Dr. Eric Loepp is a Professor in the Department of Politics, Government, and Law and Co-Director of the Center for the Advancement of Teaching, Learning, Scholarship, and Technology at the University of Wisconsin-Whitewater.
The jury may still be out on the long-term implications of artificial intelligence (AI) in higher education, but the age of Google and Wikipedia has already shown us that low-discretion assessments serve little pedagogical value when the knowledge they test is readily available and requires virtually no effort. For instance, in my field of political science, I don’t find the need to quiz students on their knowledge of how many Electoral College votes particular states get in presidential contests or which candidates won recent elections, since they can easily find the information online. Indeed, AI even does a reasonably good job of organizing coherent output related to more advanced topics, such as making the case for keeping the Electoral College versus amending it in some way. So how, exactly, should the Electoral College—or countless other subjects—be taught given the new knowledge floor available at our fingertips? What is it that students should learn? What should we be assessing?
One answer to these questions is that instructors should create highly personalized learning experiences in which students focus on the higher rungs of Bloom’s taxonomy. Though AI can mimic personal reflection, it produces considerably less helpful output when students are asked to incorporate unique experiences, local contexts, personal decision-making processes, or hypothetical pathways.
More important, though, personalized assignments shift the focus from generating a product to demonstrating learning. Students are asked not simply what they know, but how they arrived at their understanding. When we design assignments that require a student’s unique lived experience, specific local context, and distinct cognitive process, we do more than just “AI-proof” our assessments—we create deeper, more meaningful learning experiences.
Here are five specific pedagogical strategies for personalizing assignments, along with examples from my own discipline:
1. Embrace Course-Specific Prompts
Instead of asking for a summary of a common topic, craft prompts that require the synthesis of highly specific, idiosyncratic course materials. For instance, an instructor may ask students to synthesize a core concept from the textbook, a key point made by a guest speaker in a recent lecture, and/or a specific current event from the past week. By anchoring the assignment strictly within the shared, real-time ecosystem of your classroom, you compel the student to use their own judgment to prioritize.
| Less personalized | More personalized |
| Explain the role of the filibuster in the United States Senate and argue whether it should be abolished. | Synthesize Sarah Binder’s argument about minority party obstruction from our Week 3 readings with the specific parliamentary maneuvers used in the Senate last week regarding the current infrastructure bill. Contrast this dynamic with the insights our guest speaker shared in Thursday’s lecture about the shifting norms of bipartisanship. Binder, S. A. (2003). Stalemate: Causes and consequences of legislative gridlock. Brookings Institution Press. |
2. Ground the Work in Personal Context and Reflection
In many cases, AI can generate a passable, reasonably accurate essay on the suffrage movement and major moments in the history of voting rights. What it cannot do well is explain how those implications intersect with a student’s own family history, their values, their aspirations to engage in civic life, and their normative perceptions on the degree to which the ideals of these movements are reflected in the lived reality Americans experience today. Personalization shifts the focus of knowledge from the objective to the subjective. When assignments require students to act as the primary source of their own contextualization, AI can become an unreliable substitute.
Instructors can integrate lived experience components into traditional assignments. If you are teaching a sociology course on urban development, ask students to apply a theoretical framework to the specific neighborhood they grew up in. If you are teaching business management, ask them to reflect on a specific conflict they experienced in a past part-time job and analyze it using the week’s readings. This not only mitigates AI use but also operationalizes culturally responsive pedagogy by valuing the diverse backgrounds your students bring to the classroom.
| Less personalized | More personalized |
| Define political socialization and identify the primary agents that shape an individual’s partisan identification (PID). | Map your own political socialization. Conduct a short interview with a family member or community mentor who influenced your early civic views. Which of the three models of PID from Chapter 4 are most consistent with the development of your partisan identity? |
3. Focus on the Process, Not Just the Final Product
Generative AI is product-oriented; it delivers a polished final artifact in seconds. It bypasses the messy, nonlinear, and fundamentally human process of learning—the brainstorming, the dead ends, the revisions, and the breakthrough moments. Recalibrating grading schemes to emphasize and evaluate the process of creation can help minimize the temptation to engage in unproductive AI use.
