杜仲弘
杜仲弘
應用於適性化數位學習之智能代理人研發：以身心障礙學生社會技巧學習為例
Development of Smart Agent for Adaptive e-Learning: Social Skills Learning of Students with Disabilities as an Example
智慧代理人、數位學習、遊戲式學習、適性化學習、社會技巧、學習支持系統
Intelligent Agent、Smart Agent、Adaptive e-Learning、Social skills Learning、Students with Disabilities
隨著資訊科技的蓬勃發展，代理人應用於許多領域皆取得不錯的效果，在數位學習領域中也有許多代理人相關的應用。身心障礙學生具個別差異大的特質，特別需要適性化學習與輔導，若能透過代理人掌握學生於數位學習之學習狀態，如所面臨的學習困難或是當前發生的問題行為，並預測可能發生的學習困難或問題行為，分析其背後原因，了解引發學生問題行為之前事以及功能需求，給予學生適當的學習支持與輔導，當能讓學生獲得更大的學習成效。
本研究之目的在以智慧型代理人為體、資料科學方法與技術為用，設計具監控、預測、原因分析、決策、輔導、學習能力之智能代理人模型，再依此針對特殊學生社會互動學習，開發智能化社會互動學習輔導代理人，並透過遊戲式社會互動能力學習平台，驗證智能代理人之可行性與學習輔導代理人對於身心障礙學生社會互動能力提升之有效性。
本研究透過十二名台南市國小一至五年級資源班身心障礙學生，利用「次數和比率記錄」記錄方法，連續紀錄學生於遊戲時段內之正負向行為，透過視覺分析法以及簡化時間序列C統計，分析學生於遊戲時段內正負向行為頻率變化趨勢，以探討受試者在社會技巧各項目之進步狀況。實驗結果顯示受試者之問題行為頻率有明顯下降之趨勢，並且正向行為之趨勢呈現上升。透過學生遊戲動作歷程以及移動歷程可發現學生電腦操作技巧有所提升，且隨著實驗次數增加，學生從隨著系統提示幫助他人、請求他人幫助到主動與他人正向互動、合作搬運，其所展現之社會技巧有增加之趨勢，說明此學習平台對於社會技巧學習有正面效益。
The purpose of this study was to design a “smart agent model” that monitors, predicts behavior, conducts behavioral analysis, makes decisions, provides counseling, and learns through data science methods. Then, according to the model, a smart social interaction learning coaching agent is developed and applied to social interaction learning of special education students. Finally, the game-based social skills learning platform is used to verify the feasibility of special education digital learning smart agents and the effectiveness of learning and consulting agents on the social interaction ability of students with disabilities. The study arranged for 12 students from the Tainan City Elementary School to the fifth-grade resource class for disabilities to conduct experiments. Through the visual analysis method and simplified time series C statistics, the trend of the positive and negative behavior frequency of students during the game period is analyzed, and the progress of the disciplines in various social skill projects is discussed. The experimental results showed that the subject's problem behavior frequency decreased significantly and the trend of positive behavior increased. Students' computer skills can be improved through the student's game action footprint and game movement footprint. As the number of experiments increases, students help others or seek help from others by listening to the system prompts progress to actively interacting with others and cooperating to solve problems. The above situation indicates that the growth trend of students' social skills proves that this learning platform has positive benefits for social skills learning.