Why the Old Exam Model is Broken
For decades, our assessment system has been heavily skewed towards rote learning and the recall of facts. Exams often tested what students could memorise, not necessarily what they understood. This model was already showing its age before generative AI
became a household name. Now, with tools that can produce essays and solve complex problems in seconds, an exam focused on information recall is fundamentally broken. The conversation should not be about how to stop students from using these tools. Instead, it should be about acknowledging that access to information is no longer the bottleneck. The real challenge is teaching students what to do with that information, a skill the old model was never designed to measure.
The New Core Skill: Verification
In a world flooded with information, both real and AI-generated, the ability to verify is paramount. AI models are known to 'hallucinate'—that is, to invent facts, sources, and data with complete confidence. An AI-era exam, therefore, should not ask for the answer, but for the student’s process of proving that the answer is correct. Tasks should require students to cross-reference AI-generated output with reliable sources, identify potential biases, and justify the credibility of their information. This moves the assessment from testing a static piece of knowledge to evaluating a dynamic skill. It rewards the student who asks, "Is this true, and how do I know?" rather than the one who simply copies and pastes.
Attribution as the New Academic Honesty
The fear of 'cheating' with AI stems from an outdated definition of academic integrity. Using a tool is not inherently dishonest; failing to acknowledge its use is. Educational institutions, including many in India, are slowly adapting their policies to reflect this. The future of academic work involves transparency. Students should be taught to document and disclose how they used AI, whether for brainstorming, drafting, or editing. This is no different than citing a book or a website. An assessment can include a mandatory reflection statement where students detail which prompts they used and how AI assisted their thinking process. This not only maintains honesty but also provides educators with valuable insight into a student's learning journey.
Rewarding What AI Cannot Do: Independent Judgement
Ultimately, the most important human skill in the age of AI is independent judgement. An AI can generate text, but it cannot form a truly original argument, synthesise ideas with genuine insight, or apply ethical reasoning to a complex problem. This is where the focus of assessment must shift. Future exams should be project-based, centred on real-world scenarios that have no single right answer. Students could be asked to critique an AI-generated solution, refine a flawed AI proposal, or use AI-drafted materials as a starting point for their own unique analysis. Grading should focus less on the final product and more on the process: the logic, the creativity, and the critical value added by the student.
A Glimpse into the New Exam
Imagine an open-book, open-internet exam where students are encouraged to use AI tools. The task is not to write an essay on a historical event, but to act as a historical consultant for a film, using AI to gather initial data but then independently verifying it, crafting a narrative, and defending their choices against potential inaccuracies. Or perhaps a business student is asked to evaluate an AI-generated market entry strategy, identifying its strengths, weaknesses, and ethical blind spots. In these scenarios, the AI is a collaborator, not a crutch. The assessment focuses on the uniquely human skills of critical thinking, creativity, and accountability—precisely the abilities needed to thrive in a world alongside AI.
