E-Paper
AI can do the assignment. Can universities still measure learning?
This article is authored by Praphul Chandra, dean (research), Atria University.
Prefer HTon Google
At a recent education conference, the organisers gifted participants a copy of The Hitchhiker's Guide to the Galaxy. I found the gift exceptionally fitting. In the book, a supercomputer determines that the answer to the ultimate question of life, the universe, and everything is 42. The irony is revealing, because the question itself was never known. In many ways, this reflects our present moment. Today, AI can answer a remarkable range of questions, complete much of what we assign, and pass many of our exams. But do we know what questions to ask? Do our students? What does this mean for education? For learning? For universities?
When AI can answer almost any question, the value of human intelligence shifts to asking the right questions. Unfortunately, this is a skill that traditional education systems tend not to prioritise. Throughout formal education, questions are given to students, and they are trained to find the answer. Today, however, finding an answer is as simple as copy-pasting the question into an AI chatbot like ChatGPT or Claude.
How can universities measure learning in terms of a student's ability to question? Fortunately, we have a paradigm to turn to. Liberal arts education, which relies on a pedagogy of class discussions, debates, dialogues, and self-reflection, encourages students to ask questions, to question the answers offered by others, to question the evidence, and to question the question itself. These are critical skills for the age of AI. And liberal arts educators around the world have developed methodologies for evaluating them: From assessing participation in group discussions to setting assignments and exams where the questions have no correct answer, and where students evaluation rubrics grade the quality of reasoning rather than the correctness of a conclusion – and let institutions evaluate open-ended thinking.
There is no reason these skills and pedagogical techniques should remain limited to students of the liberal arts. The ability to question an AI response, to critique a solution it proposes, to verify the data it cites, and to challenge a claim is a critical skill in the age of AI. At Atria University, for example, all students — irrespective of their major — complete courses in critical thinking, data understanding, and critique in their foundation year. This is a form of interdisciplinary education whose time has come.
In STEM streams such as engineering and the life sciences, AI creates another significant challenge. Today's AI models can solve problems in mathematics, physics, biology, computer science, electrical engineering — virtually any field. This not only strains "assignments" and projects as methods of evaluation; it strikes at a deeper question: if AI can solve these problems, why should students learn these subjects at all?
The answer lies in the purpose of STEM education. Whether the goal is to understand how the world around us works, to uncover the laws of nature, or to apply those principles to real problems, STEM education tends to follow an underlying pattern: abstraction to create models of the real world, study of those models, and then application of what we learn from them to real-world challenges. It is this final step where human intelligence will remain critical — primarily because the ambiguity and complexity of real-world are never fully captured in a model, in mathematics, or in code. Nurturing scientists, technologists, and engineers who can bridge science, AI, and real-world problems is the challenge for today's universities.
One way to meet it is to bring real-world challenges into the education system: To expose students to the practical applications of STEM foundations, to give them the opportunity to understand messy real-world problems, and to challenge them to apply their knowledge and skills to solving them. It is in this translation from theory to AI to application that universities should strive to nurture today's STEM students. At Atria University, for example, we deliberately create studios and labs where learning is measured by a student's ability to solve real problems with whatever knowledge and tools they have — including AI.
A word of caution: the availability of AI does not diminish the importance of learning scientific and mathematical fundamentals. Building mental models of the real world, a core capability in STEM, requires retaining facts and internalising information. Outsourcing all recall to AI or a web search erodes the mental discipline required to hold and manipulate complex ideas internally.
Evaluating students' understanding of fundamentals remains as important as ever. Traditional methods of assessment therefore still matter. When it comes to foundational STEM subjects, application is important, but so are the fundamentals — creative solutions require students to hold and manipulate abstract concepts and models in their own minds. Exams and oral vivas remain necessary ways to measure learning; in the age of AI, however, they are no longer sufficient on their own.
In the age of AI, learning is changing, and so must evaluation. Educational institutions must assess a student's ability to ask questions, to question AI responses, and to interrogate claims presented as evidence. Universities must assess a student's ability to apply scientific concepts and technologies to real-world problems, while still ensuring a strong grasp of the fundamentals.
(The views expressed are personal)
This article is authored by Praphul Chandra, dean (research), Atria University.
