Gemini 賦能：打造有溫度的「班級小幫手」LINE 機器人 — 從導師焦慮到親師共好
講者/Speaker： 黃唯芯, Wei-Shin Huang
核心目標與系統
1.利用 Generative AI (Gemini) 實現「零程式碼」開發模式，讓不具備程式背景的教師能開發客製化的 LINE 機器人，以自動化處理催繳作業、出缺席登記等行政工作，減少導師焦慮並改善親師溝通的冰冷感。
2.系統整合了 Google 試算表（後台）與 LINE 官方帳號（前台）。
創新與特色
1.溫暖防呆機制：獨創「請假保護傘」。系統偵測到病假、事假等關鍵字時，會自動遮蔽當日缺交警告，改發溫馨關懷文案。
2.視覺化回饋：串接 QuickChart API，將段考成績數據自動轉化為「個人能力雷達圖」推播。
3.高容錯率：AI 優化程式邏輯，能自動適應教師不同的紀錄習慣（如將「缺」、「x」視為缺交）。
4.開發模式：教師擔任「產品經理」，透過「先討論邏輯」、「貼截圖除錯」、「描述日常習慣」等自然語言協作策略，引導 Gemini 撰寫 Google Apps Script 程式碼。
具體成效
1.量化：預計每週為導師節省約5-10小時的行政時間。整合 5 大類數據（出缺席、獎懲等），家長查詢效率提升 100%。
2.質化：成功降低教師科技焦慮，並透過人性化功能（如「請假保護傘」）有效提升親師信任感。
Core Objectives & System
1. Utilize Generative AI (Gemini) to realize a "zero-code" development model, enabling teachers without programming backgrounds to develop customized LINE bots. This automates administrative tasks such as chasing late assignments and recording attendance, thereby reducing homeroom teachers' anxiety and softening the impersonal nature of parent-teacher communication.
2. The system integrates Google Sheets (as the backend database) with LINE Official Accounts (as the frontend interface).
Innovations & Features
1. Empathetic Fail-Safe Mechanism: Features the original "Leave Protection Umbrella." When the system detects keywords indicating sick or personal leave, it automatically suppresses missing assignment warnings for that day and replaces them with warm, caring messages.
2. Visualized Feedback: Integrates the QuickChart API to automatically transform exam score data into a "Personal Ability Radar Chart" that is pushed directly to users.
3. High Fault Tolerance: AI optimizes the programming logic to automatically adapt to different teachers' data entry habits (for example, recognizing both "缺" [missing] and "x" as unsubmitted assignments).
4. Development Model: Teachers act as "Product Managers." They use natural language collaboration strategies—such as "discussing the logic first," "pasting screenshots for debugging," and "describing daily routines"—to guide Gemini in writing the Google Apps Script code.
Concrete Outcomes
1. Quantitative Impacts: Expected to save homeroom teachers approximately 5 to 10 hours of administrative work per week. By integrating 5 major data categories (including attendance, rewards/penalties, etc.), the efficiency of parent queries has increased by 100%.
2. Qualitative Impacts: Successfully lowers teachers' technology anxiety and effectively enhances parent-teacher trust through human-centric features like the "Leave Protection Umbrella."