Decode the Mark: Why Every Score Needs an Explanation
A student receives 6/10. But what does that score really explain? Explore how evidence, marking criteria, and meaningful feedback can make assessment more transparent and evaluation more meaningful.
The Missing Story Behind the Score
Imagine a student receiving an answer sheet after a test. At the top, in red ink, is one number.
6
She stares at it. Six. Not five. Not seven. Six. She remembers studying for the test. She remembers writing carefully. She remembers being certain about some responses and less certain about others.
But one question keeps returning.
“Why six?”
The number does not tell her.
- What has she understood?
- What has she missed?
- Where have the other four marks gone?
She turns the paper over. Nothing. Only the number. And that is when she realises something:
The mark is visible. The story behind it isn't.
The student and classroom events described in this article are illustrative. They do not represent an actual documented student case.
Why Did I Get Six?
Imagine that the next morning, she approaches her teacher.
“Ma'am, could you tell me why I got six?”
“Of course. Let's look at it together.”
They begin with the first response.
“This part demonstrates the concept. That's why you received these marks.”
Then they move to the next.
“This idea is relevant, but one important point is missing.”
And finally:
“Your vocabulary is strong here, but the grammar affects the clarity of your answer.”
The teacher breaks down the score:
- Content — 2/3
- Organisation — 1/2
- Vocabulary — 2/2
- Grammar — 1/3
The total is still 6/10.
Nothing has changed. Yet suddenly, the six feels different. She can see where the marks came from, where they were lost, and what she could improve.
“So now I understand why I got six,” she says.
The teacher smiles. The result hasn't changed. The understanding has.
This illustrative scenario captures an important distinction: awarding a mark and explaining a mark are not the same thing.
A score communicates an outcome. Meaningful feedback helps a student understand the reasoning behind it and identify what to work on next.
A mark can tell a student what was awarded. By itself, however, it may not reveal:
- What did I write?
- What evidence of understanding was recognised?
- Which marking criteria were applied?
- How did the evaluator arrive at this decision?
- How did that decision become 6/10?
Behind every evaluated response lies a sequence of decisions. The student's answer provides the evidence. The marking scheme or rubric establishes the criteria. The evaluator interprets the response and determines how it meets those criteria.
Making these connections visible can help students understand their results and help educators explain their decisions. But transparency involves more than displaying a rubric alongside a score. It also means recognising when an answer deserves another look.
Consider another illustrative situation: a student submits an answer that differs from the model response but presents a logical, defensible method.
The important question is not simply whether the response matches the model answer. It is whether the response demonstrates valid understanding. Some answers are straightforward. Others involve ambiguous wording, unconventional but valid methods, unclear diagrams, or handwriting that requires closer examination. These cases may benefit from additional human review.
A review does not necessarily mean the original evaluation was wrong. Nor must it result in a different mark. Its purpose is to examine the evidence and determine whether the original decision remains justified.
Transparency does not guarantee agreement. It makes the basis of a decision easier to understand and question.
The conversation can move from:
“Why did I get only six?”
to:
“Show me where my answer did not meet the criterion.”
That is a more specific and constructive starting point for discussion.
From Answers to Assessment Decisions
The value of a review becomes clearer when the result does not change. A reported rechecking exercise at Devi Ahilya Vishwavidyalaya (DAVV) in Indore offers a real-world example. According to The Times of India, 15 B.Ed. students observed two subject experts reviewing 45 answer scripts during a session lasting approximately two and a half hours. The evaluators explained their scoring decisions and addressed questions about partial credit and marking criteria. The report stated that no marks changed and that the students accepted the results. (Source: The Times of India, 27 November 2025.)
The significance of this reported exercise is not that it proves every evaluation was correct. Rather, it illustrates how explaining the reasoning behind a decision can help students understand an outcome, even when the score remains unchanged. The example should be understood as a reported university exercise, not as proof that transparency alone guarantees student satisfaction or eliminates evaluation errors.
For digital assessment, the corresponding challenge is to make this kind of explanation possible without requiring every evaluation to be repeated manually. That brings us to a more fundamental question: how can an automated system preserve the evidence and reasoning behind a mark?
Beyond Scoring : The Challenge of Traceable Evaluation
For a teacher, connecting a student's answer to a marking criterion may seem straightforward. The teacher reads the response, interprets its meaning, compares it against the expected criteria, and awards marks. In an automated evaluation system, these steps involve several distinct technical challenges.
The challenge is not simply to generate a score. It is to preserve the relationships between the student's response, the criteria applied, the decision made, and any subsequent review.
Consider what must happen behind the student's 6/10.
1. Evidence Integrity: Is the original Answer Preserved
Before an answer can be evaluated, the system must establish which response belongs to which student, question, and answer-sheet page.
Scanned pages may be incomplete, blurred, rotated, or incorrectly ordered. Handwriting recognition may misread words, mathematical symbols, or formulas. Even when text is extracted correctly, formatting and context may be lost.
If the evidence is captured or interpreted incorrectly, later stages may evaluate something different from what the student actually wrote.
2. Rubric Alignment : Are the Right Criteria Being Applied?
A marking rubric may contain multiple criteria, partial-credit rules, acceptable alternative methods, and subject-specific conventions.
