Home of the Uedu platformShow a desktop notification when the AI TA finishes replying
Notify me when classmates post messages in the forum
Play an alert sound whenever there is a new notification
The hardest part of setting AI usage guidelines is not finding a template, but writing clearly enough about “how far it may be used” for TAs to judge it correctly. Here, first self-assess your current governance status, then generate a chapter-by-chapter draft of the guidelines based on your situation.
Most units get stuck in three places: not knowing which sections the framework should include, not knowing how to write usage levels as criteria that can be judged, and not knowing how far they still are from being "operational".
The self-assessment for the five dimensions uses fixed scoring, so the same submission will always receive the same score. The score itself shows where improvements are needed, and also makes changes visible when the same unit is assessed again a year later.
The draft maps each type of assignment to usage levels from L0 to L4, and states the disclosure method required at each level. When a Student asks, "Can I use AI for this?", there is a clear answer.
Clauses that the research is not tied to grades, the rules must be announced in advance, and detection results may not be used as the sole basis for judgement are written directly into the draft by the system, and are not left to the model to decide whether to generate.
Course-level rules are written for Students to read and can be pasted directly into the syllabus; institution-level rules are written for Instructors and administrative staff and can go through the approval process. The two do not share templates.
單一課程的 AI 使用規範，寫完可直接放進教學大綱或課程公告。
Reader:修課學生與教學助理
校、院或系級的 AI 使用政策框架，供內部審議程序使用。
Reader:校內教師、行政單位與學生代表
Grading is deliberately capped at five levels. Any finer and Teachers cannot rate it, and TAs cannot judge it reliably. Each level also specifies both 'what can be done' and 'how to disclose it'; without the latter, the rules can only rely on post hoc detection.
The five dimensions are each divided into four levels, for a total of 15 points. Each level is written as an observable state (whether there is this document, whether there is this mechanism), rather than an imprecise description such as “fairly well developed”.
| Orientation | Why view this item | Top-tier appearance |
|---|---|---|
| 規範明確度 | 沒有成文規範時，判定會落到個別教師身上，同一所學校不同課程會出現互相矛盾的標準，學生無所適從。 | 有正式文件且對各類評量分別載明可用程度 |
| 揭露機制 | 揭露機制是整套規範能否運作的樞紐。沒有揭露，規範就只能靠事後偵測，而偵測工具的誤判會傷到誠實的學生。 | 有規定格式，且未揭露會依既定程序處理 |
| 評量調整 | 若評量方式完全不動，只加一條「不准用 AI」，實質上是把規範的執行成本轉嫁給教師的偵測能力。 | 評量已重新設計為看得見歷程與判斷，而非只看成品 |
| 資料治理 | 對話歷程含有大量可辨識資訊。沒有保存期限與存取權限的規定時，這些資料的風險會隨時間累積。 | 存取權限、保存期限、刪除程序與學生權利皆已明訂 |
| 師生培力 | 規範講的是不能做什麼，培力講的是能做到什麼。只有前者的單位，實務上會退化成全面禁用。 | 教師專業發展與學生 AI 素養課程皆已納入常態規劃 |
The following clauses are appended directly by the system and are not generated by the model. This is the platform's established position; of course you may modify it in your own final document, but the draft we provide will always include it.
| Jurisdiction | Document | Related items |
|---|---|---|
| 台灣 | 個人資料保護法 | 學習歷程與對話紀錄屬個人資料，蒐集目的、告知義務、當事人權利行使方式均須依此法規劃。 |
| 台灣 | 教育部生成式 AI 教學應用相關參考指引 | 主管機關對教學場域使用生成式 AI 的原則性指引，訂定校級政策時通常需引為依據。 |
| 台灣 | 行政院及所屬機關使用生成式 AI 之參考指引 | 公立學校的行政作業使用 AI 時，通常需與此指引的機敏資訊處理原則一致。 |
| 歐盟 | EU AI Act | 將教育場域的部分 AI 應用列為高風險，並限制在教育情境推論情緒狀態。有國際合作或境外學生時值得對照。 |
| 歐盟 | GDPR | 涉及歐盟境內學生資料時適用，對自動化決策與資料主體權利的要求高於本地規範。 |
Policy documents often need to explain 'how Student data is processed on the platform'. The following are governance facts already made public by Uedu; draft text should quote only these and must not promise anything the platform cannot do.
The self-assessment takes about 2 minutes, and draft generation takes about 30 to 60 seconds. Outputs can be saved, reviewed later and printed.
Start after logging in