星期四第三節,副校長坐在課室最後一行,手上是一張觀課表。四十分鐘裡,學生用 AI 生成了一份整齊的報告,匯報時聲音清楚,組員輪流回答提問。表格上有十二格,副校長在「學生參與度」填 4,「資訊科技運用」填 5,然後在「建議」一欄寫下「時間控制可以更好」。離開課室之前,他把表交回教務處。表上沒有一句學生講過的話,也沒有那份報告的樣本。On a Thursday in the third period, a vice-principal sits in the back row with an observation form in hand. For forty minutes the students use AI to produce a tidy report, present it clearly and take questions in turn. The form has twelve boxes. He marks 4 for participation, 5 for use of IT, and writes in the suggestions box that time control could be better. Before leaving he hands the form back to the general office. It carries no sentence a student said and no copy of the report.

這張表記錄的是觀課老師的判斷。一個 4 分說明當日有一位老師覺得不錯。半年之後,學校要向校董會、向教育局的中期報告或者向校外評核隊伍說明這節課帶來了甚麼,手上只有這個 4 分。我們在 2026 年 11 月 20 日讀了五張香港學校與教育局公開的觀課表,四張把每項判斷變成 1 至 5 分,五張合共八十個評分格,文字欄的欄名是意見、建議、備註與略述情況。The form records the observer's judgement. A 4 says that on the day one teacher thought the lesson was good. Six months later, when the school has to tell its management committee, the Bureau's mid-term report or an external review what the lesson achieved, the 4 is all it holds. On 20 November 2026 we read five observation forms published by Hong Kong schools and by the Education Bureau. Four turn each judgement into a score from 1 to 5, between them the five carry eighty scored boxes, and the text fields are labelled comments, suggestions, remarks and a brief description.

八十個評分格裡,最接近學生能力的一題是「學生運用後設認知能力解決問題」,它要的仍然是一個分數。要留住學生做過的事,表格要換成五欄,每欄留一句可以核對的話;填滿一頁大約十分鐘,之後的十分鐘用來開會。這條路徑在 AI 課堂上最有用,因為學生交上來的作品可以大部分由模型完成,而且通常比學生自己做的整齊。Among the eighty boxes, the item closest to a student capability asks whether students apply metacognitive ability to solve problems, and it still wants a number. Keeping what a student did takes five fields, each holding one sentence that can be checked; filling the page takes about ten minutes, and the ten minutes after the lesson are for the meeting. The route matters most in an AI lesson, where the work a student hands in can be mostly the model's, and usually tidier than the student's own.

摘要Summary

五張觀課表、八十個評分格,沒有一欄要求記錄學生學習的證據。一頁五欄的紀錄填得完十分鐘,之後的十分鐘用來開會。Five observation forms and eighty scored boxes, and not one field asks for evidence of student learning. A page of five fields takes ten minutes to fill, and the ten minutes after it are for the meeting.

  • 教育局通函第 156/2026 號要求學校在 2026/27 學年把數字教育寫進學校發展計劃,並把「同儕觀課」列為校內學習社群推動 AI 教學的做法[1]。Circular Memorandum No. 156/2026 requires schools to put digital education into the school development plan in 2026/27, and names peer lesson observation as one way a school's teacher group spreads AI teaching[1].
  • 同一份通函要求每名教師在三年 150 小時的持續專業發展周期內完成不少於 30 小時數字教育培訓,進階單元包括「評估革新與學習數據分析」,以及一項要「構思、執行及檢討」的 AI 課堂專題[1]。The same circular asks every teacher to complete at least 30 hours of digital education training inside each three-year cycle of 150 hours, and its advanced units include one on assessment and learning data and a project that asks the teacher to plan, teach and review one AI-integrated lesson[1].
  • 五張公開的觀課表之中,四張把每項判斷變成 1 至 5 分,五張合共八十個評分格。Four of the five published forms turn each judgement into a score from 1 to 5, and between them the five carry eighty scored boxes.
  • 五張表都留了文字欄,沒有一欄要求觀課老師寫下學生的原話,也沒有一欄要求留住學生的作品。All five leave space for a written comment, and none asks the observer to write down a student's own words or to keep the student's work.
  • 《中小學應用人工智能教學指南》的「評」一節要求教師用即時口頭查問、思考歷程與課堂內完成的課業查核學生有否濫用 AI,並把教師觀察列為評估數據的一種[2]。The assessment section of the EDB guide to using AI in teaching asks teachers to check misuse through a question asked in class, the student's record of thinking and work completed under supervision, and it counts a teacher's observation as assessment data[2].
  • 一節 AI 課要看得見三個位置:學生自己做的那一步、他解釋得出來的理由、他交上來的作品,這三樣由模型補不上[2][8]。An AI lesson has three places to watch: the step the student took, the reason he can give for it, and the work he hands in, and the model cannot supply any of the three[2][8].
  • 一份五欄紀錄之後答得了三個問題:中期報告的證據、校外評核看見的學與教、第二年續約的決定[1][9]。One five-field record later answers three questions: the evidence for the mid-term report, what an external review sees of learning and teaching, and whether to renew the course for a second year[1][9].

七點合起來是一個設計問題。評分記錄觀課老師的看法,句子記錄學生做過的事;學校要把後者放進表格,半年後的三個問題才有答案。The seven points come down to a question of design. A score records the observer's opinion and a sentence records what the student did, so the school has to put the second into the form before the three questions have an answer six months later.

觀課表問的是觀課老師的看法The form asks the observer for an opinion

四月的一個下午,同一張表回到副校長手上。這次觀的是外聘導師的 AI 課:學生用工具做了一個班會壁報,導師在旁邊提示,學生跟着做。表格上的「資訊科技運用」仍然是 5 分。離開課室的時候,他心裡有一個問題,表格上沒有位置寫:那個壁報,學生自己做過哪一部分?One April afternoon the same form is back in his hands. This time the lesson belongs to an outside instructor's AI course: the students use a tool to make a class noticeboard, the instructor prompts from the side and the students follow. The box for use of IT is another 5. Walking out, he has a question the form has no place for: which part of that noticeboard did the students make themselves?

表格要求觀課老師判斷,判斷的單位是一個分數。分數可以在離開課室之後填,也可以在走廊上填;證據做不到。學生講過的話在課室裡出現一次,作品在課後被收走或者丟掉,兩者都不會自己回到表格上。The form asks the observer to judge, and the unit of judgement is a score. A score can be filled in after leaving the room, or in the corridor. Evidence cannot. A sentence a student says appears once in the lesson, and the work is collected or thrown away afterwards, and neither returns to the form by itself.

AI 令這個問題更明顯。學生交上來的作品可以大部分由模型完成,而且通常比學生自己做的整齊,匯報的講稿也可以由工具起草。觀課老師看到的成品越漂亮,越難從成品判斷學生做過哪一步。課堂裡仍然屬於學生的三件事是:他決定了要做甚麼、他解釋得出為甚麼這樣做、他親手完成的那一部分。AI sharpens the question. The work a student hands in can be mostly the model's, and it is usually tidier than the student's own; the words in the presentation can be drafted by a tool as well. The more polished the product looks, the harder it becomes to tell from it which step the student took. Three parts of the lesson still belong to the student: he decided what to make, he can explain why he made it that way, and he finished one part with his own hands.

