孩子用 AI 做功課,家長應該禁止,還是教他正確使用?
孩子用人工智能做功課,不必一律禁止,也不宜毫無限制地使用。先遵守學校對該項課業的規定,確保思考過程仍然屬於孩子自己。家長不要只看功課是否流暢,更應問:人工智能幫了什麼、孩子檢查或改過什麼、孩子自己學會了什麼。
許多家長看見孩子的功課突然寫得很流暢,第一個念頭往往是禁止使用人工智能(AI)。家長不必一律禁止,也不宜放手讓孩子直接套用答案。應先遵守學校對該項課業的規定,並確保思考過程仍然屬於孩子自己。
如果學校容許使用,家長也不宜只看最後一篇功課是否流暢,而應看孩子有沒有自己想過、檢查過,以及學會了什麼。
先分清:幫忙思考,還是代替思考
人工智能可以幫忙,也可以代替。分別在於孩子有沒有把自己的思考交出去。
- 代替思考: 孩子把題目直接貼上,要求「寫一篇作文」或「告訴我答案」,再把生成的內容幾乎原封不動交出去。這樣很快就能交功課,但家長和老師很難判斷孩子到底學會了什麼。
- 幫忙思考: 孩子先自己試做,再向人工智能請教思路、例子或檢查方法,然後用自己的話整理,並且核對內容是否正確、是否符合題目要求。工具可以協助,但不能取代孩子的判斷。
同一道功課,兩種用法差很遠。以下是假設例子,不是真實學生習作,也不是任何學校的規定。
假設例子一:英文短文。 孩子對人工智能說「請寫一篇一百字的週末遊記」,再把全文複製到作業簿,這是代替思考。比較好的做法是:孩子先寫大綱或初稿,再問「我這段想寫去公園,請指出意思不清楚的地方,不要直接替我重寫」,然後由孩子自己修改。
假設例子二:數學應用題。 孩子把整題貼上,只問「答案是多少」,這是代替思考。比較好的做法是:孩子先列出已知條件和自己的想法,再問「請教我怎樣想,不要直接給答案」,做完後自己驗算,並能向家長說明每一步為什麼這樣做。
如果孩子離開工具就說不出思路,或者無法解釋功課裡的用字和步驟,就表示工具可能已經在代替他思考。
孩子用 AI 做功課,家長應該看答案,還是看思考過程?
據《香港文匯報》2026年11月13日〈師說新語〉專欄報道,生成式人工智能進入中小學課堂後,若仍以「成品流暢、答案正確」來評分,評估容易失真。該文作者、創知中學校長黃晶榕博士指出,學生使用人工智能做功課,並不等於作弊。關鍵在於學校訂明使用邊界、學生如實申報,評估才能呈現真實的思考。
該專欄舉例指出,觀塘福建中學為專題研習設立「人工智能誠信申報表」,要求學生記錄指令、人工智能產出、調整追問及最終成品,考測則禁用人工智能。台山商會中學則要求提交提示詞列表、訓練過程及反思表,記錄工具的限制與謬誤,評分重心放在思考過程。這些是專欄所述的學校做法,家長仍應以子女學校公布的要求為準。
對家長而言,不要只因功課看來流暢,便以為孩子已經學會。
在家可以怎樣做:交功課前先問三句
交功課前,不必先挑剔整篇文字是否完美。可以請孩子暫時合上螢幕,用自己的話說出以下三句的答案。
- 人工智能幫了什麼? 孩子應說得出自己先做了哪一步,工具後來補充了什麼。如果他只說「它全部寫好了」,就要一起停下來,先讓孩子自己想。
- 你檢查過或改過什麼? 請他指出改過的句子、數字或資料,以及為什麼要改。完全沒有改過,並不自動等於內容正確。
- 你自己學會了什麼? 請他講一個原先不懂、現在能說明的要點。如果講不出來,這次可能只是把功課交出去,還沒學會。
如果學校要求申報,就協助孩子如實記錄,而不是代他填寫,也不要代孩子隱瞞。一個穩妥的次序是:先自己試做,再按學校規則決定是否使用工具,最後用自己的話說明過程。
使用前先看年齡和私隱
知涵建議孩子七歲或以上才接觸人工智能。這是學院的年齡建議,並不表示所有工具都容許七歲兒童開設帳戶。不同工具的年齡限制和家長監護設定並不相同。使用前應先查看該工具的規定,並由家長決定是否適合自己的孩子。
無論使用哪一種工具,都不要向它提供孩子的真實姓名、學校、住址、電話或照片,也不要把同學的個人資料貼上去。人工智能生成的內容可能看起來通順,卻並不一定正確,所以查證和修改仍然要由孩子自己做。
如果希望孩子練習怎樣向人工智能提問、檢查答案,並用自己的話說明想法,可以參考知涵兒童財商學院的AI思維與溝通課程(適合7至15歲)。
常見問題
孩子用人工智能做功課,算不算作弊?
