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\u00a0emerging.\u00a0<\/p>\n
We are beginning to mistake performance for understanding.\u00a0<\/p>\n
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\u00a0are capable of generating\u00a0human-like responses across many academic tasks, raising concerns about how learning can be evaluated when performance can be externally produced (Kasneci\u00a0et al., 2023; Zhai, 2023).\u00a0<\/p>\n
The risk is that we may be evaluating the tool rather than the learner.\u00a0<\/p>\n
What Faculty Are Noticing\u00a0 <\/p>\n
Many instructors already sense this shift. A student\u00a0submits\u00a0strong written responses yet\u00a0struggles\u00a0when asked to explain reasoning, adapt to a new scenario, or justify a choice. The student appears competent on paper but uncertain in conversation.\u00a0<\/p>\n
Research increasingly reflects this experience. Students using generative AI may complete assignments successfully while\u00a0demonstrating\u00a0weaker conceptual transfer when asked to independently apply ideas (Zhai, 2023; Mollick & Mollick, 2023). The work looks correct, but the understanding is fragile.\u00a0<\/p>\n
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.\u00a0<\/p>\n
These activities require judgment.\u00a0<\/p>\n
Artificial intelligence can generate explanations, but it does not develop understanding through participation or\u00a0consequence. When assessment focuses primarily on written output, AI unintentionally separates performance from comprehension.\u00a0<\/p>\n
Why Our Assessments Are Breaking\u00a0 <\/p>\n
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.\u00a0<\/p>\n
Generative AI changes that assumption.\u00a0<\/p>\n
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.\u00a0<\/p>\n
Researchers now warn that AI-generated responses challenge traditional assessment validity because written output can no longer reliably\u00a0indicate\u00a0individual cognition (Perkins et al., 2024). When an assignment can be completed successfully without\u00a0engaging\u00a0the intended mental processes, the assessment is no longer measuring learning.\u00a0<\/p>\n
Correct answers do not always\u00a0indicate\u00a0understanding.\u00a0<\/p>\n
What Intelligence Actually Requires\u00a0 <\/p>\n
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).\u00a0Learners\u00a0construct meaning when they must interpret situations, not when they only reproduce explanations.\u00a0<\/p>\n
I describe this form of learning as Experiential Intelligence: the capacity to interpret situations, reflect on outcomes, and make responsible decisions using knowledge.\u00a0<\/p>\n
This kind of understanding develops when students must:\u00a0<\/p>\n
explain their reasoning\u00a0 adapt to unfamiliar scenarios\u00a0 respond to consequences\u00a0 <\/p>\n
In other words, understanding appears when knowledge is used, not when it is displayed.\u00a0<\/p>\n
Artificial intelligence can\u00a0assist\u00a0learning by organizing information and generating explanations. However, it cannot engage in lived situations, revise beliefs after consequences, or take ownership of decisions. Those processes\u00a0remain\u00a0human.\u00a0<\/p>\n
What This Means for Teaching\u00a0 <\/p>\n
The presence of AI does not make traditional assignments useless. It does mean they are no longer sufficient as primary evidence of learning.\u00a0<\/p>\n
Faculty may need to shift from evaluating products to evaluating thinking.\u00a0<\/p>\n
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\u00a0observe\u00a0reasoning rather than production.\u00a0<\/p>\n
Practical adjustments may include:\u00a0<\/p>\n
oral explanations of written work\u00a0 case-based or scenario-based exercises\u00a0 reflective reasoning assignments\u00a0 requiring students to justify decisions\u00a0 <\/p>\n
These approaches do not\u00a0eliminate\u00a0AI. They place learning where AI cannot\u00a0substitute:\u00a0interpretation, reasoning, and responsibility.\u00a0<\/p>\n
The Opportunity\u00a0 <\/p>\n
Artificial intelligence is often framed as a threat to education. It may instead be a clarifying moment.\u00a0For years, educators have debated what students should gain from a course. Memorized information\u00a0fades quickly. Correct answers can now be generated instantly. What\u00a0remains\u00a0valuable is the ability to use knowledge wisely.\u00a0<\/p>\n
AI exposes a distinction that has always existed: education is not only about\u00a0acquiring\u00a0information. It is about forming\u00a0judgment.\u00a0If students can complete an assignment without thinking, the assignment is measuring production rather than learning.\u00a0The challenge for educators is not preventing AI use but designing learning that requires understanding.\u00a0<\/p>\n
Artificial intelligence can generate responses.\u00a0Education must develop thinkers.\u00a0<\/p>\n
Dr. Lydia Elliott is\u00a0the Director of Faculty Development at Carle Illinois College of Medicine at the University of Illinois in Urbana-Champaign,\u00a0and 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.\u00a0<\/p>\n
References<\/p>\n
Kasneci, E., Sessler, K.,\u00a0K\u00fcchemann, S., Bannert, M., Dementieva, D., Fischer, F., \u2026\u00a0Kasneci, 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<\/a>\u00a0<\/p>\n Luckin, R., Holmes, W., Griffiths, M., & Forcier, L. B. (2016). Intelligence unleashed: An argument for AI in education. Pearson.\u00a0<\/p>\n Mollick, E., & Mollick, L. (2023). Assigning AI: Seven approaches for students, with prompts. The Wharton School Research Paper. https:\/\/ssrn.com\/abstract=4475995<\/a>\u00a0<\/p>\n Perkins, M., Furze, L., Roe, J., &\u00a0MacVaugh, 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.\u00a0https:\/\/doi.org\/10.53761\/q3azde36<\/a>\u00a0<\/p>\n