一張留得住學習證據的觀課表,會在這三件事出現的時候把它們寫下來。分數記錄的是觀課老師,句子記錄的是學生。一份有句子的紀錄,半年後仍然答得出「這節課學生做過甚麼」,因為它抄下了當時的答案。A form that keeps evidence of learning writes these three down while they are happening. A score records the observer; a sentence records the student. A record holding sentences still answers the question of what the students did six months later, because it copied the answer at the time.

四十分鐘裡,一位觀課老師能看到多少?一班 30 人,坐在最後一行,他看得清楚的是前排幾個;走近一點,一次只能停在兩三張桌子旁邊。所以觀察一定要有焦點:一節課看一欄,下一節換另一欄。表格要求他一次判斷十二項,等於每一項分到三分鐘,其中大部分時間用來看教師。How much can one observer see in forty minutes? With thirty students in the room, the back row gives a clear view of the first few desks; a walk round stops beside two or three tables at a time. That is why an observation needs a focus, one field this lesson and another the next. A form that asks for twelve judgements gives each one about three minutes, and most of those minutes go on watching the teacher.

取消評分是另一個題目。學校要評核教師,觀課是其中一個來源,分數有它的用途;這裡講的是同一次觀課的另一個用途:了解學生在課堂裡做到了甚麼,並用同一份紀錄決定下一節改哪一樣。兩個用途可以分開兩份文件,也可以在一份文件裡分開兩欄;做不到的,是用同一個 4 分兼做兩件事。Abolishing the score is a separate question. A school appraises its teachers, observation is one source for that, and a score has its use; the other use of the same forty minutes is to find out what the students did and to decide on one change for the next lesson. The two uses can live in two documents, or in two parts of one document. What does not work is one 4 doing both jobs.

通函把同儕觀課寫進教師的培訓要求The circular puts peer observation inside the training requirement

這一年的觀課有一份時間表。教育局通函第 156/2026 號要求學校在 2026/27 學年把數字教育寫進學校發展計劃及/或學校周年計劃,並要求每名教師在每三年 150 小時的持續專業發展周期內完成不少於 30 小時數字教育培訓[1]。This year's observation comes with a timetable. Circular Memorandum No. 156/2026 requires schools to put digital education into the school development plan or the annual plan in 2026/27, and it asks every teacher to complete at least 30 hours of digital education training inside each three-year cycle of 150 hours[1].

30 小時乘以教師人數,是一間學校三年內要安排的培訓時數。60 名教師就是 1,800 小時,相等於 300 個六小時的教師發展日。一個數字解釋了學校為甚麼要把觀課放進培訓計劃:觀課可以在校內發生,由學校自己的人帶,紀錄留在學校手上。Multiply 30 hours by the staff and the total is what a school has to arrange inside three years. Sixty teachers come to 1,800 training hours, which is 300 staff development days of six hours. That number explains why observation belongs in the training plan: it happens on site, it is led by the school's own staff, and the record stays with the school.

通函附錄的進階單元把要求寫得更具體。第 9 項叫「總結專題」,要求教師「構思、執行及檢討一項融入人工智能的課堂、教學單元或校務安排」;第 8 項要求教師「透過反思及實證為本的方式,策劃、推行及檢視教與學層面的人工智能創新措施」;第 5 項是「人工智能下的評估革新與學習數據分析」[1]。三項合起來,是一位教師要設計一節課、教一次,然後拿出證據檢討。The advanced units in the circular's annex are more specific. Unit 9, called the concluding project, asks a teacher to plan, teach and review one AI-integrated lesson, teaching unit or administrative arrangement; unit 8 asks for AI initiatives in learning and teaching to be planned, delivered and examined in a way that is evidence-based; unit 5 covers assessment and the analysis of learning data[1]. Together they describe a teacher who designs one lesson, teaches it, and then produces the evidence to review it.

通函在「教師專業層面」一節點名同儕觀課:鼓勵學校成立「數字教育教師小組」,推動同儕觀課、經驗分享與專業交流[1]。觀課因此由人事評核的一環,走進了培訓要求;培訓要求的那一半,要的是檢討的證據。The section on the teacher as a professional names peer observation: schools are encouraged to form a digital education teacher group that promotes peer observation, sharing and professional exchange[1]. Observation therefore moves from appraisal into the training requirement, and the training requirement asks for the evidence of a review.

這就解釋了表格為甚麼忽然重要。一張只記錄觀課老師看法的表,可以滿足程序,滿足不了這一節:通函要求的是檢討,檢討要有材料。學校今年要交的是 30 小時的紀錄與一節課的檢討,兩份文件可以由同一張五欄表產生。That is why the form suddenly matters. A page that records only the observer's opinion satisfies the paperwork and not this clause: the circular asks for a review, and a review needs material. What the school has to show this year is 30 hours of training and the review of one lesson, and one five-field page can produce both.

五張表加起來有八十個評分格The five forms carry eighty scored boxes between them

我們在 2026 年 11 月 20 日讀了五張公開的觀課表。方法是搜尋「觀課表」與「觀課紀錄表 site:edu.hk」,把找得到的表格逐項讀完,數它們要求觀課老師填甚麼。五張表分別來自教育局的資優教育指引、教育局與賽馬會「校本多元」計劃的學習支援材料、一間小學的學校報告、一間中學的先導計劃,以及一間特殊學校的自評工具。On 20 November 2026 we read five published observation forms. The method was to search for 觀課表 and for 觀課紀錄表 on school sites, open every form the results led to, and count what each one asks the observer to fill in. They come from the Bureau's gifted education guidelines, from learning support material produced with the Jockey Club, from a primary school's annual report, from a secondary school's pilot scheme, and from a special school's self-evaluation tools.

表格Form 出版者與年份Publisher and year 評分項Scored items 文字欄Text field
同儕觀課評估表Peer observation form 教育局資優教育指引附件十二[3]Bureau gifted education guidelines, annex 12[3] 18 項,全部以 1 至 5 分評級。Eighteen items, each scored 1 to 5. 一個建議欄,另有一行要求列明原因。One suggestions box and one line asking why.
觀課記錄表Lesson observation record 教育局與賽馬會「校本多元」計劃[4]Bureau with the Jockey Club diversity project[4] 42 項,每項以 1 至 5 分評級,另設不適用。Forty-two items scored 1 to 5, with a not-applicable option. 每行一個意見欄。A remarks column on every row.
同儕觀課紀錄表Peer observation record 聖士提反女子中學附屬小學,2017-18 學校報告[5]St Stephen's Girls' Primary School, 2017-18 report[5] 12 項,以 1 至 5 分評級。Twelve items scored 1 to 5. 五組優點與建議欄,另設教師自評頁。Five pairs of strengths and suggestions, plus a page for the teacher.
課堂觀察評估表Classroom observation form 高主教書院,2024-25 先導計劃[6]Raimondi College, 2024-25 pilot[6] 8 項,以 1 至 5 分評級。Eight items scored 1 to 5. 公開樣本沒有文字欄。The published sample has no text field.
觀課紀錄表Observation record 甘迺迪中心,校本自評工具[7]Kennedy Centre, self-evaluation tools[7] 沒有評分,每項勾選一個描述。No score; each item ticks one description. 一欄略述情況,一格其他意見。One line for a brief description and one box for other comments.

