十二年國民基本教育課程綱要（簡稱108課綱）自實施以來，強調核心素養導向，倡導多元發展與適性揚才，然亦因課程結構調整導致部分高中階段之課程時數刪減與難度降低。本研究旨在探討新課綱背景學生進入大學後，於基礎學科的成績表現上是否面臨銜接困難，並探索如何在高等教育場域中延續108課綱所重視的探究學習核心理念。
本研究以國立清華大學一年級學生修習之必選課程「普通物理B」為研究場域，對象為非理學院但具理工背景之學生。課程中導入探究式試題組以及該校由研究團隊結合OpenAI的AI模型所開發的虛擬助教「小TAI」與作為學習輔助工具，兩者皆融入探究導向設計思維，試圖延伸高中教育階段的探究理念至大學教學實踐中。
研究採用混合方法進行：量化部分透過學生成績資料與小TAI系統使用紀錄，分析學習表現與工具使用行為之關聯；質性部分則以問卷調查與學生訪談為主，蒐集其使用經驗與回饋意見。研究涵蓋110至113學年間共六個學期，並聚焦於113學年為主要觀察時段，各學期樣本數皆逾百人。
研究結果顯示，新課綱背景學生在學期成績上呈現逐年下滑趨勢，112學年度平均成績較110學年度下滑逾13分（p < .001，d = 1.22, 1.09）。然於113學年導入探究工具後，觀察到成績止跌回升之初步現象，上、下學期平均分別提升 5.39 分與 4.00 分（p = .001、.015；d = -0.52、-0.35）。其中，小TAI作為個人化AI學習輔助工具，對穩定使用者在段考表現上展現出T分數平均變化量提升（0.89 至 4.40），雖統計效力有限（p > .05），但整體趨勢正向；相較之下，受限於即時回饋不足與學習負擔增加，探究式題組的使用意願偏低。整體而言，AI與探究工具之導入展現強化學習成效與深化探究理念之潛力，顯示其應用於高等教育的發展價值。

The 2019 Curriculum Guidelines for 12-Year Basic Education in Taiwan emphasize core competencies and diversified learning but have also led to reduced instructional hours and content simplification in some high school subjects. This study investigates whether students under the new curriculum face learning difficulties in foundational university courses and explores how inquiry-based learning principles can be extended into higher education.
This research focuses on a first-year university physics course (“General Physics B”) for non-science-major students with a background in science and engineering. The course integrated two inquiry-oriented tools: (1) the inquiry-based problem sets and (2) “TAI”, an AI virtual assistant developed based on OpenAI's model. Both tools aim to support learning through personalized interaction and structured inquiry.
Using a mixed-methods approach, the study analyzed student performance data and AI usage records, along with feedback collected through questionnaires and interviews. The study spanned six semesters (academic years 110 to 113), with a focus on the 113 academic year, each semester involving over 100 student participants.
Results showed a steady decline in semester performance among new curriculum students, with Year 112 scores dropping over 13 points from Year 110 (p < .001 ; d = 1.22, 1.09). After introducing the inquiry tools in Year 113, scores showed a modest rebound, with average gains of 5.39 and 4.00 points in the fall and spring semesters, respectively (p = .001, .015; d = -0.52, -0.35). TAI showed potential benefits for consistent users, with improvements in T-score ranging from 0.89 to 4.40 (p > .05). In contrast, use of the problem sets was limited due to lack of feedback and added workload. Overall, integrating AI and inquiry tools shows promise for improving outcomes and sustaining inquiry practices in higher education.