Recommended AI Toolstack for Teaching

核心觀念

• 教學即服務 (TaaS)：將備課、授課、評量等環節模組化，利用 AI 工具將重複性高的部分自動化，讓老師能更專注於與學生的互動和課程設計。
• 素材先行，遊戲後製：先準備好核心的教學素材（PDF、文字檔），再利用 AI 工具（如 Padlet Arcade）快速生成多種互動遊戲，一魚多吃。
• 駕駛艙整合：將簡報、互動工具（抽籤、計時器）、評量、線上資源全部整合到一個「教學駕駛艙」網頁中，避免在多個視窗間切換，保持教學流暢性。
• 視覺化輔助：善用 AI 生圖功能（Claude + Image 2），為考卷、簡報、學習單快速生成情境插圖，提升教材的視覺吸引力和學生的理解力。

整合工作流

• 研究與萃取：使用 Google Gemini 的 Deep Research 收集主題資料，再用 NotebookLM 匯入教材與報告，萃取出核心教學重點。
• 教材視覺化：
• 指示 Claude + Image 2 根據教學內容生成簡報插圖、情境圖片。
• 指示 NotebookLM 或 Gemini 將文字大綱生成資訊圖表。
• 評語系統建立：在 Google 試算表建立評語代碼，利用 NotebookLM 結合自訂 YAML 提示詞，建立自動生成學習評語的系統。

2. 授課階段：互動式教學體驗

• 建立教學駕駛艙：
• 在 Claude Code 中，指示 AI 讀取所有備課素材（簡報圖片、資訊圖表）。
• 要求 AI 生成一個包含簡報輪播、互動工具（抽籤、計時器）、評量連結的單頁式 HTML「教學駕駛艙」。
• 將此網頁部署到 GitHub Pages，方便課堂上直接使用。
• 即時互動遊戲：
• 將備課的 PDF 教材上傳至 Padlet。
• 選擇「Arcade」功能，讓 Padlet 自動將教材內容轉換成記憶卡、排序、填空等多種互動遊戲。
• 在課堂上透過 QR Code 或連結讓學生直接遊玩，並顯示排行榜。

3. 評量與反饋階段：AI 輔助評量

• 生成考卷：在 Claude 中，結合 Image 2 生成帶有情境插圖的 Word 考卷。
• 自動批改與分析：將學生的作業（如作文）匯入 NotebookLM，設定好評分量表作為來源，讓 AI 根據標準進行初步評分與回饋。
• 影片測驗：將教學影片（可由 NotebookLM 簡報+AI配音生成）上傳至支援字幕的 AI 測驗平台，自動生成互動式測驗題目。

最佳金句

「AI時代想要讓你的教學有效率，你就要學會使用這些工具啊。」 「我把這個界面叫做教學駕駛艙。」 「它是根據我上傳的素材，去自動生成這些題型。」

教學切入建議

常見誤區

• 工具碎片化：學了很多單點的 AI 工具，但無法將它們整合成一個順暢的工作流，導致備課時仍在不同工具間反覆切換。
• 為互動而互動：過度使用花俏的互動遊戲，卻忽略了遊戲是否服務於核心的教學目標。
• 忽略學生體驗：設計的互動環節需要學生安裝 App、註冊帳號，或介面複雜，導致技術問題蓋過學習本身。
• 完全信任 AI 評分：將 AI 的評分或評語結果未經審核直接發給學生，可能產生不公平或不恰當的回饋。

---en---

Core Concepts

• Teaching as a Service (TaaS): Modularize stages like lesson preparation, instruction, and assessment; automate repetitive tasks using AI tools, allowing teachers to focus more on student interaction and curriculum design.
• Content First, Game Post-Production: Prepare core teaching materials (PDFs, text files) first, then use AI tools (e.g., Padlet Arcade) to quickly generate various interactive games, maximizing content utility.
• Cockpit Integration: Integrate presentations, interactive tools (random selector, timer), assessments, and online resources into a single “Teaching Cockpit” webpage, avoiding switching between multiple windows and maintaining teaching flow.
• Visual Aid: Leverage AI image generation (Claude + Image 2) to quickly create situational illustrations for exams, presentations, and worksheets, enhancing the visual appeal of materials and student comprehension.

Integrated Workflow

• Research and Extraction: Use Google Gemini’s Deep Research to collect topic-related data, then import textbooks and reports into NotebookLM to extract core teaching points.
• Material Visualization:
• Instruct Claude + Image 2 to generate presentation illustrations and situational images based on teaching content.
• Instruct NotebookLM or Gemini to generate infographics from text outlines.
• Feedback System Establishment: Create a feedback code in Google Sheets, then use NotebookLM combined with custom YAML prompts to build a system for automatically generating learning feedback.

2. Instruction Phase: Interactive Teaching Experience

• Build a Teaching Cockpit:
• In Claude Code, instruct AI to read all preparatory materials (presentation images, infographics).
• Request AI to generate a single-page HTML “Teaching Cockpit” including a presentation carousel, interactive tools (random selector, timer), and assessment links.
• Deploy this webpage to GitHub Pages for direct use in class.
• Real-time Interactive Games:
• Upload prepared PDF teaching materials to Padlet.
• Select the “Arcade” function to have Padlet automatically convert the materials into various interactive games like memory cards, sequencing, and fill-in-the-blanks.
• In class, students can play directly via QR Code or link, and a leaderboard can be displayed.

3. Assessment and Feedback Phase: AI-Assisted Assessment

• Generate Exam Papers: In Claude, combine with Image 2 to generate Word exam papers with situational illustrations.
• Automated Grading and Analysis: Import student work (e.g., essays) into NotebookLM, set up a grading rubric as a source, and let AI conduct preliminary grading and feedback based on the standards.
• Video Quizzes: Upload teaching videos (which can be generated from NotebookLM presentations + AI voiceover) to an AI quiz platform that supports subtitles, to automatically generate interactive quiz questions.

Best Quotes

“In the AI era, to make your teaching efficient, you must learn to use these tools.” “I call this interface the ‘Teaching Cockpit’.” “It automatically generates these question types based on the materials I upload.”

Teaching Entry Points

Common Pitfalls

• Tool Fragmentation: Learning many individual AI tools but failing to integrate them into a smooth workflow, leading to continuous switching between tools during lesson preparation.
• Interaction for Interaction’s Sake: Overusing flashy interactive games, neglecting whether the games serve the core teaching objectives.
• Ignoring Student Experience: Designing interactive elements that require students to install apps, register accounts, or have complex interfaces, causing technical issues to overshadow learning itself.
• Complete Trust in AI Grading: Directly sending AI’s grading or feedback results to students without human review, potentially leading to unfair or inappropriate feedback.