四張表把判斷變成數字,第五張把判斷變成勾選。八十個評分格裡,題目寫的是「本課節能達到既定的教學目標」、「學生積極參與」、「批判性思考能力」、「課堂氛圍」。最接近學生能力的一題是「學生運用後設認知能力解決問題」,它要的仍然是一個 1 至 5 的分數。Four forms turn the judgement into a number and the fifth turns it into a tick. Across the eighty boxes the items read: the lesson achieves the stated objectives, students take part actively, critical thinking, classroom atmosphere. The item closest to a student capability asks whether students apply metacognitive ability to solve problems, and it still wants a score from 1 to 5.

表格長成這樣,與它們原本的用途有關。五張表都是評估工具,用途寫在標題或者出處裡:試教評估、同儕評估、教師自評、學校自評。評估要比較,比較需要數字,評分表就是這樣長出來的。數字方便比較,句子方便理解;一份用來比較的表格,自然以分數為主,學生的話放在備註裡。The shape of the forms follows from what they were built for. All five are evaluation instruments, and the purpose is in the title or the source: a trial lesson, peer evaluation, teacher self-evaluation, school self-evaluation. Evaluation compares, comparison wants numbers, and a rating sheet is what that produces. Numbers compare well and sentences explain well; a form built to compare fills itself with scores and puts the student's words in a remarks box.

五張表都有文字欄,欄名是建議、意見、優點、備註與略述情況。觀課老師可以在這些欄裡寫任何事,表格沒有要求他寫下學生講過的那句話,也沒有要求他留住學生交上來的作品。三張表的欄目夠闊,一位認真的觀課老師放得進證據;表格本身沒有提出這個要求。All five carry a text field, labelled suggestions, comments, strengths, remarks or a brief description. An observer may write anything in them, and the form never asks for the sentence a student said or for a copy of the work the student handed in. Three of the fields are wide enough for a careful observer to put evidence in; the form does not ask for it.

這是搜尋找得到的五張表,樣本偏重小學、資優與特殊教育,因為這幾類學校較多把表格公開;兩張沒有日期,一張是 2017-18 年度。重點在工具的年齡。表格比 AI 課更老:它們在 AI 進入課室之前已經定形,現在用來觀察 AI 課。These are the five forms a search returned, and the sample leans towards primary, gifted and special education, because those schools publish their forms more often; two carry no date and one comes from 2017-18. The point is the age of the instrument. The forms are older than the AI lesson: they were settled before AI reached the classroom, and they are now used to observe it.

有一間學校已經拿走分數One school has already removed the score

拿走分數的學校存在。陳瑞祺(喇沙)書院的 2018-19 周年校務報告描述校內的「同儕互助觀課計劃」:觀課前共同備課,觀課期間不使用量表,觀課後議課,三部分都寫進同一份紀錄表[10]。A school has already done it. The 2018-19 annual report of Chan Sui Ki (La Salle) College describes a peer observation scheme in which colleagues plan the lesson together beforehand, observe without a rating scale, and hold a discussion afterwards, with all three parts written into one record[10].

這是一個真實的改進。觀課老師不用把一節課壓成幾個數字,紀錄變成敘述,議課有具體內容可以討論。That is a real improvement. The observer no longer compresses a lesson into a handful of numbers, the record becomes a narrative, and the discussion afterwards has a specific point to work on.

紀錄的重心仍然是會議。觀課老師填「觀課」與「議課」兩欄,學生的作品只有在他選擇描述的時候才出現。拿走分數只是第一步,第二步要有人決定:紀錄裡要留住學生的哪一句話、哪一件作品。The record still centres on the meeting. The observer fills the observation column and the discussion column, and the student's work appears only if the observer chooses to describe it. Removing the score is the first step. The second needs someone to decide which sentence and which piece of work the record keeps.

第二步寫在另一份材料裡。中文大學 QSIP 的《科組領導手冊》單元三講觀課議課,紀錄的要點有三條[8]:The second step is written down elsewhere. Unit three of the QSIP subject-leadership handbook, published by the Chinese University's school improvement programme, sets out three points for the record[8]:

  1. 客觀地紀錄事實。Record the facts objectively.
  2. 記錄學生回應教師問題的頻次、對答情況與參與程度。Record how often students answer, what they say and how far they take part.
  3. 拍攝學生的課堂表現與習作樣本,帶回議課。Photograph what students do in the lesson and samples of their work, and bring them to the discussion.

手冊的議課問題有一條:「有多少學生能掌握課堂的重點?如何知悉?」這一問把紀錄的重心從「我覺得」推到「我見到」。教育局的《中小學應用人工智能教學指南》走得更遠,直接把教師觀察列為評估數據的一種[2]。One of the handbook's questions for the discussion asks how many students grasped the point of the lesson and how the observer knows. That question moves the record from what I felt to what I saw. The Bureau's guide to using AI in teaching goes further and counts a teacher's observation as one kind of assessment data[2].

這套做法的成本可以算出來。一次完整的觀課循環包括 20 分鐘的課前會議、40 分鐘的課堂、20 分鐘的議課,一位教師一個循環 80 分鐘;60 名教師輪一次,是 80 小時,分散在一個學年裡,每星期不到三小時。與 30 小時的培訓要求相比,這個時間不大,而它產生的紀錄正好是培訓要求要檢討的那份證據。The cost of the practice can be counted. One full cycle runs twenty minutes of meeting before the lesson, forty minutes in the room and twenty minutes after it, which is eighty minutes per teacher; sixty teachers come to eighty hours spread across a school year, under three hours a week. Against a requirement of 30 hours of training, that is a small cost, and the record it produces is the evidence the requirement asks to be reviewed.

一節 AI 課要觀察甚麼What to watch in an AI lesson

一節 AI 課要看得見三件事:學生自己做的那一步、他解釋得出來的理由、他交上來的作品。這三件事由模型補不上,觀察它們,就看得出學生有沒有在學習。An AI lesson has three points worth watching: the step the student took, the reason he can give for it, and the work he hands in. The model cannot supply any of the three, and watching them shows whether learning is happening.

《中小學應用人工智能教學指南》的「評」一節沒有用「觀課」兩個字,它把同一個問題寫得很清楚:AI 可以令評估由學習成果走向學習過程,教師要判斷的是學生有沒有把思考外判給工具。指南給出三種查核方法[2]。The assessment section of the Bureau's guide never uses the word observation, and it states the same problem plainly: AI lets assessment move from the finished result towards the learning process, and the teacher has to judge whether the student has outsourced the thinking to the tool. The guide names three ways to check[2].

