01
將檢測得分定位為輔助線索
AIGC 疑似度評分不應作為評判學術失信的唯一標準，而應作為發現和定位需要人工審閱、輔導或關懷段落的輔助線索。
多維度分析學生文稿的 AIGC 疑似特徵。建議結合學生的寫作大綱、筆記及日常表現進行綜合評估，避免單憑百分比得分下結論。
Document input
僅影響檢測規則，不會翻譯正文。
檢測要點
AIGC 檢測面板
AI 特徵預估
貼上或上傳文稿後開始檢測，這裡會顯示分數與分句信號。
預檢
01正文字元
0
02詞數
0
03語言
繁體中文
04來源
貼上
指南
多維度分析學生文稿的 AIGC 疑似特徵。建議結合學生的寫作大綱、筆記及日常表現進行綜合評估，避免單憑百分比得分下結論。
步驟 1
文稿特徵掃描
步驟 2
比對歷史習作
步驟 3
私下溝通與輔導
指南
將檢測得分定位為輔助線索
防範潛在的演算法誤判
結合寫作過程性證據
01
AIGC 疑似度評分不應作為評判學術失信的唯一標準，而應作為發現和定位需要人工審閱、輔導或關懷段落的輔助線索。
02
非母語者寫作、高度程式化的學術報告，或者經過語法校對軟體大幅修正潤飾的文本，均可能被演算法判定為偏高的 AIGC 疑似特徵。
03
建議結合學生的心智圖、文獻卡片或歷史修改版本。親自撰寫文稿的學生通常能夠清晰地闡述其觀點取捨與論證邏輯。
What we check
The teacher page is built around classroom review, where text length, assignment format, and student history all change how a detector result should be read.
Scan the full paper and inspect flagged passages instead of acting on one global percentage.
Keep citations and formal prose in context, because academic style often looks more predictable than casual writing.
Discussion posts and lab notes can be noisy. Short-text flags should be compared with classroom writing baselines.
Use the score as a review trigger before grading, then ask process questions where the signal is concentrated.
Teacher workflow
The classroom problem is not only whether a detector can find AI-like text. The real issue is whether the teacher can make a fair, defensible decision after the score appears.
Use TurnitPass when the school LMS score feels too thin or when you need a sentence-level review before acting.
Compare sentence rhythm, vocabulary, and argument structure with in-class writing or earlier submissions.
Use the result to ask better process questions, not to replace academic integrity judgment.
Review path
Cross-check the result
If only one system flags the work, treat the signal as weak. Agreement across tools is stronger but still not a verdict.
Review the flagged passages
Look for concentrated blocks in introductions, conclusions, and transition paragraphs rather than random isolated lines.
Ask for process evidence
Draft history, source notes, outlines, and a private explanation usually matter more than any single score.
False positives
False positives are part of every detector workflow. A fair review process keeps the score, the text, and the student conversation together.
If only one system flags the work, treat the signal as weak. Agreement across tools is stronger but still not a verdict.
Look for concentrated blocks in introductions, conclusions, and transition paragraphs rather than random isolated lines.
Draft history, source notes, outlines, and a private explanation usually matter more than any single score.
Compare the workflow around the score, not just the existence of a detector.
TurnitPass
Classroom signal plus revision context
Breakdown
Sentence and pattern guidance
Next step
False-positive workflow
Turnitin AI
Institutional bundled score
Breakdown
Limited visible evidence
Next step
Depends on school policy
GPTZero
Single detector classifier
Breakdown
Sentence details on some plans
Next step
Requires teacher process review
Classroom policy
Treat the scan as a review trigger. The useful workflow keeps the score, the flagged passage, the student process, and the final decision in the same record.
Policy
Define what AI use is allowed, disclosed, or prohibited before assignments begin.
Evidence
Pair detector estimates with drafts, notes, prior writing, and a private explanation.
Decision
Do not penalize from a score alone; document the full review context.
流程
Review record
01
Draft source
Full submission + class context
02
Detector signal
Clustered passages, not a verdict
03
Next action
Conversation or no action
Further reading
不能。疑似度得分只能作為輔助線索，必須與寫作過程性證據及教師的專業學術判斷相結合。
不會。系統僅提供即時文本掃描與分析，不保存任何學生名冊或作業，全面保障學生隱私。