The system must connect relevant parts of the response to the appropriate criteria and use the correct version of the marking scheme. Without explicit criterion mapping, an automated system may produce a plausible score without making clear which parts of the answer satisfied which requirements.
3. Decision traceability: How was the score reached?
A final mark is an outcome, not an explanation.
A traceable evaluation should connect the response to the applicable criterion, the evidence identified, the decision made, and the marks awarded. This becomes more challenging when several models, rules, or scoring components contribute to the final result. Without structured links between these stages, reconstructing the reasoning may require repeating the evaluation or relying on incomplete logs.
4 Review Traceability: What Changed and Why?
Suppose the initial score is 6/10, but a subsequent review identifies additional evidence that justifies another mark.
The record should distinguish the original decision from the revised one, connect the change to the review, and preserve the reason for the adjustment. Simply replacing six with seven would retain the latest score but lose the history of how the decision evolved.
5. Beyond Automation : Where Human Judgement matters.
Some responses are straightforward. Others involve ambiguous wording, unconventional but valid methods, or competing interpretations.
A robust evaluation design should recognise that not every response is equally suitable for automated judgment. It should define when additional review may be appropriate and preserve the role of qualified educators in decisions that require interpretation.
These challenges are interconnected. Accurate text recognition does not guarantee correct rubric application. Correct criterion mapping does not guarantee accurate aggregation. And a correct final score does not, by itself, make the evaluation explainable.
The technical requirement is to preserve the connections throughout the evaluation lifecycle:
Student Response → Captured Evidence → Applicable Criterion → Evaluation Decision → Awarded Mark → Review → History
This requires more than a model that generates a score. It calls for a structured evaluation process, explicit relationships between records, appropriate validation, controlled rubric versions, and a reliable record of decisions and changes.
The question is no longer simply whether a system can award six out of ten. It is whether the system can establish which evidence supported those six marks, which criteria governed the decision, and what happened if the evaluation was reviewed later.
Beyond Individual Scores: Learning at scale.
These connections matter beyond individual answer sheets.
Imagine a school examining patterns across thousands of evaluations. Are students repeatedly missing the same concept? Is a particular rubric criterion producing frequent disagreements? Are certain types of responses regularly being flagged for review?
Such patterns could reveal opportunities to improve teaching, clarify assessment criteria, or investigate inconsistencies in evaluation.
Repeated misunderstandings might indicate a learning gap. Frequent ambiguity around a rubric criterion might suggest that the marking guidance needs clarification. Repeated review requests might point to areas that warrant closer examination.
These are possibilities to investigate, not conclusions to assume from scores alone. With appropriate evidence and analysis, evaluation can contribute to a continuous feedback loop:
Evaluate → Explain → Review → Learn → Improve
Assessment then becomes more than a process for assigning marks. It can also help educators understand what students have learned, where difficulties remain, and what may need attention next.
ONETUZA: Beyond Scores Towards Explainable Evaluation.
This is the thinking behind ONETUZA.
We believe educational assessment should be more than the production of a score. It should support a clear relationship between what a student has written, what the assessment requires, how the response is evaluated, and how the final result is reached.
Our evaluation philosophy is guided by five design principles.
1. Evidence before conclusions
An evaluation should be grounded in the student's response. The score should not become detached from the evidence on which it is based.
2. Criteria before scoring decisions
The applicable marking criteria should provide the basis for evaluating a response. A meaningful explanation should help connect the decision to those criteria.
3. Structured evaluation rather than a single unexplained output
Assessment involves several distinct tasks, from capturing the response to interpreting it and calculating the result. Treating these as identifiable stages creates a clearer basis for validation and review.
4. Human judgment where interpretation matters
Automation should support professional judgment, not assume that every response can be assessed reliably without human involvement. Ambiguous or unusual answers may require additional scrutiny.
5. Traceability when decisions change
An evaluation should be considered a process, not merely a final number. Where reviews or modifications occur, preserving the history and reasons for those changes is important to accountability.
These principles describe how we believe an explainable evaluation system should be designed. They should not be mistaken for confirmation that every capability is already implemented or verified in ONETUZA's production environment.
Specific capabilities—such as response-to-criterion traceability, versioned rubric records, review escalation, confidence-based routing, and complete audit histories—must be assessed against the product's actual implementation and tested behaviour before being presented as live features.
The distinction matters. A design principle establishes what a system should achieve; a verified production capability establishes what the deployed system demonstrably does.
For ONETUZA, the ambition is to help make assessment more structured, transparent, and useful to educators, while ensuring that product claims remain grounded in demonstrable capabilities.
A Mark Should Tell More Than a Number
Return, one last time, to the student and her answer sheet.
There it is.
6/10
The number has not changed. But she now knows what questions to ask.
What did I write?
What was expected?
How was my response evaluated?
Why did I receive this mark?
Was anything changed during review?
She understands that a meaningful assessment process should make these questions easier to answer. A mark tells us what was awarded. Evidence helps us understand why. A traceable history can show how a decision was reached and whether it changed.
The mark may stay the same. The student's understanding of the mark can change. Because a score is not the whole story.
A mark should not only be awarded. It should be explainable.
That is the principle at the heart of ONETUZA's evaluation philosophy: connect the evidence, make the reasoning clearer, preserve the importance of human judgment, and make assessment more meaningful for the people who depend on it.