指南的查核The check 它要找出甚麼What it finds out 紀錄要留住甚麼What the record keeps
即時口頭查問A question asked in class 學生能否即場解釋自己的選擇。Whether the student can explain a choice on the spot. 學生的原話,以及當時問的那條問題。The student's words and the question that produced them.
思考歷程The record of thinking 學生走過哪幾步,哪一步由他決定。Which steps the student took and which one he decided. 步驟的先後,以及他停下來的地方。The order of the steps and where he stopped.
課堂內完成的課業Work finished in class 這件作品在課室裡做得出來嗎?Whether this piece could be made inside the room. 作品的相片、存放位置與完成時間。A photograph, where the file sits and when it was finished.

指南另外寫了兩句可以直接放進校內指引的話。「學生不應直接提交由人工智能生成的課業。」「任何情況下,教師不應未經專業判斷便直接採用人工智能給予學生的評分和回饋。」[2] 兩句合起來的意思很直接:分數要由教師判斷,教師手上就要有判斷的材料。The guide carries two sentences a school can lift straight into its own policy. Students should not submit work generated by AI as their own, and under no circumstances should a teacher adopt a score or comment produced by AI without professional judgement[2]. Together they say one thing simply: the score belongs to the teacher, and the teacher needs material to base it on.

指南還要求學校為 AI 在課業與評估中的使用訂立準則,並列明部分課業(例如寫作)只適宜在課堂及教師監督下完成[2]。這一條把觀察的位置向前移:判斷要在作品出現的時候做,成品交上來之後已經太遲。The guide also asks schools to set rules for AI in coursework and assessment, and notes that some work, writing among them, is better done in class under supervision[2]. That moves the observation earlier: the judgement has to happen while the work is being made, because by the time it is handed in it is too late.

觀課表上要填的是甚麼What an observation form should hold

一頁 A4,五欄,十分鐘填得完。五欄是任務與學生那一步、學生的一句原話、交上來的作品、AI 交出了甚麼、下一次的一條決定。前四欄是這一節課的學習證據,第五欄把它接到下一節。One A4 page, five fields, ten minutes to fill. The fields are the task and the student's step, one sentence the student said, the work handed in, what the AI produced, and one decision for next time. The first four are the evidence of learning in this lesson and the fifth joins it to the next one.

欄Field 填甚麼What goes in it 證明甚麼What it shows 誰填Who fills it
任務與學生那一步The task and the step 這節課的任務,以及哪一步由學生自己完成。The task for the lesson and which step the student did alone. 學生做過的工作。The work the student did. 觀課老師,課前讀任務紙。The observer, from the task sheet before the lesson.
一句原話One sentence 學生被問「為甚麼這樣做」之後講的那句話,逐字記錄。The sentence a student says when asked why he did it that way, written word for word. 學生理解自己的選擇。That the student understands his own choice. 觀課老師,課堂內。The observer, during the lesson.
交上來的作品The work handed in 作品的相片或檔案位置,以及哪一部分由學生完成。A photograph or the file location, and which part of it the student made. 學生獨立完成的部分。The part the student finished without help. 觀課老師,課後十分鐘。The observer, in the ten minutes after.
AI 交出了甚麼What the AI produced 同一個任務由工具產生的版本,一段就夠。The tool's version of the same task, one passage is enough. 學生的工作與工具的工作相差多遠。How far the student's work sits from the tool's. 觀課老師。The observer.
下一次的一條決定One decision for next time 下一節改哪一樣、誰負責、幾時再做。What changes next lesson, who carries it and when it happens. 這份紀錄有沒有帶來改動。Whether the record changed anything. 觀課老師與授課教師,議課時一起填。The observer and the teacher, together in the discussion.

五欄的次序有作用。先寫任務,再寫學生,最後才寫工具;寫的過程裡,觀課老師會發現自己看見哪幾欄、看不見哪幾欄。看不見的一欄,就是課後要問的問題。The order does work. The task first, the student second and the tool last, because filling it in that way shows the observer which fields he can see and which he cannot. The field he cannot fill is the question to ask afterwards.

一句原話勝過一個四分,因為句子半年後仍然在,而且可以核對:同一個問題在學期末再問一次,學生的答案會告訴學校他記住了幾多。A sentence is worth more than a 4, because the sentence is still there in six months and can be checked: ask the same question at the end of term and the student's answer tells the school how much stayed.

這張表不打分,所以它排不出教師的名次。它是一份紀錄,用來說明一節課裡學生做了甚麼,以及下一節改哪一樣。學校要用同一份紀錄評核教師,就要另設一份文件,或者在同一頁上分開兩欄,讓兩件事各有各的位置。The form carries no score, so it cannot rank teachers. It is a record of what the students did in one lesson and what changes in the next, and a school that wants the same page for appraisal needs a second document, or two clearly separate parts of one page, so that each job has its own place.

五欄填完之後,紀錄要有人收。收的人在科組,不在教務處:科主任每一節收一份,按月看一次,把同一欄的句子放在一起讀。六節課之後,他手上有的是一班學生在六個任務裡做過的事,這份材料可以用來排下一個單元,也可以用來回答家長的提問。Someone has to collect the pages, and the subject panel is where it works: the panel head takes one page per lesson, looks through them once a month and reads the sentences in the same field together. After six lessons he holds what one class did across six tasks, which is material for planning the next unit and for answering a parent's question.

一節課填出來是這個樣子What one filled page looks like

示範用的是一節中學數字人課:約 30 名學生,六節各 90 分鐘,取自我們的課堂紀錄[11]。紀錄沒有學校名稱,學生的一句話在取得書面同意後匿名引用。The example comes from one secondary-school digital human course: about thirty students over six sessions of ninety minutes, taken from our delivery record[11]. The record carries no school name, and the student's sentence is quoted anonymously under written consent.

示範紀錄:一節數字人課的五欄A sample record: five fields from one digital human lesson

欄Field填好的內容What it says
任務與學生那一步The task and the step學生要設計一個 DSE 口試考官,指令要它指出邏輯漏洞。提示詞由學生自己寫,考官要嚴到甚麼程度也由他決定。Students design a DSE oral examiner and instruct it to point out gaps in reasoning. The student writes the prompt, and the student decides how strict the examiner should be.
一句原話One sentence「平時舉手答錯驚俾全班笑,對住個 AI 考官錯十次都唔尷尬,反而係我自己唔忿氣想拗贏佢。」一位中四學生說,書面同意日期 2026-10-16。"In class, getting it wrong means the whole room laughs. With the AI examiner I can be wrong ten times and it is fine; what keeps me going is that I want to win the argument." A Secondary 4 student, written consent dated 16 October 2026.
交上來的作品The work handed in提示詞的文字,以及一段對話紀錄。學生自己寫的三句規則是不准寒暄、要指出邏輯漏洞、要追問理由。The prompt text and a transcript of one exchange. The three rules the student wrote are: no small talk, name the gap in the reasoning, and press for the reason.
AI 交出了甚麼What the AI produced考官的提問與追問。第一次讚得太早,學生把規則改嚴之後,追問才出現。The examiner's questions and follow-ups. It praised too easily at first, and the follow-up questions appeared only after the student tightened the rules.
下一次的一條決定One decision下一節的頭 45 分鐘鎖定純文字邏輯練習,之後才開語音;導師在第三節前確認。The first forty-five minutes of the next lesson stay text-only for the logic work, and voice opens afterwards; the instructor confirms before the third session.