要看學校對該項課業的規定。有些課業容許使用但要如實申報,有些考測則禁用。即使被容許使用,把生成內容原封不動交出去,老師仍可能無法判斷孩子是否真正學會。
應不應該在家裡一律禁止孩子使用人工智能?
不必一律禁止,但也不宜毫無限制地使用。先遵守學校規則,再教孩子把工具用來幫忙思考,而不是代替思考。
功課看起來很流暢,是否代表孩子已經學會?
不一定。流暢只說明成品好看,不能證明孩子理解內容。家長更宜看他能否說明思路、指出修改,以及說出自己學會了什麼。
如果學校禁止使用,家長還要不要在家教孩子用?
該項課業就應遵守學校規定。若家長想讓孩子在其他場合認識工具,也要分清家中練習和要交回學校的評核,不要把兩者混在一起。
孩子幾歲可以開始接觸人工智能?
知涵建議七歲或以上才開始接觸。各工具的開戶和年齡限制並不相同,即使孩子已滿七歲,也不等於可以自行開設每一種工具的帳戶。使用前應先查看該工具的規定,並由家長陪伴。
可以讓孩子把個人資料交給人工智能嗎?
不可以。不要提供真實姓名、學校、住址、電話或照片,也不要上傳同學的資料。
參考來源
黃晶榕,〈【師說新語】AI時代校內評估應關注學習歷程〉,《香港文匯報》,2026年11月13日
When children use AI for homework, should parents ban it, or teach them to use it properly?
If a child uses artificial intelligence for homework, there is no need for a blanket ban, nor should use be left completely unrestricted. First follow the school's rules for that assignment, and make sure the thinking still belongs to the child. Parents should not only look at whether the work reads fluently; they should also ask: what did the AI help with, what did the child check or change, and what did the child learn?
Many parents, seeing a child's homework suddenly written very fluently, first think of banning artificial intelligence (AI). Parents do not need a blanket ban, nor should they leave the child free to paste in answers. They should first follow the school's rules for that assignment, and make sure the thinking still belongs to the child.
If the school allows AI to be used, parents should still not look only at whether the finished homework reads fluently. They should look at whether the child has thought it through, checked it, and learned something.
First, tell the difference: helping the child think, or thinking in the child's place
Artificial intelligence can help, and it can also take over. The difference is whether the child has handed over their own thinking.
- Thinking in the child's place: The child pastes in the question, asks it to "write an essay" or "tell me the answer," and then submits the generated content almost unchanged. The homework can be handed in quickly, but parents and teachers will find it hard to tell what the child has actually learned.
- Helping the child think: The child first tries the work themselves, then asks the AI for an approach, examples, or a way to check, then puts it in their own words and verifies whether the content is accurate and meets the requirements of the question. The tool can assist, but it cannot replace the child's judgment.
The same assignment looks very different under these two uses. The following are hypothetical examples. They are not real student work, and they are not the rules of any school.
Hypothetical example 1: a short English composition. The child tells the AI, "Please write a 100-word weekend outing," then copies the whole text into their exercise book. That is thinking in the child's place. A better approach is: the child first writes an outline or a draft, then asks, "I want this paragraph to be about going to the park. Please point out where the meaning is unclear, and do not rewrite it for me," and then makes the changes themselves.
Hypothetical example 2: a maths word problem. The child pastes in the whole question and only asks, "What is the answer?" That is thinking in the child's place. A better approach is: the child first lists the known information and their own ideas, then asks, "Please teach me how to think about this, and do not give me the answer directly," checks the work themselves afterwards, and can explain to a parent why each step was done that way.