One way to do this is to implement scaffolded assessments. Break a major term paper or project into a series of smaller, iterative deliverables: a topic proposal, an annotated bibliography, an outline, a rough draft, and a peer-review workshop. Most important, require metacognitive reflections at each stage. Ask students to write a brief paragraph explaining why they chose a specific source, how their thesis evolved after peer feedback, or what the most challenging part of drafting their argument was. While an AI can write an essay, faking the ongoing, week-by-week cognitive struggle of drafting that essay is highly impractical for the student—and relatively easy for an instructor to spot if attempted.
| Less personalized | More personalized |
| Write a 10-page research paper on the factors that influence voter turnout and campaign strategy in American elections. | Week 3: Submit a 1-page proposal selecting a specific, highly competitive state or congressional race happening during the current election cycle (e.g., a swing US House district in your home state). Detail the district’s demographic makeup and recent historical voting margins using Census data. Week 5: Submit a “messaging and mobilization map.” Track specific campaign ads, debate sound bites, and social media posts from both candidates over the past three weeks. Identify exactly which local voter coalitions each campaign is attempting to activate. Week 7: Submit an annotated bibliography that synthesizes recent localized polling data, recent FEC campaign finance filings for both candidates, and three local news editorials. Week 8: Before the election, present your final predictive analysis of the race’s outcome to the class. This must be accompanied by a 500-word reflection explaining how a specific, unexpected event (e.g., an “October surprise,” a sudden shift in polling, or peer feedback during our Week 12 workshop) forced you to pivot or revise your initial hypothesis from Week 3. |
4. Leverage Hyperlocal Current Topics
Most large language models have a knowledge cutoff date, and even those connected to the live internet struggle to grasp the nuance of hyperlocal, community-specific issues, many of which are evolving in real time. They are trained on macro-level data, which makes micro-level analysis a distinctively human advantage.
To do this, instructors may shift the focus of assignments from historical or global abstracts to local, immediate realities. For instance, instead of asking students to write a generic marketing plan for a hypothetical company, have them engage directly with the community by partnering with a local small business and develop a strategy based on an in-person interview with the owner. Instead of an environmental science paper on global coral bleaching, ask them to analyze the specific water quality data of the river running through town.
| Less personalized | More personalized |
| Explain how redistricting and gerrymandering impact the incumbency advantage and voter turnout in U.S. House elections. | Using an online mapping tool (e.g., Dave’s Redistricting App) and recent county-level election results and/or Census data, select a state and redraw its congressional districts to achieve a specific, skewed political outcome that creates a maximum partisan gerrymander for one political party. Accompany your digital map with an analysis explaining exactly which local cities, counties, or neighborhoods you chose to “pack” or “crack” to achieve your goal, and how the physical geography of your state’s voter distribution made this task easy or difficult. |
5. Shift to Multimodal and Interactive Formats
The text-based essay has long served as a primary tool that instructors use to assess student learning, yet it is precisely the format that text-based AI models can (largely) replicate with ease. Written essays can still play a vital role in assessment, but in some contexts diversifying the ways students demonstrate competency is a powerful way to personalize learning and minimize unauthorized AI assistance.
Multimodal assessments support this goal. Rather than written deliverables, students can create a podcast episode, design an infographic, curate a digital portfolio, or deliver a live presentation followed by a Q&A session. While students might still use AI to brainstorm or outline these projects—which can be a valid use of the tool if the syllabus permits it—the final execution requires their own voice, their own visual design choices, and their own real-time cognitive agility. Interactive assessments, such as town-hall-style debates or in-class oral exams, provide immediate visibility into a student’s true understanding of the material.
| Less personalized | More personalized |
| Write a research paper comparing realist and liberal theories of international relations. | In-class crisis simulation. You will be assigned to represent a specific nation-state during a fictional unfolding geopolitical crisis introduced at the start of class. You must: Draft a rapid-response policy memo (written in class without internet access) outlining your state’s immediate objectives. Participate in a 45-minute live negotiation session with other “state actors.” Record a 3-minute post-simulation audio reflection via our class podcast analyzing which international relations theory best explains why the negotiations succeeded or broke down. |
Further Considerations
Of course, several crucial caveats are necessary: First, the strategies described above require a considerable investment in time and energy on the part of instructors. Full stop. Furthermore, some activities may not be applicable to some courses due to the nature of the content, the size of the class, technological limitations, or other reasons.1
To the extent we do reimagine our teaching and learning practices, though, it is important to underscore that instructors need not redesign every aspect of a course all at once. A more realistic approach is to identify one or two assessments each year that are both important to student learning and particularly vulnerable to AI completion. This may mean revising a term paper to include a proposal, checkpoint, and reflection. Or it may mean replacing a multiple-choice quiz with a short podcast or video presentation. In some cases, individual activities may be reworked into team exercises that reduce the total number of submissions to review. Incremental changes to a small number of high-impact assignments can produce meaningful improvements without requiring wholesale course redesign.
