We all know that this summer in India, a national examination question paper was leaked. The exam was NEET. But this makes us all think about impartial examinations. The teacher who sets the exam and the protocols that follow. NEET is a national exam. But many exams are not national, such as college exams. College and university exams are held for a smaller number of students. This also needs improvement to make examinations impartial and to assess students using global metrics.
In this article, I present a computer algorithm to generate question papers that are not biased by the setters and teachers who teach the subject to students. This ensures the results are not dependent on a something that has bias. At the same time, the method is leak-proof. Let’s make global standards in examination systems. Let global exams use the same metrics.
This is an intelligent computer algorithm, not an LLM based AI application.
There are steps in this process that need to be automated. The following steps are needed.
A few days before, make sure exam center printers are working, and centers have enough printing paper.
Step 1. Create a question bank and add questions to it before each exam. This way, the question bank grows every 6 months.
Step 2. Annotate questions as 2-mark, 5-mark, and 10-mark questions. These numbers can change as per exam and university requirements. The questions must cover all credits a course includes; that is, all syllabus topics must be covered.
Step 3. Write a computer algorithm that generates a random number and generates, say, 10 questions worth 2 marks, 6 questions worth 5 marks, and 5 questions worth 10 marks. This makes a 100-mark score. But then we test whether these questions span the entire syllabus. For that, we use decision algorithms such as spanning sets. Spanning set algorithms ensure they cover the entire syllabus. The spanning set chooses one or more questions from each unit. The span should be maximized, meaning the question paper covers the entire syllabus. Next, we need to account for the marks assigned to each credit in the syllabus. For this, we must take a weighted average of the marks proportion and the question’s spanning capability (i.e., the topic it covers). This should ensure the question paper covers all essential topics according to each unit’s weightage. For this, annotate syllabus topics with the right tags. Use tags such as “linked list,” “stacks,” “queue,” “knapsack,” “TSP,” and “Graph” for the advanced algorithm course, along with topic weights. The maximum value of the spanning set can be that of a perfect question paper. This is a one-time task: to assign tags and weights to course material. As the course changes over time, you can update the weights and tags. You can also use other decision algorithms and optimisation techniques.
Step 4. Send the paper to human experts for review before printing, so the final paper can be attested by human experts. This is just before the exam. Add a digital signature of the human expert team that evaluates the examination paper just before the exam.
Step 5. If the paper fails requirements, another paper can be generated.
Step 6. Print the final question paper at the exam center at the time of the exam.
Step 7. Start the exam with human teachers as duty in-charge.
Step 8. Collect the answer sheets on time. Include extra printouts of question papers with answer sheets.
This way, the questions selected for student evaluation are independent of who is teaching the course. This can be used for national examinations such as NEET in India, which recently made global headlines.
Thank you for reading.
Subscribe for updates.
Artificial Intelligence and Future 