本研究旨在設計並初步導入家庭AI聊書助理，探討其在親子聊書歷程回顧、AI提問建議及家校閱讀資料轉譯中的應用情形。既有學校閱讀活動與師生聊書資料雖逐漸累積，家庭中的親子聊書互動仍較少被具體保存，學校端資料也未必能直接轉化為家庭可接續的對話資源。因此，本研究整合親子聊書錄音、語音轉錄、角色辨識、摘要、CROWD提問方向分析、AI建議問題、歷程回顧與家校資料整合等功能，協助家長保存、理解並延伸親子聊書。 本研究採設計研究法（design-based research， DBR），以質性資料分析為主，問卷、系統操作紀錄與AI輸出檢核為輔。研究包含兩次家庭端系統導入與一次教師觀點訪談，資料來源包括家長前導背景問卷、親子聊書逐字稿、系統紀錄、家長訪談、家校整合內容與教師觀點訪談。研究者先依家庭與研究階段整理使用脈絡，再對照逐字稿、系統紀錄、AI輸出與訪談資料，分析家長如何接觸、理解、參考及轉化系統內容，並整理各階段資料如何支持後續設計調整。 研究結果顯示，系統可將家庭中容易流失的口語互動整理為逐字稿、摘要與提問方向，提供家長回顧互動及尋找後續聊書線索的入口；家長亦不將AI提問建議視為固定腳本，而會依孩子回答、家庭語氣與文本內容進行篩選、縮短、改寫或暫不使用，因此AI較適合作為可選擇與可調整的對話支持。學校端閱讀與聊書資料則需經摘要化、觀察化、脈絡化與行動化整理，才較容易被家長理解，並形成家庭聊天、選書或後續聊書的參考。教師訪談亦補充下一輪家校聊書共備所需的資料內容、角色分工與推動條件；由於本研究未正式實施共備活動，故不推論其成效。本研究指出，家庭閱讀AI工具應協助家長理解、判斷與接續親子互動，而非取代家長或教師；未來研究可延長導入期程、增加參與家庭並納入孩子觀點，以進一步檢視其長期使用條件與實務價值。;This study designed and preliminarily implemented a family AI book-talk assistant to examine its use in reviewing parent–child book-talk， providing AI-generated questions， and translating home–school reading data. Because parent–child book-talk is rarely preserved and school-based data may not readily become usable family resources， the system integrated recording， transcription， speaker identification， summaries， CROWD-based question analysis， AI suggestions， history review， and home–school data integration. The study adopted design-based research (DBR)， using qualitative analysis as the primary approach and questionnaires， system logs， and AI-output evaluation as supplementary data. It included two family implementation stages and one teacher interview stage. Data included parent questionnaires， parent–child book-talk transcripts， system logs， interviews， and home–school integrated content. Data were organized by family and research stage to examine how parents accessed， understood， referred to， and transformed system-generated content. The findings showed that the system transformed easily lost oral interactions into transcripts， summaries， and question-direction records that supported review and later planning. Parents did not treat AI-generated questions as fixed scripts， but selected， shortened， revised， or set them aside according to their children’s responses and conversational styles. AI suggestions were therefore positioned as optional， adjustable conversational support. School-based reading and book-talk data required summarization， observation， contextualization， and action-oriented translation before becoming useful references for family conversation， book selection， or later book-talk. Teacher interviews identified the data content， role distribution， and implementation conditions needed for future home–school book-talk co-planning; because such activities were not formally implemented， no effects were inferred. The study suggests that AI tools for family reading should support parental understanding， judgment， and continued parent–child interaction rather than replace parents or teachers. Future research may extend implementation， involve more families， and include children’s perspectives.
