Education
Testing Thinking, Not Answers: What Assessment Must Become in the AI Age

Many years ago a faculty interview panel asked me what the toughest thing about teaching was, and I said grading without a moment's hesitation. I had no idea then how much harder it would get. Generative artificial intelligence can now write a coherent essay, solve a complex equation or produce clean code in seconds, and our traditional measures of student learning are crumbling under the pressure.
The obvious response is better detection software, which I think is the wrong battle. We have to rethink what we assess and how we assess it, so that we judge the thinking behind an answer more than the answer itself.
What our grades are measuring now
When information is everywhere, the professor is no longer the only gatekeeper of knowledge, and that raises uncomfortable questions for higher education. What should we teach when any fact is a few keystrokes away? I keep coming back to a harder one.
If a student can deploy an algorithm to secure top marks, what do our grading systems actually measure?
In India the policy intent is already there. The University Grants Commission's 2023 advisory asked universities to frame comprehensive AI guidelines, and the Central Board of Secondary Education has urged schools to move away from rote learning towards critical thinking. Classroom practice has not caught up. Our assessment methods are archaic, and the AI detectors that many are pinning their hopes on are unreliable.
Four assessment designs that measure real thinking
What we need are assessments that happen in the classroom, where technology cannot do the work for the student. I would start with four.
1. Bring back the viva voce
A meaningful share of the grade should move to in-person oral defences. A file uploaded minutes before a midnight deadline tells me very little. I learn far more when a student stands before faculty or peers and has to defend, dissect and unpack an argument under live questioning. Doctoral programmes have long relied on this, and undergraduates deserve the same standard.
2. Use live-patch diagnostics
Assess execution as well as the end product. Give students deliberately broken code, a flawed mathematical proof or a weak historical argument, and ask them to fix it in real time with no digital help. That is how problems turn up at work, imperfect, incomplete and usually urgent.
3. Make AI auditing an assignment
Banning these tools gets us nowhere. Ask students to audit what the AI produces instead. Can they spot a factual hallucination and check the sources? Can they see where the machine's reasoning runs out? That builds real digital literacy and guards against passive dependence.
4. Grade for mastery
Partial-credit systems often reward compliance more than competence. An "edit-to-mastery" contract grading model works differently. Major assignments follow a Complete/Revise loop. Work is marked Complete only when it meets strict professional standards published in advance; otherwise it goes back for rewriting. A case analysis in Strategic Management, for instance, would pass only if it applied the required analytical frameworks and made recommendations backed by data. Under such a system an 'A' means the student has shown mastery, and top marks start to mean something again.
Why this is now a philosophical problem
Grading used to be hard because we had to separate conceptual understanding from ordinary human error. Now we have to separate authentic human thought from machine output, and I see that as a philosophical problem more than an administrative one. It forces us to ask what the physical classroom is for. My answer is less automated assessment and more cultivation of deep, focused thinking, in an age designed to distract our students.
What only a teacher can do
AI can synthesise data, but it cannot mentor. A teacher notices where each student is heading, challenges their assumptions and holds them to a standard higher than any machine can meet. Assessment built around thinking makes that work visible to students, employers and institutions.
What Indian universities can do now
Generative AI has shown how much of our assessment rewarded the production of answers over the quality of reasoning. With the National Education Policy 2020 and the UGC directions still to come, Indian universities already have a framework for changing this. I believe we have a real chance to turn this technological shock into a lasting improvement in how we teach and how we judge learning.


