalized tariff and exchange-rate shocks. The goal is not an “AI-proof” assignment, but one in which AI augments learning without substituting for the economic reasoning the course is meant to develop. The session shares the design, lessons learned, remaining challenges, and strategies adaptable to other data-intensive courses.
How can AI move beyond novelty to meaningfully support economics teaching and learning? This session examines the research, learning science, and development principles shaping Macmillan Learning’s approach to AI. Attendees will learn how evidence, educator expertise, and responsible design guide the creation of tools that deepen student understanding, support purposeful practice, and encourage active engagement while preserving the instructor’s essential role.
The conference is in person and free to attend. Parking is free, no permit needed, and lunch is provided during the midday keynote.
For GPS, use Miller Auditorium, 1341 Theatre Dr., Kalamazoo, MI 49008. Lot 33 is directly south of it, along Howard Street.
Parking in Lot 33 is free for conference guests on October 16. You do not need a permit, a meter, or the Parkmobile app. Please park only in Lot 33: the free parking does not extend to other campus lots or meters.
Head north beside Miller Auditorium, take the stairs up to Miller Plaza, then turn right toward Brown Hall. Enter on the plaza side and continue to Room 2028, where registration and coffee open at 8:45 a.m. We will post signs along the way.
This route includes stairs. For accessible routes and building entrances, see wmich.edu/maps.
