Rethinking Assessment: Why AI Means We Must Test Thinking, Not Answers

Thirty years ago, during a faculty interview at a premier research university, a world-renowned scholar asked me an unexpected question: “What is the toughest thing about teaching?” My answer then was “grading.” Today, as generative artificial intelligence transforms traditional assignments into an algorithmic arms race, that answer feels prophetic. However, the solution to this crisis does not lie in superior AI detection software; it requires a radical, unapologetic rethinking of how and what we assess.

If grading was a pedagogical challenge three decades ago, generative AI has turned it into an existential crisis. When algorithms can generate coherent essays, solve complex equations, and write clean code in seconds, traditional metrics of evaluation crumble. This forces higher education to confront uncomfortable structural questions: What must educators teach when information is ubiquitous and the professor’s role as the sole gatekeeper of knowledge is obsolete? If a student can deploy an algorithm to secure top marks, what do our grading systems actually measure?

“If a student can deploy an algorithm to secure top marks, what do our grading systems actually measure?”

In the Indian context, the policy intent to address this shift exists, but classroom execution lags behind. The University Grants Commission’s (UGC) 2023 advisory asked universities to frame comprehensive AI guidelines, yet implementation remains stagnant. Similarly, while the Central Board of Secondary Education (CBSE) has urged schools to pivot from rote learning to critical thinking, actual assessment methods remain archaic. We are attempting to protect legacy metrics with AI detectors that are fundamentally unreliable. To survive this technological disruption, higher education must replace the transactional, point-scoring model with tech-immune, classroom-centric assessments.

Revive the Viva Voce

First, we must revive the viva voce. Grading weight must shift significantly toward in-person oral defences. A student’s evaluation should depend not on a digital file uploaded minutes before a midnight deadline, but on their ability to stand before faculty or peers to defend, dissect, and unpack arguments under live questioning. Long standard in doctoral programs, this rigor must be brought into undergraduate classrooms.

Adopt Live-Patch Diagnostics

Second, institutions should adopt live-patch diagnostics that evaluate execution rather than the final product. Presenting students with intentionally broken code, flawed mathematical proofs, or weak historical arguments forces them to troubleshoot in real time without digital assistance. This approach mirrors actual workplace problem-solving.

Introduce AI Auditing

Third, we must introduce AI auditing. Rather than banning these tools fruitlessly, assignments should require students to critically evaluate AI-generated outputs—identifying factual hallucinations, verifying sources, and critiquing the cognitive limits of the machine. This cultivates genuine digital literacy rather than passive digital dependence.

Move to Edit-to-Mastery Grading

Finally, higher education must move away from the partial-credit systems that merely reward compliance, adopting instead an “edit-to-mastery” contract grading model. Under this framework, major assignments operate on a rigorous “Complete/Revise” loop. A submission is marked “Complete” only when it meets strict, pre-published professional standards. For instance, a case analysis in Strategic Management must successfully apply specific analytical frameworks and offer data-backed recommendations to pass. If it falls short, it is marked “Incomplete” and sent back for iterative rewriting. This mirrors global institutional efforts to combat grade inflation and restores institutional signaling value to top marks; an ‘A’ is earned through demonstrated mastery, not accumulated compliance points.

From Administrative Problem to Philosophical One

Thirty years ago, grading was difficult because educators had to discern conceptual understanding from human error. Today, it is difficult because we must parse authentic human thought from machine output. Our primary challenge is no longer administrative; it is philosophical. We must redefine the purpose of the physical classroom—moving away from automated assessment and toward cultivating deep, focused human intellect in an era designed to distract it.

“AI has stripped education to its absolute core. It can synthesize data, but it cannot mentor.”

AI has stripped education to its absolute core. It can synthesize data, but it cannot mentor. The true value of an educator in the age of AI lies in executing the irreducibly human work: to notice individual student trajectories, to challenge assumptions, and to hold learners to a standard higher than what an algorithm can replicate. By leveraging the National Education Policy (NEP) 2020 and pending UGC directives to redesign our classrooms, Indian universities can turn this technological shock into a historic opportunity.

About the Author

Dr. Madhukar G. Angur is Founder Chancellor of Alliance University, Bangalore, and former Distinguished Professor at the University of Michigan-Flint. His new book is Next Point: A 10-Minute Defense System for Your Brain at Work.