Plagiarism & integrity.

Let us be precise about what this module is, because the word "plagiarism" invites expectations it does not meet. This is not Turnitin. It does not search the web, published papers, other assignments, or previous years. It has no external service behind it and no AI paraphrase detector. What it is: a guard, built entirely inside the database, that catches the most common academic shortcut in a class — one student copying another's answer on the same assignment.

The whole feature is one table of flags and one review screen at /dashboard/lms-plagiarism. There is nothing for a teacher to run, schedule, or configure. A flag appears when the database notices two submissions are too alike; a teacher reads it and either lets it stand or dismisses it as a false positive. That is the entire loop, and its honesty is the point — a school knows exactly what it is relying on.

Catch the obvious copy between classmates, surface it plainly, and leave the judgement to the teacher.

The detection is automatic and runs at the moment a student submits. When a text submission is written or updated, a database trigger wakes up, normalises the text (lower-cased, whitespace collapsed), and compares it against every other submission on the same assignment in the same school. It works in two passes:

A few honest boundaries baked into the trigger: it only looks at text submissions of a reasonable length (very short answers are skipped, since two one-liners are naturally alike); it only ever compares within a single assignment; and it never reaches outside the school. File uploads, PDFs, and images are not analysed at all — only typed text. And the whole check is best-effort: if the text-similarity capability is unavailable, the trigger simply skips rather than blocking the submission.

A teacher or admin with LMS-assignment view rights opens /dashboard/lms-plagiarism. The screen is a single board — every flag in the school, split into what needs looking at and what has already been handled. Across the top, four counts:

• Active — flags not yet dismissed; the working queue.
• High severity — the strongest matches, the ones to open first.
• Exact match — the character-for-character copies.
• Dismissed — the flags already cleared as false positives.

Each active flag is a card carrying its similarity, a tag saying whether it was an exact or a near match, when it was detected, and the identifiers of the two submissions involved — sorted so the strongest matches sit at the top. The only action is Dismiss as false positive: the reviewer must type a short reason (a template answer everyone was told to reuse, a shared quotation, a legitimate collaboration), and the flag moves into the history section with that reason attached. There is no un-dismiss — clearing a flag is a considered, recorded decision.

Being clear about the ceiling is as useful as describing the floor, so a school can size its expectations correctly:

• No web or corpus check. A student who copies from a website, an encyclopedia, or a friend at another school is not caught — only similarity between two submissions on the same assignment is measured.
• No cross-assignment or cross-year memory. Reusing last year's essay, or the same answer on a different assignment, does not raise a flag.
• Text only. Uploaded files, PDFs, images, and audio are never analysed; a submission has to be typed text to be compared.
• Not paraphrase-proof. The near-match measure is character-based, so a thorough rewrite that keeps the ideas but changes the words will slip under the threshold.
• The teacher decides. There is no severity ladder that punishes on its own, and no student notification — a flag is information for a human, not a verdict.

None of this is a defect to hide; it is the deliberate scope of a minimum guard. If a school needs web-scale originality checking, that is a different, larger integration — noted in the roadmap below.

This module is intentionally the minimum guard, and its author labels it exactly that. The larger integrity product is roadmap, not shipped — here is the straight list.

The integrity check sits under the LMS assignment flow, feeds the teacher's judgement in the gradebook, and shares the anti-cheat instinct of the examinations engine.