Decision Support and Responsible Use
Version 0.2-draft · Effective 2026-11-07
DRAFT: for review by the Data Protection Officer and school administration.
The system includes "decision support" tools. They look at numbers, such as how students did on past questions, and give suggestions, predictions and warnings. This page explains what they do and how to use them responsibly.
The main rules
- The teacher always makes the final decision. The system only flags and recommends.
- Predictions and recommendations are estimates. They can be wrong.
- There are no automated grade or disciplinary decisions. The system never changes a grade, never fails or passes a student by itself, and never decides that a student cheated. This includes the flag for similar answer patterns ("possible collusion"). That flag is a reason to look at the papers. It is not proof, and it must not be used alone to accuse or punish anyone.
- Scanner results must be reviewed before you confirm them. The camera and the handwriting reader can make mistakes. Look at the photo and the answers shown, correct what is wrong, and only then confirm. Nothing is recorded until you confirm.
- Every recommendation shows its evidence. Each one has a rule code (for example DSS-01), the numbers it is based on, and Accept, Dismiss and Note buttons. Your choice is saved with your name and the date. Opening the page again updates the numbers but never undoes your decision.
What the tools do
All the rules below are available now. The limits they use (for example the mastery threshold of 60%) are System Settings that the school can change.
| Code | What the system does | What it looks at | What you see | Status |
|---|---|---|---|---|
| PER-01 to PER-05 | Splits a test's items by grading period (Prelim, Midterm, Finals) | The exam type's period rule (or the teacher's split for that test), the topics of each period and their hours | A TOS grouped by period with subtotals; only questions of the right period can fill a cell | Built |
| WIF-01 to WIF-04 | Predicts how a test may turn out before it is made (What-if) | The chosen topics and levels, and how students did on those questions in the past | Expected average score, expected passing rate, and a label: Easy, Moderate or Difficult | Built |
| GEN-01 | Ranks the questions that fit each cell of the Table of Specifications | How well a question separates strong and weak students, how close it is to the wanted difficulty, how long since it was used, how often it was used | A ranking and score per question. The teacher can replace any question | Built |
| IA-01 to IA-07 | Measures how each question performed after a test | How many students answered correctly, and whether strong students did better than weak ones | A suggestion for each question (Retain, Revise or Reject), problem distractors, a difficulty relabel suggestion, and test reliability. The teacher decides what changes in the question bank | Built |
| SCN-01 to SCN-07 | Reads the photo of an answer sheet | How dark each circle is, the QR code, the handwriting in the boxes | What the student shaded or wrote, how sure the scanner is, and photo problems. The teacher confirms every paper | Built |
| CHK-01 to CHK-03 | Flags answers that may be read wrongly | Scanner confidence, more than one shaded circle, faint or erased marks, photo quality | Highlighted items the teacher marks right or wrong, or a request to scan again | Built |
| CHK-04 | Warns that an answer key may be wrong | Whether most of the top scorers chose the same answer that is not the key | A warning with the evidence. If the teacher corrects the key, every recorded paper is re-scored and a before/after list is shown | Built |
| CHK-05 | Points out a possible similar answer pattern | How many identical WRONG answers two papers share | A side-by-side view for the teacher, who may dismiss it or write a note. It is never proof of cheating | Built |
| CHK-06 | Finds missing, duplicate and replaced papers | The class list compared with the recorded papers | Students with no paper are listed; a second paper for the same student is blocked unless the teacher replaces the first | Built |
| DSS-01 | Points out students who may need support (private: teacher and Program Head only) | Latest score, trend across tests, KNOW vs HOTS, missing papers, failed tests | A risk level (Low, Moderate, High) with the reasons and suggested actions. Never shown to students or posted | Built |
| DSS-02 | Finds weak topics (topic × KNOW/HOTS, shown with the topic's grading period) | How much of the possible points the class earned in each area | Re-teach, add practice questions, give it more weight in the next TOS | Built |
| DSS-03, DSS-04 | Compares KNOW with HOTS results, and the What-if prediction with what really happened | KNOW % and HOTS %, predicted and actual mean and passing rate | A note when HOTS lags, the prediction error, and a KNOW/HOTS split to try next time | Built |
| DSS-05, DSS-06 | Item and question bank health | Item Analysis decisions, usage, questions available per topic and level | Questions to revise or retire, and topics that need more questions | Built |
| DSS-07 | Suggests teaching alternatives | The kind of problem found | 2–3 strategies to choose from (the teacher picks one or none) | Built |
The complete list of rules, with formulas, thresholds and the settings that control them, is in the project document docs/DSS-RULES.md. Ask the Administrator if you want to read it.
Why the results can be wrong
- Little data. A prediction for a new question or a small class rests on default values, not real history. The What-if page tells you how many items had no history.
- Averages hide people. A class average says little about one student.
- Different situations. A question that was easy for last year's class may not be easy for this year's class.
- Reading errors. The scanner can misread a faint bubble, a smudge, or messy handwriting.
- Rules are simple. Thresholds such as "revise this question" are rules of thumb, not facts about the question or the student.
- A risk level is a weighted sum of a few numbers. It does not know about a student's health, family, or a bad day. It can rank a student too high or too low.
How predictions get better
The What-if prediction made when a TOS is approved is saved. After the test is checked, the system shows how far the real mean and passing rate were from the prediction, and which KNOW/HOTS split would have aimed better. When you run Item Analysis, each question's real difficulty is stored in the question bank, and the next What-if uses it instead of the "Easy / Average / Difficult" default. So predictions improve as you use the system, but they stay estimates.
Using the results well
- Treat a Retain, Revise or Reject suggestion as a prompt to read the question again. You decide whether to change it.
- Treat the at-risk list as private. Only the class teacher and the department's Program Head can see it; it is never shown to students. Use it to offer help early (a consultation, remedial work, extra practice). Do not post it, tell other students, or use it to punish or to grade.
- The Program Head sees department summaries without student names, and can open a section to read it and add notes. Only the class teacher accepts or dismisses a recommendation.
- Teaching alternatives are general ideas, not instructions. Pick one that fits your class, or none.
- If a wrong answer key warning appears, check the key yourself. If you correct it, papers are re-scored and the correction is logged.
- If an answer pattern flag appears, look at the papers and the seating, talk to the students if needed, and follow the school's rules. The decision belongs to you and the school, never to the system.
- Check generated test papers and answer keys before you give them out.
Questions or concerns
If a result looks unfair or wrong, please tell us through the Contact and Support page. Students who think a score or flag is wrong can ask their teacher for a re-check. See the Student Data Notice.