Objective vs subjective questions: a key or a marker
What is a subjective question? Compare objective and subjective questions with definitions, examples and guidance on answer keys, rubrics and marking.
An objective question has a predetermined correct answer or a defined set of acceptable answers. A subjective question requires judgement against criteria, usually expressed in a marking rubric. The difference is how the response is scored, not how difficult the question is.
For example, “What is 12 × 8?” is an objective question: the answer is 96. “Which solution best addresses this case, and why?” is a subjective question when the response is assessed for the quality of its reasoning against a rubric.
On NUADU, multiple-choice, matching, ordering, code and other objective formats are scored deterministically, by rules set when the assessment is configured, while essays and spoken responses are marked against a rubric and signed by a named marker. Exam officers and IT or compliance leads usually want two things confirmed: where that line falls, and that no model sits between a candidate's answer and an objective score.
Objective vs subjective questions: what is the difference?
An objective question has a predetermined correct answer, or a defined set of acceptable answers, so every response can be scored the same way by the same rule. A subjective question asks for a response that has to be judged against criteria, so two qualified markers could reasonably reach different scores.
Typical examples on each side:
- Objective: multiple choice, matching, ordering, a numeric answer within a tolerance, a short typed answer checked against a key, code run against test cases.
- Subjective: an essay, a short constructed response marked against descriptors, a spoken response, a case analysis.
The difference is not difficulty. A multiple-choice item can test hard reasoning, and an essay prompt can be easy. The difference is what does the scoring: a rule decided in advance, or a person applying a rubric. The same split scales up to objective vs subjective assessment in education — a paper can be mostly one or the other, and many mix both.
Choose by what the evidence has to show. Objective questions suit knowledge with a checkable answer, and cohorts where every response must be scored the same way. Subjective questions suit what only a performance can show — an argument, a piece of writing, a spoken response — and they bring the cost of judgement with them: a rubric, a marker, and a second look wherever the policy requires one.
It also decides a procurement question. "AI-assisted marking" travels badly through a review; by the time the phrase reaches an IT or compliance lead it has sometimes become "the AI marks the exams", and the first question back is whether that includes the multiple-choice paper. On NUADU it does not.
What you configure for objective questions
Objective formats are scored by rules, not by a judgement made response by response. Partial credit scoring is one of those rules: it awards part of the marks for a partly correct response instead of scoring it all or nothing. On NUADU, scoring for MCQ, matching, ordering, code and other objective formats is deterministic, and what gets configured is:
- Partial credit — how much a partly correct response earns.
- Synonym acceptance — which alternative answers count as correct.
- Tolerance ranges — how close a numeric answer must be to count.
- Code execution against test cases — a submission runs against the test cases set for the item, as configured.
Each is a setting, not a call made per response. Once configured, the same rule applies to every candidate in the cohort, so consistency is a property of the rule rather than of a marker's stamina on the day. The rule produces the number; a named educator still signs the result, and that sign-off is logged like every other.
Deterministic where the format allows. A named marker where it does not.
Is automatic grading AI?
Not on NUADU's objective items. The phrase covers two different things — a key applied by a rule, and a model's judgement of free text — and many explanations blur them. NUADU keeps them apart. Objective sections are scored automatically where the rubric allows, by rules, with no model between the response and the score. Constructed and spoken responses, with AI assistance on, receive draft scores and rubric-anchored explanations that a marker reviews, overrides where needed and signs; AI never makes the final grade decision.
Why not let a model score the objective items too, for consistency? Because it would add nothing and cost something. A key is already consistent. A model has an opinion; a key does not. Putting a model behind a key would turn a reproducible rule into an output that has to be explained — the opposite direction from defensible.
That is also why the objective side is unaffected by the AI switch. Everything on this platform works with every AI capability switched off, and for objective items that sentence describes normal operation, not a fallback: switch AI on and, in scoring, only constructed and spoken responses change.
Where the line falls on a mixed paper
A mixed paper carries two kinds of score, and the record keeps them apart: objective sections scored by rules, and a writing or speaking task with a marker's name on it. Both are logged. Only one involves a model, and that one always involves a person too.
Rule-based scoring also produces evidence about the items themselves. For objective items, distractor analysis is available for item review — which wrong options candidates chose, and whether an option the strongest candidates pick points at a flaw in the item rather than in the candidate. Reading those statistics well is what item analysis in assessment is for.
The same items feed two other decisions. For high-stakes programmes, NUADU draws parallel test forms from a bank of items with calibration statistics, and equating keeps the forms comparable in difficulty. And when an existing bank moves, items can be imported with metadata, framework mapping and rubric structure, the starting point for any item bank migration; every item, objective or subjective, still passes a named approver before it is used.
Marking, rubrics and moderation, on the Grade page, covers both sides of the line, and exam authoring and the item bank in Create is where items are written, reviewed and versioned.