If the child cannot explain the reasoning once the tool is closed, or cannot explain the wording and steps in the homework, the tool may already be thinking in their place.
When a child uses AI for homework, should parents look at the answer, or at the thinking process?
According to a 13 November 2026 〈師說新語〉 column in Hong Kong's Wen Wei Po (《香港文匯報》), after generative artificial intelligence entered primary and secondary classrooms, assessment can easily become distorted if work is still graded by whether "the finished product is fluent and the answer is correct." The author, Dr 黃晶榕, principal of 創知中學, points out that a student using AI for homework is not the same as cheating. The key is that the school sets clear boundaries for use, students declare it honestly, and assessment can then show genuine thinking.
The column gives these examples: 觀塘福建中學 has set up an "AI integrity declaration form" for project work, requiring students to record their prompts, the AI output, follow-up questions and adjustments, and the final product, while AI is prohibited in tests and examinations. 台山商會中學 requires students to submit a list of prompts, a record of the working process, and a reflection sheet, noting the tool's limits and errors, with marking focused on the thinking process. These are school practices as described in the column; parents should still follow the requirements published by their child's school.
For parents, the point is not to assume a child has learned something just because the homework looks fluent.
What parents can do at home: ask three questions before the work is handed in
Before the homework is handed in, there is no need to first pick apart whether every sentence is perfect. Parents can ask the child to close the screen for a moment and answer the following three questions in their own words.
- What did the AI help with? The child should be able to say which step they did first, and what the tool later added. If they only say, "It wrote everything," stop together and let the child think first.
- What did you check or change? Ask them to point out the sentences, figures or information they changed, and why. Having changed nothing does not automatically mean the content is correct.
- What did you learn yourself? Ask them to explain one point they did not understand before and can now explain. If they cannot, this time they may only have handed in the homework, without having learned it.
If the school requires a declaration, help the child keep an honest record; do not fill it in for them, and do not conceal the use of AI on their behalf. A sound sequence is: try it yourself first, then decide whether to use the tool according to the school's rules, and finally explain the process in your own words.
Check age and privacy before use
Childpreneur Academy (知涵) recommends that children only start using AI from the age of seven. This is the academy's age recommendation; it does not mean every tool allows a seven-year-old to open an account. Age limits and parental-control settings differ from tool to tool. Check the tool's rules before use, and let the parent decide whether it is suitable for their child.
Whatever the tool, do not give it the child's real name, school, address, phone number or photos, and do not paste in classmates' personal information. AI-generated content may look smooth without being accurate, so checking and revising still have to be done by the child.
If you would like your child to practise asking AI questions, checking answers, and explaining their ideas in their own words, you may refer to Childpreneur Academy's AI Thinking and Communication Course (for ages 7 to 15).
Frequently asked questions
If a child uses AI for homework, does that count as cheating?
It depends on the school's rules for that assignment. Some assignments allow use if it is declared honestly; some tests and examinations prohibit it. Even when use is allowed, submitting generated content unchanged may still leave the teacher unable to tell whether the child has truly learned it.
Should parents impose a blanket ban on AI at home?
There is no need for a blanket ban, but use should not be left completely unrestricted. First follow the school's rules, then teach the child to use the tool to help them think, not to think in their place.
If the homework looks very fluent, does that mean the child has learned it?
Not necessarily. Fluency only shows that the finished product looks good; it does not prove the child understands the content. Parents would do better to see whether the child can explain their reasoning, point out changes, and say what they have learned.
If the school prohibits AI, should parents still teach the child to use it at home?
For that assignment, the school's rules should be followed. If parents want the child to learn about the tool in other settings, they should also keep home practice separate from work that will be submitted for school assessment, and not mix the two.
At what age can a child start using AI?
Childpreneur Academy (知涵) recommends starting only from the age of seven. Account-opening rules and age limits differ from tool to tool. Even if a child is already seven, that does not mean they can open an account on every tool themselves. Check the tool's rules before use, and accompany the child.
Can a child give personal information to AI?
No. Do not provide a real name, school, address, phone number or photos, and do not upload classmates' information.
Sources
黃晶榕, 〈【師說新語】AI時代校內評估應關注學習歷程〉, 《香港文匯報》 (Wen Wei Po), 13 November 2026.