這份紀錄沒有分數。它證明兩件事:那三句規則由學生自己寫,而且他在考官讚得太早時把規則改嚴。這一步是他自己做的,紀錄寫得出來。它證明不了一件事:一個月後這個學生還講不講得出同樣的話;那個問題要另一種測試,學校與家長可以用同一個方法量。The record carries no score. It shows two points: the three rules are the student's own, and he tightened them when the examiner praised him too easily. That step is his own work, and the record can show it. It cannot show one thing: whether the student can still say the same a month later. That needs a different test, one a school and a parent can run the same way.

五欄填完大約十分鐘,其中一半時間用在第二欄:等一條值得記下來的回答。學生不會自動講出理由,觀課老師要問,而且要問幾次;同一個句式「你為甚麼這樣做」在課堂裡問上三次左右,才會遇到一句放得進表格的話。The five fields take about ten minutes, and half of that goes into the second one, waiting for an answer worth writing down. Students do not volunteer their reasons, so the observer asks, and asks more than once; a question put as "why did you do it that way" has to be asked three or four times in a lesson before one answer is worth a line on the page.

觀課之前要談好甚麼What to agree before the lesson

三件事:這一節看哪一樣、哪一步由學生自己完成、授課教師自己想改哪一樣。三件事談完大約 15 分鐘,可以在課前會議完成,也可以在走廊上完成。Three points: what this lesson is being watched for, which step belongs to the student, and what the teacher wants to change. Fifteen minutes settles all three, in a meeting before the lesson or in the corridor.

第一件決定紀錄的焦點。QSIP 的手冊要求觀課前先協商檢視焦點,並讓施教教師清楚了解觀課的目的、焦點和安排[8]。焦點寫成一句話,例如「這一節看學生能否自己寫出提示詞」;一句話的焦點,五欄都填得滿,因為觀課老師知道要等哪一個動作。The first settles the focus. The QSIP handbook asks for the focus to be negotiated beforehand and for the teacher to know the purpose, the focus and the arrangements[8]. Write the focus as one sentence, such as whether students can write the prompt themselves, and all five fields become fillable, because the observer knows which moment to wait for.

第二件決定第一欄填不填得成。任務紙要寫明哪一步由學生完成,觀課老師拿着它進課室,才知道要記錄哪一個動作。手上沒有任務紙,第一欄只能課後補,補出來的是回憶。The second decides whether the first field can be filled at all. The task sheet states the step the student takes, and the observer carries it into the room so that he knows which moment to record. Without the sheet the first field is written afterwards, and what gets written is a memory.

第三件把觀課接到教師自己的目標上。授課教師講一件他想改的事,觀課老師在課堂裡特別留意那一件;紀錄的第五欄由這一件開始,會議的第三條問題也由它開始。The third ties the observation to the teacher's own aim. The teacher names one thing he wants to change, the observer watches for it, and the fifth field starts from it, which is where the third question in the meeting starts too.

觀課老師同時要決定自己坐在哪裡、會不會巡視學生的桌面。QSIP 的手冊允許觀課老師巡視學生的學習進度,條件是不影響課堂氣氛[8]。巡視得越近,第一欄與第三欄填得越實;坐在最後一行,寫下的多半是第二欄。The observer also decides where to sit and whether to walk round the desks. The handbook allows a walk round to check how students are getting on, on condition that it does not disturb the room[8]. The closer the observer gets, the more the first and third fields hold; from the back row most of what gets written belongs in the second.

觀課之後十分鐘要問甚麼What the ten minutes after the lesson are for

三條問題,按次序問,總共十分鐘。第一條由授課教師答,第二條由觀課老師用紀錄答,第三條由兩人一起答,並寫進第五欄。Three questions, asked in order, and ten minutes in total. The teacher answers the first, the observer answers the second from the record, and the two of them answer the third together, which then goes into the fifth field.

  1. 這節課你想學生做到哪一步?Which step did you want the students to take in this lesson?
  2. 有多少個學生做得到,你怎麼知道?How many students took it, and how do you know?
  3. 下一節改哪一樣?What changes in the next lesson?

次序有作用。先讀句子,後講意見;觀課老師把第二欄的句子讀出來,授課教師通常會先解釋,解釋的內容往往就是下一節要改的地方。句子讀得越慢,解釋得越具體。The order does work. The sentence comes before the opinion: the observer reads out what the student said, the teacher usually explains, and the explanation tends to contain the change for next lesson. The slower the sentence is read, the more specific the explanation.

QSIP 手冊的「有多少學生能掌握課堂的重點?如何知悉?」放在第二條,因為它把「我覺得」擋在門外[8]。十分鐘不夠的時候,先保留第二條;沒有第二條,會議會變成兩個人的印象交換。The handbook's question about how many students grasped the point, and how the observer knows, sits in second place because it keeps "I thought" outside the room[8]. When ten minutes will not stretch, keep the second question; without it the meeting becomes two people trading impressions.

授課教師答不出第一條,是一個有用的結果。答不出,說明這一節沒有設定學生要走到哪一步,或者那一步在教案裡寫得不清楚;紀錄的第一欄於是大半空白。這個空白本身要寫進第五欄:下一節先把那一步定下來,再安排觀課。A teacher who cannot answer the first question has produced a useful result. It means the lesson set no step for the students, or the step was not clear in the plan, and the first field of the record stays mostly empty. That emptiness belongs in the fifth field: fix the step first, then arrange the observation.

這些紀錄之後答得了甚麼問題What the records answer later

三個。一個來自教育局,一個來自校外評核隊伍,一個來自學校自己。Three, and each comes from a different reader: the Bureau, an external review team, and the school itself.

誰來問Who asks 問甚麼What they ask 紀錄答得了甚麼What the record answers
教育局的中期報告The Bureau's mid-term report 課堂紀錄證明得到進度嗎?Do the classroom records show progress? 每節課的日期、任務、學生作品,以及那一條改動[1]。The date, the task, the student's work and the one change, lesson by lesson[1].
校外評核隊伍An external review team 學校的觀課紀錄看得出學與教嗎?Do the school's observation records show learning and teaching? 學生表現的句子與作品樣本,而一份成績分佈表答不了這條問題[9]。Sentences about what students did, and samples of their work, where a distribution of scores cannot answer at all[9].
學校續約的決定The school's renewal decision 第二年還買不買這門課。Is this course worth a second year? 六節課累積下來的模式:哪一類任務學生做得到,哪一類離開工具就做不到[11]。A pattern built from six lessons, showing which tasks students finish and which they cannot finish without the tool[11].

一年的紀錄如果只有分數,第二年的決定只能靠印象。五欄紀錄累積六節之後,學校看得到哪一類任務學生做得到,哪一類任務離開工具就做不到。這個分別與家長那個一個月測試是同一個問題的兩面:學校看課堂,家長看孩子。A year of scores leaves next year's decision resting on impressions. Six five-field records show which tasks students can finish and which they cannot finish once the tool is gone. That distinction is the same question the month-later test puts to a parent, seen from the other side: the school watches the lesson and the parent watches the child.