It’s also important to underscore that personalization should not be treated as a panacea to AI concerns. Technology is improving every day, and generative AI can already simulate personal reflection and contextual reasoning. Our goal, therefore, cannot be to fully AI-proof our assessments, but to create AI-mindful assignments that center our assessment on pertinent learning goals in the age of ubiquitous AI. Authentic engagement is harder to fake when assignments require personal context, supporting evidence, instructor interaction, and iterative reflection.
Finally, although personalization can help mitigate unproductive AI use, we should also consider the potential pedagogical benefits of AI and use personalization strategies to augment rather than suppress them. AI literacy is increasingly recognized as a foundational student skill. With agentic AI—where AI can serve as a tutor, research assistant, and colleague to students—on the horizon, personalized experiences can also help prepare students for an AI-enabled workplace where human subjectivity, intent, and unique experiences shape agentic workflows.
Conclusion
Indeed, AI can help us do more with the precious and limited time we get to spend with our students, human-to-human. More than one student has shared with me that AI helped them clarify their understanding of key concepts and come to class with greater confidence and interest in engaging with their colleagues. This, in turn, creates time and space in the classroom to interact thoughtfully with the big questions that require a human component—debating the limits of free speech, pondering changes to electoral rules and processes, establishing an appropriate balance between liberty and security, among countless others. In many ways, AI is creating space for a larger human element in our courses—but to use it well may require a steady evolution of our assessment strategy.
Suggested Readings
Assessment Redesign and AI
Dawson, P., Bearman, M., Dollinger, M., & Boud, D. (2024). Validity Matters More Than Cheating: Reframing Assessment in the Age of Generative AI. Assessment & Evaluation in Higher Education. DOI: https://www.tandfonline.com/doi/full/10.1080/02602938.2024.2386662
Mollick, E. (2024). Co-Intelligence: Living and Working with AI. New York: Portfolio. https://www.penguinrandomhouse.com/books/741805/co-intelligence-by-ethan-mollick/
Perkins, M., Furze, L., Roe, J., & MacVaugh, J. (2024). The Artificial Intelligence Assessment Scale (AIAS): A Framework for Ethical Integration of Generative AI in Educational Assessment. Journal of University Teaching & Learning Practice. https://arxiv.org/pdf/2312.07086
Essays, Writing, and Assessment in the AI Era
Midford, S. (2025). Is the Essay History? Rethinking Effective Assessment in the Age of Generative AI. History Australia.
https://www.tandfonline.com/doi/full/10.1080/14490854.2025.2570457
Syska, A. (2025). We Tried to Kill the Essay—Now Let’s Resurrect It. LSE Higher Education Blog.
https://blogs.lse.ac.uk/highereducation/2025/02/27/we-tried-to-kill-the-essay-now-lets-resurrect-it/
Warner, J. (2025). More Than Words: How Writing Is Thinking in the Age of AI.
https://www.hachettebookgroup.com/titles/john-warner/more-than-words/9781541605510/?lens=basic-books
Higher-Ed Commentary on Writing and AI
Villasenor, J. (2025). AI Has Rendered Traditional Writing Skills Obsolete. Education Needs to Adapt. Brookings Institution.
https://www.brookings.edu/articles/ai-has-rendered-traditional-writing-skills-obsolete-education-needs-to-adapt/
Hsu, H. (2025). What Happens After AI Destroys College Writing? The New Yorker. https://www.newyorker.com/magazine/2025/07/07/the-end-of-the-english-paper
Mollick, E. (2024–Present). One Useful Thing (blog). https://www.oneusefulthing.org
MEET THE AUTHOR
Dr. Eric Loepp is a Professor in the Department of Politics, Government, and Law and Co-Director of the Center for the Advancement of Teaching, Learning, Scholarship, and Technology at the University of Wisconsin-Whitewater. He teaches courses and conducts research in American government, political behavior, and research methods. He is the recipient of the American Political Science Association’s 2018 CQ Press Award for Teaching Innovation.
- It is imperative that universities recognize and support initiatives that promote assessment personalization. We cannot do this alone.” ↩︎