Opinion: Ensuring student learning with AI-enhanced education

The Utah State Board of Education announced a statewide partnership with Google's Gemini for Education to integrate AI into classrooms, but an opinion piece argues that without safeguards such as proctored exams, students can use AI to bypass learning. The author, citing Gemini's own response, states that no software can prevent students from using external AI, and recommends controlled, in-class assessments as the only valid way to ensure learning.

Within education, from K-12 to university, AI integration is top of mind. The Utah State Board of Education recently announced https://www.ksl.com/article/news/utah/science-and-tech/utah-introduces-ai-tools-in-classrooms-statewide/51506597 a statewide partnership with Google’s Gemini for Education https://www.deseret.com/opinion/2026/06/10/ai-in-classrooms-protecting-utah-kids/ to take steps toward achieving this purpose. At the university level, phrases like “AI-enhanced” and “AI literacy” are becoming ubiquitous. With all this, educators must not overlook the importance of confirming whether students are learning the material they are being taught — a concept sometimes described as assessment or assurance of learning. Educators must also recognize that AI enhancement and integration notwithstanding, without appropriate safeguards in place, students can use AI to utterly bypass learning. How do we make sure students are actually learning? Given the State Board of Education’s adoption of Gemini for Education, I asked Gemini about it, and inquired: “What prevents a student from using AI to simply answer the questions posed in assignments and quizzes?” Gemini responded: “Here is the straight answer: no software can physically stop a student from pulling out a phone or opening another browser tab to query an external AI. Gemini itself cannot police a student’s screen.” It went on to explain that restrictions can be imposed on district-managed devices and that assessment approaches can be adapted to be class-specific. It concluded though that “high-stakes” evaluations must occur in “controlled, in-class environments oral defenses, hands-on demonstrations, or proctored locked-screen exams .” Let’s consider the feasibility of each of these approaches to learning assessment. Oral defenses are impracticable at nearly all levels of education outside a PhD program or elite and expensive private schools where faculty-student ratios are measured in single digits. Hands-on demonstrations may be more feasible in certain academic disciplines and settings. Exercises that require students to engage in face-to-face interaction with others, make presentations with real-time Q&A, or demonstrate competence through practicum, may still lead to generally valid assessment. And to be sure, experiential learning should be encouraged more than ever to help students develop their human-side skills in an era in which screen-based learning is so prevalent. But generally, these types of hands-on demonstrations are instructionally labor intensive and impractical in learning environments with large cohorts of students. As such, this approach will not lend itself to many academic situations. This leaves proctored locked-screen exams, or even, if you like, using Scantrons and blue books to achieve the same result. This is the only way to validly determine whether instruction, enhanced by AI or not, is actually leading to student learning. If this type of checkpoint to learning is not installed, the implications are serious. Human tendency is like water: it takes the path of least resistance. If students know that there is no actual point at which they will be responsible for the material taught in class, many or most will simply plug unproctored assignments, quizzes and exams into their AI tool of choice and completely undermine their own learning. Emphasizing the importance of learning Do we want high school graduates who read at a fourth-grade level? Do employers want employees with college degrees who never gained the cognitive skills, including critical thinking, a college education purports to deliver? As school starts this year for K-12, parents would do well to talk with their children about the appropriate use of AI. The temptation to use AI to bypass learning is real and ought to be acknowledged and addressed, even as AI attempts to enhance the learning process through products like Gemini for Education. Parents would do even better to talk with their children’s teachers about whether student learning is ultimately being measured through proctored locked-screen evaluation tools. College students should similarly evaluate their professors to determine whether learning will be validly assessed. It should be a warning signal if the professor’s syllabus has take-home exams or bases large parts of the student’s grade on unproctored projects or assignments. This is even more so if class grading must adhere to standardized means and distributions. Finally, this is a call to educators and educational administrators at all levels, including and specifically the Utah State Board of Education and the Utah Board of Higher Education: AI integration must be accompanied by AI disintegration when it comes to measuring whether or not students are learning. This requires top-down policy development and implementation that appropriately equips and incentivizes educators to validly assess their students’ learning. All the AI enhancement in the world means nothing — and our students run the real risk of squandering their education — if safeguards are not in place to ensure this.