「智啟學教」的中期報告要用課堂紀錄證明進度,兩份表格的每一格另有一頁逐格拆開,見中期報告的四個部分;學校挑選供應商的時候,報價邀請要問的四條問題也有一頁說明,見評估一份報價。兩件事最後都要讀一份課堂紀錄,而這份紀錄的格式可以現在就定。The mid-term report of the AI-in-education grant proves progress with classroom records, and every field of the two forms is taken apart on the page about the report, and a school choosing a supplier has four questions for the quotation, set out on the page about evaluation. Both come down to reading a classroom record, and the shape of that record can be settled now.

學年完結的時候,同一批紀錄還有第三個用途:學校報告裡的評估一節。報告要寫的是這一年做了甚麼、成效怎樣;五欄紀錄給的是日期、任務與作品,比起一段感想更容易寫成一句可以核對的話。At the end of the year the same records serve a third purpose, in the evaluation section of the school report. That section has to say what the year did and what came of it; the five fields supply dates, tasks and work, which is easier to turn into a sentence someone can check than a paragraph of reflection.

外評之外,重點視學也會讀同一批文件:學校發展計劃、周年計劃、學校報告,以及指定科目的教學進度表與學生課業[9]。五欄紀錄與學生作品放在一起,正好是視學隊伍會翻開的那一疊。Alongside an external review, a focus inspection reads the same file: the development plan, the annual plan, the school report, and the schemes of work and student work for the subjects it names[9]. The five-field pages placed with the student work make up the stack an inspection team opens.

這張表證明不了的事What the form cannot show

一節課只是一個樣本。40 分鐘是一學期幾十節課裡的一節;同一個班看兩節,比兩個班各看一節有用。學校如果一學期只觀一節,五欄紀錄記下的是那一節,其餘幾十節仍然沒有紀錄。One lesson is a sample. Forty minutes is one period out of several dozen in a term, and two lessons with the same class tell an observer more than one lesson with each of two classes. A school that observes once a term records that one lesson, and the other dozen carry no record at all.

觀課老師的偏好會留在紀錄裡。喜歡那位老師的觀課老師,寫下的句子會偏向正面;所以第二欄要求逐字記錄,讓偏好留在句子裡,看得見。同一份紀錄由兩個老師分別填,兩句原話放在一起,偏見就會現形。The observer's preference leaves a trace in the record. An observer who likes the teacher writes warmer sentences, which is why the second field asks for the words themselves: the preference stays inside the sentence, where it can be seen. Have two colleagues fill the same record and put the two sentences side by side, and the bias shows.

如果這張表變成評核,教師會為表格準備一節課。五欄紀錄一旦與考績掛鈎,最整齊的一節課會出現在觀課那天。兩份文件要分開存放,這一點比表格的設計更重要。A form used for appraisal gets a rehearsed lesson. Once the five fields count towards a teacher's review, the tidiest lesson of the term arrives on the day of the observation. Keep the two documents apart, which matters more than the design of the page.

它量不到一個月後的保留。同一個任務隔一個月再做一次,沒有筆記、沒有工具,才答得到「還記得幾多」;家長版的一個月測試用的就是這個方法。It cannot measure what survives a month. Set the same task again four weeks later, with no notes and no tool, and the record answers how much stayed; the month-later test for parents uses the same method.

兩個人讀同一句話,可以得出兩個結論,這是一件好事。紀錄留住原話,分歧就有材料可以對照:同一個學生在兩個任務裡講過的話放在一起,觀課老師與授課教師的判斷就有了共同的依據。分歧寫進第五欄,下一節就是一次實驗。Two people can read the same sentence and reach two conclusions, which is useful. Because the record keeps the words, a disagreement has material to check against: put what one student said in two different tasks side by side and the observer and the teacher have shared ground. Write the disagreement into the fifth field and the next lesson becomes the test.

紀錄的價值在累積,不在單張。六節課之後,模式才出現;一節課的紀錄只是一句話。A record earns its value by accumulating. Six lessons produce a pattern, and one lesson produces a sentence.

示範紀錄取自我們的交付紀錄:一節課,約 30 名學生,2026-11-13 由團隊答覆並由創辦人確認數字準確。The sample record comes from our delivery records: one lesson, about thirty students, answered by the team and confirmed by the founder on 13 November 2026.

我們交出的是課堂紀錄What we hand over is the classroom record

我們做的一邊:每節課前交任務紙,寫明哪一步由學生自己完成;課後交學生的作品樣本與提示詞,讓學校的觀課老師填得滿五欄。學校買到的不只是一節課,還有一份可以放進報告的紀錄。模組、科目與學習階段的對應,寫在到校 AI 課程那一頁。Our side of it: before each lesson we hand over the task sheet, which states the step the student takes alone, and afterwards we hand over samples of student work and the prompts, so the school's observer can fill all five fields. What a school buys is a course and a record that can go into a report. The modules, their subjects and their learning stages are set out on the school course page.

我們不做的一邊:我們不坐在觀課的位置。學校的觀課由學校的人做,導師在場只負責教;一份由我們填的紀錄,證明的是我們的課。The side we stay out of: we do not sit in the observer's chair. The school's observation belongs to the school's own staff, the instructor is there to teach, and a record we fill in would prove a point about our lesson.

我們的缺口:五欄格式未套用到每一節課。現時的紀錄以課程為單位,作品樣本按班存放,未有一頁一節的版本。這是我們要補的,補完之後這一頁會更新。Our gap: the five fields are not yet used for every lesson. The record we keep is organised by course, the work samples sit with each class, and there is no one-page-per-lesson version. That is what we have to build, and this page will change when it exists.

我們不會做的一件事:給學校一份填好的紀錄去抄。填好的紀錄證明的是我們的課;學校要證明的是自己的課,兩份紀錄要由兩個人在兩個位置填。One thing we will not do is hand a school a filled record to copy. A filled record proves a point about our lesson, the school has to prove a point about its own, and the two records have to be filled by two people in two places.

