A student taps an option in the Learning Game, the screen turns red, and they are certain the screen is the one that is mistaken. Usually it is not: being wrong and finding out at once is what practice is for. But sometimes the student is right, because the question itself was wrong.
Quiz questions in PDP Shikshya are generated rather than typed out by hand, and a generated question can come back with its answer key pointing at the wrong option: the working in the explanation is right, and the option marked correct is not the one that working arrives at. We would rather say so plainly than claim an accuracy we cannot promise. What matters is what the product does about it.
Catching what can be caught, and admitting the rest
Before a generated question is ever shown to a child, an automatic consistency check reads it against its own explanation and discards the ones where the two clearly disagree. That check is deliberately cautious — it is tuned to avoid throwing away good questions — so it stays silent whenever it cannot read the reasoning confidently, and those questions are served.
Nor is a bad question a one-round annoyance. Questions are kept in a shared bank and re-used for some time afterwards, so one mis-keyed question can go on telling class after class that they were wrong when they were right. The student who notices is the most reliable detector we have.
What a student actually does
The route is deliberately small. After answering — and only on a question that was marked wrong — a quiet line appears under the explanation: "This answer is wrong". Tapping it opens a short box with the question repeated and one optional field. A student who wants to explain can write a sentence; a student who just wants to flag it can send nothing at all. It is framed as reporting a question, not as complaining about a mark, and no adult has to be asked first.
Nothing is sent mid-round. The flag is held until the round ends, so play is not interrupted, and everything reported is then sent for re-checking while the student is looking at their results. Each report comes back with its own outcome on that screen.
The re-check is a second opinion, not a rubber stamp
A challenge is neither taken on trust nor waved through. The question is examined again, and the order of that examination is the whole design: the checker works the question out for itself first, and is only then shown the answer key we hold, explicitly labelled as a claim that may be wrong. A checker shown the key first agrees with it far too readily.
Three further things hold it steady. The question and its options are rebuilt from our own stored copy rather than from whatever the student's device says they were, since the options are shuffled each time a question is served. Extracts from the class's own textbook, for that subject and grade, are supplied alongside and outrank general knowledge where they speak to the question — a question can be wrong in general and right for this syllabus. And the student's own sentence is passed on only as a hint about which part felt unfair; it is explicitly not allowed to move the verdict, because a child writing in order to have their score changed should not be able to simply ask for it. The finding comes back in the language the question was set in, so a Nepali-medium question is settled in Nepali.
Three possible outcomes, and only one of them moves a mark
- The student was right and the key was wrong. The score is corrected there and then — the round's total and points are recalculated on the results screen — and the question stops counting as a wrong answer in what teachers later see.
- The original answer stands. The student is told so, with a short plain-language explanation of why that option is correct, which is a better outcome than an unexplained red mark.
- It cannot be settled confidently: the question is ambiguous, badly worded, or the re-check is simply unsure. The score is left unchanged, the student is told exactly that, and the question is still put in front of the school's staff.
A mark is only moved when the re-check is confident and its own independently derived answer agrees with the option the student chose; a re-check that contradicts itself does not get to change a child's record. And a challenge can only ever add — only a question marked wrong can be challenged, and upholding one never takes a mark away.
What happens to the question matters more than one score
Correcting one child's mark repairs one child's afternoon. If the question stays in circulation, every other class that meets it is told the same untruth. So an upheld challenge withdraws the question from service for that school. It is not deleted — staff still need to see what it said — it simply stops being offered to anyone.
We withdraw the question rather than quietly rewrite its answer key, and that is a deliberate trade with a real cost. Rewriting a key on one automated opinion no teacher has reviewed would fail silently if it were wrong. Withdrawal is permanent and thorough: if the same question is generated again later, it still does not come back. Usually the question itself was fine and only the key was wrong, so a sound question is lost along with a bad one. The bank refills with others, and that is the better direction to fail in.
What the school can see
Every challenge and its outcome is kept, and the school's own staff can see them — a teacher for the students in their department, a school administrator across the school. The record is grouped by question rather than by child, and ordered by how many students reported each one. Five students reporting one question is a single question to pull, not five complaints; a list of pupils who objected to their marks would read as a discipline problem when it is really a maintenance queue.
The limits, stated honestly
Pressing the button again on a question already challenged returns the same finding rather than a fresh opinion — the answer is not re-rolled until it is favourable. Each student may raise a small number of challenges an hour and a slightly larger number a day. And if a round was played offline and the score is still waiting to save, the challenge cannot be filed yet: the screen says so and asks the student to report it next time they play online, rather than pretending it was sent.
Why a school should want this
Software can promise it is never wrong, or it can give the person most likely to notice a way to be heard. Only one of those survives contact with a classroom. A student who reports a bad question and gets a straight answer — you were right, or you were not, and here is why — learns something about evidence that a silently corrected database never would. And a teacher told "the app marked me wrong" has somewhere to send the child other than a shrug.