Find the Gap: How to Check Diagnostic Quality
The 'Find the Gap' tool helps teachers turn student answers into a list of likely misunderstandings and suggest targeted practice. However, automatic analysis must be checked: the model may misinterpret the formulation, miss an important gap or suggest exercises off-topic. In this guide we will examine how to evaluate the result quality by three criteria - accuracy, completeness and suitability - on a specific example with percentages. We do not promise perfect conclusions: the tool gives a diagnostic draft, and the decision remains with the teacher.
Open tool: Knowledge gap finder ↗What the Tool Does and What Is Input
"Find the Gap" is a text tool: you paste the source material (student answers, a work fragment, an oral answer transcript) and state the goal - which skill or topic you want to check. On output you get one text result in a separate panel: presumed misunderstandings and practice recommendations. The result can be copied or downloaded as a plain TXT file.
Important to understand the limits: the tool does not check spelling like a proofreader, does not compare document versions, does not highlight changes and does not publish materials. It does not build charts or generate images. Everything it does is analyze text and offer an interpretation. Therefore the output quality depends directly on how accurately you describe the goal and how informative the source text is.
Practical tip: before running the tool remove unnecessary details from the source material that are not related to the topic being checked. If you are checking percentage understanding, do not paste the entire test text - only those tasks and answers where the student's logic is visible. The cleaner the input, the easier it is to evaluate whether the tool hit the real gap.
Source material: "20% discount, then another 10%. I added the percentages: 20 + 10 = 30%. So the final price is minus 30%.". Goal: check understanding of sequential percentage changes.
Criterion One: Diagnostic Accuracy
Accuracy means the tool correctly named the error cause, not just rephrased the wrong answer. In the percentage example a weak result would say: "student miscalculated." A strong result would point to a specific misunderstanding: "student adds percentages, instead of applying them sequentially to the decreasing base."
To check accuracy ask yourself: if I were explaining the student's error myself, would my formulation match the tool's formulation? If yes - the diagnosis is useful. If the tool names the reason vaguely or incorrectly, the result needs to be reformulated manually.
Remember that the model can make mistakes. It does not know the context of your course and has not seen the student's previous works. Therefore accuracy is always evaluated by you, not taken on faith. It is useful to save successful diagnostic formulations in a separate file - gradually you will build a bank of typical misunderstandings.
Weak: "Student miscalculates percentages." Strong: "Student applies additive model: adds percentage rates, not considering that the second discount is calculated from the new price."
Criterion Two: Completeness of Coverage
Completeness is how many real gaps the tool noticed. In one student answer there may be several problems: misunderstanding a basic operation, computational error, misreading the condition. A good result separates them, rather than mixing them into one general phrase.
Check completeness by writing down all errors and gaps you see in the source text yourself, then compare with what the tool proposed. If it found fewer - add the missed ones. If it found more - check if it invented a problem not present in the text.
Do not strive for maximum list length. The goal is not to list everything possible, but to highlight what really prevents the student from moving forward. Extra items blur attention and make practice choice difficult.
In the percentage example there are two real gaps: additive model and misunderstanding the base for the second discount. If the tool output only the first, the second must be added manually.
Criterion Three: Suitability of Proposed Practice
Suitability is the most practical criterion. Even accurate diagnosis is useless if suggested exercises are impossible to perform or do not train the needed skill. Check that the practice matches the student's level and your program.
A good recommendation is specific: it describes the task type, not general words. Instead of "repeat percentages" better "solve three tasks where the second discount is applied to the already reduced price, and compare the result with erroneous addition." Such a task can be given to the student immediately.
Also evaluate volume: if the tool suggests ten exercises for one gap, that is overload. Choose two or three that give maximum effect and leave the rest for later. Remember that recommendations are a draft, not a ready homework assignment.
Unsuitable: "Repeat the percentage topic." Suitable: "Give a task: item 2000 rubles, 20% discount, then another 10%. Ask to calculate in two ways and explain the difference."
How to Check the Result Step by Step
Step one: read the output fully and mark statements that seem questionable to you. Step two: return to the source text and check each statement - is there confirmation in the student's answers? Step three: compare the gap list with your own list.
Step four: evaluate the practice formulations - can they be used immediately? Step five: make manual corrections. The tool does not save history and does not keep a student database, so the final document you form yourself.
It is useful to keep a simple log: date, topic, what the tool proposed, what you corrected. After several weeks you will see in which topics the diagnosis is more reliable, and where manual check is needed. This is not a statistical study, but a working method to tune the process to yourself.
Checklist: 1) error cause named specifically? 2) all gaps from text accounted for? 3) practice executable? 4) no extra items? 5) formulations clear for student?
Limitations and Typical Errors
Main limitation: the tool works only with the text you gave it. It does not know how the student solved the problem, does not see drafts and does not hear explanations. If the source text is short, the output will be superficial. If the text is overloaded with details, the model may go the wrong way.
Second mistake - taking the output as the final diagnosis. The model can make mistakes, and service availability may change. Always leave yourself the right to reformulate or reject the proposal.
Third mistake - using the tool for topics where personal responsibility is needed: financial, medical or legal decisions. Here the material must remain educational, without personal recommendations. The tool is suitable for educational diagnosis, not for consulting.
If the source text is one phrase "I didn't understand percentages", the tool cannot accurately determine the gap. Specific student answers are needed.
Realistic Usage Scenario
Imagine a math teacher who received from a student written answers to five percentage tasks. He inserts the answers into the source material field, and in the goal field writes: "Check understanding of sequential percentage changes." He presses Generate and gets a text report.
The teacher reads the report, compares with the student's answers, adds the missed gap about the base for the second discount and selects two exercises from the proposed ones. Then he copies the final list into his lesson plan. The whole process takes several minutes, but requires attentive reading.
This is a fictional example, it shows the working logic, not a real result of a specific service. Your situation will differ depending on the topic, student level and quality of source answers.
Goal: "Check understanding of sequential percentage changes.". Source material: student answers to five tasks. Result: diagnostic draft that the teacher edits manually.
"Find the Gap" is a useful assistant for rough diagnostics, but the result quality is determined by three checks: accuracy of the error cause, completeness of the gap list and suitability of the suggested practice. Compare the output with your own analysis, make corrections and do not accept automatic formulations as the final diagnosis. The tool does not save history, does not publish materials and does not replace the teacher. Use it as a routine work accelerator, and the decision about what the student really needs remains with you.
Open tool: Knowledge gap finder ↗Interface translations are machine-assisted and may need refinement.