AI test analysis

Score alone is the least informative number on the page.

• Correct, incorrect and unattempted counts
• Accuracy overall and per subject
• Time spent per question and per section
• Chapter-wise and difficulty-wise performance
• Repeated mistake types across recent tests

A dropped mark can be a knowledge gap, a timing failure or a careless error. The fixes are completely different — revision, pacing drills and checking discipline respectively. Analysis exists to tell them apart.

Every analysis should end in at most three actions. Re-test those chapters within two weeks and compare — that comparison, not the raw score, is your progress signal.

If you ask AI for the answer, you get the answer and lose the skill. Escalate your prompts in this order and your accuracy in the real exam improves instead of your dependence.

• Weak: "Give me the answer."
• Better: "Give me one hint, not the solution."
• Best: "Here is my attempt and my reasoning — tell me exactly where it breaks."
• Target loop: Think → Attempt → Get feedback → Correct → Practise → Master.
• Anti-pattern: Ask AI → copy answer → move on.

Every benefit below has a matching failure mode. Name them out loud with your student so they are managed rather than discovered in February.

• Hallucination: convincing but wrong facts, formulae or values — verify against NCERT and official syllabi.
• Overdependence: problem-solving stamina drops when AI does the first step every time.
• Academic integrity: work submitted without understanding teaches nothing and shows up in tests.
• Data privacy: educational AI processes sensitive student data; UNESCO's guidance stresses human-centred, age-appropriate, privacy-respecting use.
• Bias: models reproduce the biases in their training data and design.
• Reduced human contact: teachers, mentors, peers and family still drive most of a student's progress.