- 學位論文
視覺化模擬輔助人工智慧教學之研究-以類神經網路為例
Learning Artificial Intelligence with Visualization and Simulation: The Case of Neural Networks
摘要
當今社會中，人工智慧影響我們的生活面向甚廣。目前國際上的教育相關政策也將人工智慧議題納入探討，並期望從小教導孩子人工智慧相關知能。本研究針對高中年段的學生，設計視覺化模擬輔助人工智慧教學策略，並發展學習平台，透過「概念理解」、「概念反思」、「概念應用」三個教學步驟引導學生進行概念學習，教學主題聚焦於類神經網路。本研究以實證研究探討視覺化模擬輔助人工智慧教學與傳統講述式教學對高中生之人工智慧學習成就、學習態度影響之差異，以及學生對於視覺化模擬輔助教學的感受。從教學實驗結果發現： 一、視覺化模擬輔助教學對人工智慧學習成就之影響 本研究發展之視覺化模擬輔助教學能透過模擬操作幫助學生建立概念：(1)， 提供學生操作與調整參數、觀察實驗結果的機會，使之能於操作過程中檢驗概念；(2) 輔助進行運算過程，以降低認知負荷並聚焦重要學習概念；(3) 模擬工具的設計融入日常生活情境，以幫助學生以解實際應用。實驗組學生因而能建立較正確清晰的人工智慧概念，進而增進其概念理解上的學習成就。但由於本研究的教學中程式設計相關教學內容較少，因此與傳統教學相較，雖亦使演算法程式實作有更好的表現，但其差異未達顯著。 二、視覺化模擬輔助教學對人工智慧學習態度之影響 實驗結果發現，使用視覺化模擬輔助教學之學生，由於能透過模擬操作測試概念並即時得到概念學習的回饋，對於自身學習成果的信心顯著高於接受傳統教學之學生。但兩組學生在「學習動機」、「自我效能」、「資訊科學抽象概念/程序之學習感受」面向沒有顯著差異。 三、學生對視覺化模擬輔助教學之感受 基於本研究的量化資料與質性訪談資料分析結果，視覺化模擬輔助教學相較於傳統教學，能使學生有更高的學習成就，實驗組學生也普遍認為視覺化模擬輔助教學對他們學習人工智慧相關知識有所幫助，此助益對學習艱深複雜的概念更加明顯。此外，訪談結果亦顯示學生認為模擬平台能夠幫助他們學習較抽象、具複雜運算的課程概念。
並列摘要
Artificial Intelligence (AI) is growing rapidly to fit the needs of our everyday life. To introduce students to the AI world, many advanced countries start to discuss how to develop and implement effective AI instruction for k-12 students. However, existing AI instruction focuses more on higher education. In addition, AI topics involve abstract and complex concepts and skills, which are difficult for k-12 students. This research aims to design and develop an AI instruction for high school students by employing visualization. Students learn with a simulation-based platform, and the proposed instruction consists three stages: concept comprehension, concept reflection, and concept application. This study conducts an empirical study to explore the effects of the proposed simulation-based instruction on learning achievement, attitude toward AI learning, and perceptions of simulation-based instruction. The research findings are as the following: 1. The effectiveness of the simulation-based AI instruction on learning achievement: The proposed simulation-based AI instruction provides students with the opportunity of adjusting parameters and observing the changes, which helps clarify the concepts. The computing function of the simulation platform can also help reduce the cognitive load. In addition, the authentic simulation context connects abstract concepts to real-world applications, so that students can grasp the AI concepts more deeply. 2. The effectiveness of the simulation-based AI instruction on learning toward AI learning: The experimental group with the simulation-based AI instruction has significantly higher confidence in their AI learning than the control group with traditional instruction. However, there is no significant difference between the two groups of students in terms of "learning motivation", "self-efficacy" and "perceptions of learning abstract concepts/skills". This might be because the experimental period is not long enough. 3. Students' perceptions of the simulation-based AI instruction The experimental group with the simulation-based AI instruction has significantly higher learning achievement than the control group. Students in the experimental group also generally believe that the simulation-based AI instruction is beneficial for their AI learning. The simulation tools help students manipulate the abstract and complex AI concepts, and then understand more clearly.
參考文獻
延伸閱讀
- 王聿（2021）。視覺化模擬輔助人工智慧教學之設計與評估〔碩士論文，國立臺灣師範大學〕。華藝線上圖書館。https://doi.org/10.6345/NTNU202101374
- 賴思妤（2022）。基於認知神經科學之運算思維導向程式設計視覺化輔助學習平台設計〔碩士論文，國立臺灣師範大學〕。華藝線上圖書館。https://doi.org/10.6345/NTNU202201839
- Jui, H. P. (2023). 以視覺為基礎之示範學習與協作機器人系統 [doctoral dissertation, National Taiwan Normal University]. Airiti Library. https://doi.org/10.6345/NTNU202300385
- 林郁展（2022）。人工智慧實作教學研究：以電腦視覺為例〔碩士論文，國立臺灣師範大學〕。華藝線上圖書館。https://doi.org/10.6345/NTNU202300021
- 鄭又銘、陳文輝（2019）。Using Attention Models on Neural Networks for Human Activity Recognition。國立宜蘭大學工程學刊，(13)，50-65。https://doi.org/10.6176/BCE.201912_(13).0005