常見問題Questions

觀課表要打分嗎?Should the form still carry a score?
可以同時做兩件事,兩件事要各有各的位置。分數回答「這位老師教得怎樣」,五欄紀錄回答「這班學生做到了甚麼」;同一頁上分兩欄,或者分兩份文件,兩邊都不會互相污染。如果只有一格,學校會用那一格回答兩個問題,兩個答案都不準。It can do both jobs, as long as each job has its own place. A score answers how the teacher taught; the five fields answer what the students did. Two parts of one page, or two documents, keep the two answers clean. With a single box, a school uses it for both questions and both answers go wrong.
一張表夠不夠?Is one form enough?
不夠,一節課只是一個樣本。同一個班一學期看兩節,第二節要看的正是第一節那條決定有沒有做到;兩節合起來才有比較。兩節課之間隔三、四星期,學生忘記了的部分就看得出來。No, because one lesson is a sample. Two lessons with the same class in a term are the minimum, and the second one checks whether the decision from the first was carried out; the pair gives the comparison. Space them three or four weeks apart and what the students have forgotten becomes visible.
誰來觀課?Who observes the lesson?
學校的人,可以是科主任、副校長或者同科同事。觀課前要拿到任務紙,知道哪一步由學生自己完成;外聘導師在場只負責教,不填這份紀錄。同一位老師的課,最好由同一批觀課老師看兩次,句子的寫法才會穩定。Someone from the school: a panel head, a vice-principal or a colleague in the same subject. They need the task sheet before the lesson so they know which step belongs to the student. An outside instructor teaches and does not fill the record, and the same observer or pair of observers seeing the same teacher twice keeps the sentences consistent.
觀課紀錄要保存多久?How long should the record be kept?
通函第 221/2025 號要求財政紀錄與單據保留不少於七年,課堂紀錄不在這一句之內[1]。課堂紀錄屬於學校的學與教檔案,按校本政策保存;實際的做法是保存一個學校發展周期,讓外評隊伍與下一任科主任都讀得到上一輪的紀錄。Circular Memorandum No. 221/2025 requires financial records and receipts to be kept for at least seven years, and classroom records fall outside that sentence[1]. They belong to the school's learning and teaching file and follow school policy; in practice a school keeps one development cycle, so an external review team and the next panel head can read the previous round.
外聘導師的課怎樣觀?How do you observe an outside instructor's lesson?
與校內教師的課一樣觀,另加一項:課前要拿到任務紙與課堂計劃,確認哪一步由學生自己完成。外聘導師的課最容易出現導師示範、學生跟着做;第一欄就是為了分出這一步,紀錄上寫不出來,就說明這一課學生沒有自己動手的部分。The same way, with one addition: ask for the task sheet and the lesson plan beforehand and confirm which step the student takes alone. An outside instructor's lesson drifts easily into a demonstration that students copy, and the first field exists to separate that step; if the observer cannot write it down, the lesson has no part the students did themselves.
AI 課一定要關掉工具才觀得到嗎?Does the tool have to be switched off to see anything?
不用。要關掉工具的只是其中一個任務,一個單元裡有一節就夠;觀察的是學生在沒有工具時做得到甚麼,然後與有工具時的版本比較[2]。指南列明部分課業只適宜在課堂及教師監督下完成,這一條給學校一個現成的理由。No. One task is enough, once in a unit, and what it shows is what a student can do without the tool, which is then compared with the version made with it[2]. The guide notes that some work is better done in class under supervision, which gives a school a ready reason to set the task that way.
這些紀錄與「智啟學教」報告有甚麼關係?How do these records relate to the grant report?
中期報告與總結報告要的是進度與數字,五欄紀錄是那些數字的底稿:每節課的日期、任務、學生作品與改動,加起來就是報告裡的教學例子與學生活動[1]。報告的兩份表格另一頁逐格拆開;那些格子背後要有的底稿,就是五欄紀錄。The mid-term and final reports ask for progress and counts, and the five-field records are the working papers behind those counts: the date, the task, the student's work and the change, lesson by lesson, add up to the teaching examples and student activities the form asks for[1]. The two forms are taken apart on another page, and the working papers behind their boxes are the five fields.

下一步Where to go next

帶來你現時用的觀課表Bring the form you use now 告訴我們那一頁上的欄目,我們用其中一節到校課填一次五欄給你看,讓學校在買之前先看見一份紀錄是甚麼樣子。Tell us what is on the page, and we will fill the five fields once from one of our school lessons so that you can see what a record looks like before buying anything. 到校 AI 課程與 19 個模組School courses and the 19 modules 每個模組屬於哪個科目與學習階段、對應哪一項承諾,與收費界線,都寫在頁面上。Which subject and stage each module belongs to, which commitment it evidences, and where the fees stop, all on the page.

相關文章Related articles

中期報告要用課堂紀錄證明進度The mid-term report proves progress with classroom records 兩份表格的四個部分、兩個方格與一個簽名,以及每一格背後要有的文件。Four parts of two forms, two boxes and one signature, and the documents each box needs behind it. 成果要看一個月後還做得到甚麼What a course taught shows a month later 家長版的七項條件與一個 20 分鐘的家用測試,問的同一個問題。Seven conditions for a parent and a twenty-minute test at home, asking the same question.

