What Can You Promise About a Result? The Question Every Institution Will Eventually Be Asked 

Blog Series · Academia Track · Part 5 

So far this track has worked through the parts of assessment that institutions usually trust without checking. Part 1 showed that a scorecard is not evidence on its own. Part 2 looked at whether two examiners would mark the same script the same way. Part 3 followed the paper trail behind a single result. Part 4 asked what breaks when one exam runs on ten thousand screens at once. 

Part 5 asks the question all of that leads to. A parent writes in, a student appeals, or a lawyer sends a letter. Each wants to know how this particular result was reached. What can the institution actually promise them? 

Most institutions would say “the result is fair.” That is a belief, not a promise. A promise is something you can show, on request, for one named candidate, months after the marks went out. 

The law already assumes you can explain a result 

In several major systems, the right to see how a result was reached is already settled. It does not wait for a scandal. 

In India, the Supreme Court held in CBSE v. Aditya Bandopadhyay (2011) that evaluated answer books are information a candidate can request under the Right to Information Act. The Court rejected the argument that examining bodies hold answer books in a fiduciary relationship that would exempt them. 

In the UK, the Information Commissioner’s Office is clear that a candidate’s own answers are exempt from subject access, but the marks and the examiner’s comments are not. If the request arrives before results are published, the institution gets extra time, but it still has to answer. 

In the US, FERPA gives parents, and students once they turn 18 or enter college, the right to inspect education records within 45 days of asking. 

The common thread is simple. The law expects the marks, the marker’s reasoning and the record behind them to exist, and to be producible for one person. An institution that can only produce a final grade is already behind what these rules assume. 

What happens when the method can’t be defended 

The clearest recent example of a result that could not be explained is England’s 2020 exam season. With exams cancelled, Ofqual used a standardisation model to adjust teachers’ predicted grades. Within days of results, the model’s grades were abandoned and replaced with the teachers’ centre assessed grades. 

When Ofqual’s chair gave evidence to the House of Commons Education Committee in September 2020, the defence of the statistics was not the problem. He pointed out that most grades sat at or within one grade of the teacher’s assessment. The problem was that individual students had no meaningful way to challenge their own result. He acknowledged that the appeals route was narrow and felt unfair to individuals, and called the belief that the approach would be publicly acceptable the fundamental mistake. 

That is the lesson for every institution. A method can be defensible on average and still indefensible for one person. Parents and courts don’t ask about averages. They ask about this student. 

Four promises an institution can actually keep 

An institution can’t promise that every result is right. It can promise that every result can be shown, checked and corrected. In practice, that comes down to four commitments. 

1. We can show you how this mark was reached- Not the grade boundary for the cohort, but the marks against each criterion, who awarded them, and the comments they left. This is exactly what the ICO and RTI rulings already expect. 
2. We can show you it was checked- Moderation, second marking and consistency checks only count if they are recorded against the script. “We moderate a sample” is a policy. “This script was in the sample, and here is what changed” is evidence. 
3. We can tell you what affected the conditions- If a candidate’s session dropped, their accommodation was applied late, or a proctoring flag was raised and cleared, that belongs on their record. Part 4 showed how often scale produces these events. The promise is that they are not invisible afterwards. 
4. We can tell you how to challenge it, and it will be looked at by a person- A review route that only re-adds the marks is not a review. The candidate should know what grounds count, who decides, and how long it takes. 

None of these needs new technology to state. All of them need records that exist before anyone asks. 

AI raises the bar, it doesn’t lower it 

It is tempting to think AI-assisted marking makes these promises harder to keep. Regulation is moving the other way. It is making them a stated requirement. 

Under the EU AI Act, systems used to evaluate learning outcomes, or to detect prohibited behaviour during tests, are classed as high-risk (Annex III, point 3). Article 86 then gives a person affected by a decision based on such a system the right to a clear and meaningful explanation of the AI’s role and the main elements of the decision. 

Read together, that is close to a legal version of the first promise above. If an AI model flagged a script, suggested a mark or ranked an answer, the institution has to be able to say so, and say what a person did with it. 

This is where the principle from Who Grades the Grader? becomes practical. AI can surface the inconsistent marker or the unusual pattern. A qualified, named person decides what it means, and that decision is written down. An AI score with no human sign-off attached is not something an institution can stand behind in front of a parent, let alone a court. 

Why this connects back to the rest of the series 

Every earlier part of this track was, underneath, about this moment. The scorecard that isn’t evidence, the marker no one checked, the paper trail that stops at the final grade, the sitting that broke at scale: each one turns into the same problem when someone asks how their result was reached. 

The institutions that answer well are not the ones with the best policies on paper. They are the ones whose records were built to be read by an outsider, for one candidate, long after the fact. 

See how Nirvana Intellect supports academic institutions → · For examination bodies → 

Next in the series: Part 6 looks at what happens to that evidence after the result is settled, and whether a transcript can ever say what a learner is actually able to do.