Across higher education, artificial intelligence has rapidly changed how students complete academic work. Essays, summaries, explanations, and structured problem-solving can now be generated instantly. The first concern has understandably been academic integrity. A deeper issue, however, is emerging.
We are beginning to mistake performance for understanding.
Students can now produce correct answers without necessarily developing the ability to interpret situations, make decisions, or apply knowledge responsibly. This is not simply a cheating problem. It is a learning problem. Our systems were designed to evaluate output, and artificial intelligence now produces output with remarkable fluency. Large language models are capable of generating human-like responses across many academic tasks, raising concerns about how learning can be evaluated when performance can be externally produced (Kasneci et al., 2023; Zhai, 2023).
The risk is that we may be evaluating the tool rather than the learner.
What Faculty Are Noticing
Many instructors already sense this shift. A student submits strong written responses yet struggles when asked to explain reasoning, adapt to a new scenario, or justify a choice. The student appears competent on paper but uncertain in conversation.
Research increasingly reflects this experience. Students using generative AI may complete assignments successfully while demonstrating weaker conceptual transfer when asked to independently apply ideas (Zhai, 2023; Mollick & Mollick, 2023). The work looks correct, but the understanding is fragile.
This becomes especially visible in professional education. In fields such as healthcare, teaching, and leadership, knowledge is not merely recalling information. A physician must interpret incomplete evidence. A teacher must respond to unpredictable classroom behavior. A leader must act under uncertainty.
These activities require judgment.
Artificial intelligence can generate explanations, but it does not develop understanding through participation or consequence. When assessment focuses primarily on written output, AI unintentionally separates performance from comprehension.
Why Our Assessments Are Breaking
For decades, education has measured learning through visible products: essays, quizzes, discussion posts, and standardized responses. These approaches worked because producing the work required the learner to do the thinking.
Generative AI changes that assumption.
AI systems produce language by predicting patterns across massive datasets. They do not interpret meaning, evaluate consequences, or assume responsibility for decisions. Yet many assessments only ask whether a response is coherent and correct.
Researchers now warn that AI-generated responses challenge traditional assessment validity because written output can no longer reliably indicate individual cognition (Perkins et al., 2024). When an assignment can be completed successfully without engaging the intended mental processes, the assessment is no longer measuring learning.
Correct answers do not always indicate understanding.
What Intelligence Actually Requires
Educational research consistently shows that understanding develops through application, reflection, and contextual use of knowledge rather than exposure to information alone (Luckin et al., 2016). Learners construct meaning when they must interpret situations, not when they only reproduce explanations.
I describe this form of learning as Experiential Intelligence: the capacity to interpret situations, reflect on outcomes, and make responsible decisions using knowledge.
This kind of understanding develops when students must:
explain their reasoning adapt to unfamiliar scenarios respond to consequences
In other words, understanding appears when knowledge is used, not when it is displayed.
Artificial intelligence can assist learning by organizing information and generating explanations. However, it cannot engage in lived situations, revise beliefs after consequences, or take ownership of decisions. Those processes remain human.
What This Means for Teaching
The presence of AI does not make traditional assignments useless. It does mean they are no longer sufficient as primary evidence of learning.
Faculty may need to shift from evaluating products to evaluating thinking.
Scholars in AI-supported education now recommend oral defenses, authentic tasks, and iterative feedback as more reliable measures of learning than static written submissions (Perkins et al., 2024; Mollick & Mollick, 2023). When students must explain how they reached an answer, instructors can observe reasoning rather than production.
Practical adjustments may include:
oral explanations of written work case-based or scenario-based exercises reflective reasoning assignments requiring students to justify decisions
These approaches do not eliminate AI. They place learning where AI cannot substitute: interpretation, reasoning, and responsibility.
The Opportunity
Artificial intelligence is often framed as a threat to education. It may instead be a clarifying moment. For years, educators have debated what students should gain from a course. Memorized information fades quickly. Correct answers can now be generated instantly. What remains valuable is the ability to use knowledge wisely.
AI exposes a distinction that has always existed: education is not only about acquiring information. It is about forming judgment. If students can complete an assignment without thinking, the assignment is measuring production rather than learning. The challenge for educators is not preventing AI use but designing learning that requires understanding.
Artificial intelligence can generate responses. Education must develop thinkers.
Dr. Lydia Elliott is the Director of Faculty Development at Carle Illinois College of Medicine at the University of Illinois in Urbana-Champaign, and the creator of Experiential Intelligence, a framework describing how people develop judgment and understanding through lived experience. She works in medical education supporting faculty teaching, feedback, and assessment practices, and her work focuses on learning and decision-making in an AI-influenced world.
References
Kasneci, E., Sessler, K., Küchemann, S., Bannert, M., Dementieva, D., Fischer, F., … Kasneci, G. (2023). ChatGPT for good? On opportunities and challenges of large language models for education. Learning and Individual Differences, 103, 102274. https://doi.org/10.1016/j.lindif.2023.102274
Luckin, R., Holmes, W., Griffiths, M., & Forcier, L. B. (2016). Intelligence unleashed: An argument for AI in education. Pearson.
Mollick, E., & Mollick, L. (2023). Assigning AI: Seven approaches for students, with prompts. The Wharton School Research Paper. https://ssrn.com/abstract=4475995
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 and Learning Practice, 21(6), Article 06. https://doi.org/10.53761/q3azde36
Zhai, X. (2023). ChatGPT for next generation science learning. SSRN Electronic Journal. https://ssrn.com/abstract=4331313