資料來源Sources

  1. 教育局通函第 156/2026 號《2026/27 學年校本數字教育整體規劃及相關教師培訓》,2026 年 8 月 25 日發出。全份文件共 14 頁,2026-11-20 核對。文中引用:第 2 段(學校須於 2026/27 學年把數字教育納入學校發展計劃及/或學校周年計劃)、第 3 節(鼓勵成立「數字教育教師小組」,推動同儕觀課、經驗分享與專業交流,第 5 頁)、第 10 段(每三年 150 小時周期內不少於 30 小時數字教育培訓,第 3 頁)、附錄的進階單元(第 5 項評估革新與學習數據分析、第 8 項以實證為本策劃推行檢視、第 9 項構思執行及檢討一項融入人工智能的課堂,第 13 頁)。Education Bureau Circular Memorandum No. 156/2026, school-based planning for digital education and the related teacher training in 2026/27, 25 August 2026, and runs to 14 pages; checked on 20 November 2026. Used here: paragraph 2 on putting digital education into the school development plan or the annual plan in 2026/27; section 3, page 5, which encourages a teacher group that promotes peer lesson observation, sharing and professional exchange; paragraph 10, page 3, on the 30 hours of digital education training inside each three-year cycle of 150 hours; and the advanced units in the annex, page 13, including unit 5 on assessment and learning data, unit 8 on planning, delivering and reviewing by evidence, and unit 9 on planning, teaching and reviewing one AI-integrated lesson. edb.gov.hk (PDF)
  2. 教育局《中小學數字教育發展藍圖》,2026 年 6 月,附篇二《中小學應用人工智能教學指南》,「評(評估)」一節在第 2-8 頁,2026-11-20 逐頁核對。文中引用:評估由學習成果走向學習過程的三種取向;教師觀察列為評估數據之一;「學生不應直接提交由人工智能生成的課業」;「任何情況下,教師不應未經專業判斷便直接採用人工智能給予學生的評分和回饋」;以即時口頭查問、要求提交思考歷程、部分課業在課堂及教師監督下完成三種方式查核;學校為 AI 在課業與評估中的應用訂立準則。Education Bureau, the Blueprint for Digital Education in Primary and Secondary Schools, June 2026, and its Supplement II, the Guide to Using Artificial Intelligence in Teaching in Primary and Secondary Schools, whose assessment section begins on page 2-8 and was checked page by page on 20 November 2026. Used here: the move from assessment of learning towards assessment for and as learning; a teacher's observation counted as one kind of assessment data; the statement that students should not submit AI-generated work directly; the statement that under no circumstances should a teacher adopt an AI score or comment without professional judgement; the three ways of checking, namely a question asked in class, a record of the student's thinking, and work completed in class under supervision; and the call for schools to set rules for AI in coursework and assessment. edb.gov.hk (PDF)
  3. 教育局《校本資優教育發展計劃指引》附件十二〈同儕觀課評估表〉。全表一頁,2026-11-20 核對。文中引用:18 項以 1 至 5 分評級(A 教學目標 2 項、B 教學活動 4 項、C 教學程序 4 項、D 課後回應 8 項)、一個建議欄與一行要求列明原因、題目「本課節能達到既定的教學目標」與「教學活動能提高學生的學習興趣」。Education Bureau, guidelines on school-based gifted development programmes, annex 12, a peer observation form of one page, checked on 20 November 2026. Used here: eighteen items scored 1 to 5, being two on objectives, four on activities, four on procedure and eight on the observer's response; one suggestions box and one line asking for a reason; and the wording of two items, that the lesson achieves the stated objectives and that the activities raise students' interest. edb.gov.hk (PDF)
  4. 賽馬會「校本多元」計劃學習支援組〈觀課記錄表〉,載於教育局特別教育資訊網專業發展資料庫。全份共 14 頁,2026-11-20 核對。文中引用:教師因素 21 項、學生因素 7 項、課程及資源因素 9 項、實地環境因素 5 項,全部以 1 至 5 分評級並設不適用與意見欄;題目「學生積極參與」與「學生運用後設認知能力解決問題」。Jockey Club "Diversity at Schools" project, learning support team, lesson observation record, held in the Bureau's special education professional development archive; the document runs to 14 pages and was checked on 20 November 2026. Used here: 21 teacher items, seven student items, nine curriculum and resource items and five environment items, all scored 1 to 5 with a not-applicable option and a remarks column; and the wording of two items, that students take part actively and that students apply metacognitive ability to solve problems. sense.edb.gov.hk (PDF)
  5. 聖士提反女子中學附屬小學《2017-2018 年度學校報告》附件十三〈同儕觀課紀錄表〉,第 85 至 90 頁,2026-11-20 核對。文中引用:學生學習 7 項與教師專業 5 項,各以 1 至 5 分評級;觀課重點五組(預習工作、學習策略、延伸活動、電子學習、其他),每組設優點與建議欄;電子學習一組的三條「能/不能」;末頁的授課教師自評與兩個簽署欄;題目「批判性思考能力」與「自學能力」。St Stephen's Girls' Primary School, 2017-18 annual school report, annex 13, a peer observation record on pages 85 to 90, checked on 20 November 2026. Used here: seven student items and five teacher items, each scored 1 to 5; five observation foci covering preparation, learning strategies, extension activities, e-learning and anything else, each with a strengths box and a suggestions box; the three yes-or-no questions in the e-learning focus; the teacher's self-evaluation page and the two signature lines; and the wording of two items, critical thinking and self-directed learning. ssgps.edu.hk (PDF)
  6. 高主教書院《意德省察先導計劃 2024-2025》分享簡報內的〈課堂觀察評估表〉,2026-11-20 核對。文中引用:八個評估項目(課堂氛圍、教師引導、時間安排、省察深度、反思分享、資源運用、目標達成、整體滿意度),以 1 至 5 分表示滿意程度;公開樣本沒有文字欄。Raimondi College, a classroom observation form reproduced in the sharing deck for its 2024-25 pilot scheme on reflective practice, checked on 20 November 2026. Used here: eight items covering atmosphere, teacher guidance, time, depth of reflection, sharing of reflection, use of resources, attainment of objectives and overall satisfaction, each rated 1 to 5, and a published sample with no text field. raimondi.edu.hk (PDF)
  7. 甘迺迪中心〈觀課紀錄表〉,載於該校校本自評工具。全表一頁,2026-11-20 核對。文中引用:三個部分(課堂目標、目標實踐效益、學生表現),以勾選描述代替評分;一欄「略述情況」與一格「其他意見/備註」;「學生表現」一組的描述選項包括「投入學習、表現積極」與「態度散漫、冷漠無聊」。Kennedy Centre, observation record, one page of the school's self-evaluation tools, checked on 20 November 2026. Used here: three parts covering objectives, the effect of those objectives and student performance, with ticked descriptions instead of scores; one line for a brief description and one box for other comments; and the paired descriptions in the student section, which run from engaged and positive to listless and indifferent. web.jfk.edu.hk (PDF)
  8. 香港中文大學香港學校改進及伙伴協作中心(QSIP)《「拓展選擇」項目 科組領導手冊 單元三 推動觀課議課》。全份共 6 頁,2026-11-20 核對。文中引用:「紀錄的要點」三條(客觀地紀錄事實;記錄學生回應教師問題的頻次、對答情況、參與程度;拍攝學生的課堂表現與習作樣本帶回議課),以及議課問題「有多少學生能掌握課堂的重點?如何知悉?」。Chinese University of Hong Kong, Centre for School Improvement and Partnership (QSIP), subject-leadership handbook, unit 3, on running observation and the discussion after it; the unit runs to six pages and was checked on 20 November 2026. Used here: the three points for the record, being to record facts objectively, to record how often students answer and what they say, and to photograph student performance and work samples for the discussion; and the discussion question asking how many students grasped the point and how the observer knows. qsip.cuhk.edu.hk (PDF)
  9. 教育局〈學校發展與問責架構及學校自我評估研討會:校外評核與重點視學的重點〉簡報,2020 年,2026-11-20 核對。文中引用:「外部期間的安排」一節列明校外評核隊伍會閱覽學校文件及學生課業、觀察學校活動及觀課、進行小組會議與個別面談。Education Bureau, seminar on the school development and accountability framework, key points on external school review and focus inspection, a deck dated 2020 and checked on 20 November 2026. Used here: the section on what happens during an external review, which lists reading school documents and student work, observing school activities and lessons, and holding group and individual meetings. edb.gov.hk (PDF)
  10. 陳瑞祺(喇沙)書院《2018-2019 周年校務報告》,同儕互助觀課計劃一節,2026-11-20 核對。文中引用:觀課前共同備課、觀課期間不使用量表、觀課後議課,三部分寫入同一份觀課紀錄表。Chan Sui Ki (La Salle) College, 2018-19 annual school report, the section on the peer observation scheme, checked on 20 November 2026. Used here: joint preparation before the lesson, observation without a rating scale, and a discussion afterwards, with all three written into one observation record. cghc.edu.hk (PDF)
  11. Edcosys 交付紀錄,2026-11-13 由團隊答覆並由創辦人確認數字準確:一個約三十名學生、六節各 90 分鐘的中學數字人課程;學生自行撰寫嚴格的 DSE 口試考官提示詞,有人練習約二十分鐘;早期約三分之一連線嘗試失敗,改用分段排隊、單向耳機與前半節文字邏輯後失敗很少。紀錄沒有學校名稱,學生引述另有一份 2026-10-16 的匿名書面同意,未經第三方審計。Edcosys delivery records, answered by the team and confirmed by the founder on 13 November 2026: one secondary-school digital human course of about thirty students across six 90-minute sessions, where students wrote a strict DSE oral examiner prompt and some practised for about twenty minutes, and where about a third of connection attempts failed early on before staggered queuing, one-way headsets and a text-only first half made failures rare. The record carries no school name, the student quotation rests on a separate anonymous written consent dated 16 October 2026, and nothing here has been audited by a third party.
← 返回資源中心← Back to the Resource Centre