Creating AI Agent Safeguards (for now) through Pragmatic Instructional Design to Deliver Assignments and Assessments

As you prepare your syllabus for the upcoming semester, a question I’ve explored before looms large. In a previous post, “The Death of the LMS in Higher Ed,” I argued that its role as a simple repository for assignments is becoming obsolete. With AI agents now ubiquitous, how do we prevent the LMS from becoming a mere drop-box for bot-generated work from an autonomous AI Agent?

The solution isn’t to fully abandon the LMS (for now), but to evolve its purpose and create humanizing instructional barriers in the tasks and assessments we deliver to our students. This post presents a few practical strategies (I am implementing this term in my own graduate-level course) to revitalize our courses by shifting the focus from the final product, which AI can generate in seconds, to the learning process, which it cannot. This is going to my attempt to defend against students utilizing AI Agents. While I know this may not be one-hundred percent effective, I do have a hunch it will mitigate AI Agents to a degree, at least for the time being, this semester.

Using AI to Build Powerful Retrieval Practice Activities to Supercharge Student Learning Opportunities

The “testing effect,” a term championed by researchers like Roediger and Karpicke, demonstrates that every time a student actively recalls information, they strengthen the neural pathways associated with it. This effortful retrieval tells the brain that the information is important and makes it easier to access in the future. Ultimately, over time, if students are able to actively recall correctly, this means they likelihood of them learning the content, is much higher. Although the challenge for teachers has always been the time it takes to create a rich variety of material, such as flashcards, practice tests, and quizzes, that facilitate this process.

With powerful AI tools, we can now create engaging, interactive retrieval practice activities in minutes. Tools like Gemini, Claude, and ChatGPT have built-in “Canvas” features that can generate the code for these resources, and you don’t need to be a coding expert to use them.