金雅倫

Ya-LunChin

碩士論文（2020）

人工智慧為基之學習輔導技術研發：以身心障礙學生合作問題解決能力發展為例

Research on AI based Technologies for Learning Tutoring:Collaborative Problem Solving Skills Development of Students with Disabilities as an Example

關鍵字 Keywords

人工智慧、身心障礙學生、合作問題解決、數位遊戲式學習、機器學習、特殊教育

Artificial intelligence、Disabilities、Collaborative Problem-Solving、Digital Game-Based Learning、Machine Learning、Special Education

摘要

合作問題解決能力為現代公民不可或缺的技能。身心障礙學生因身心特質，普遍具有同理心低落、認知受限、社會互動技能發展遲緩與不足的現象，導致難以主動與人開啟對話、維護與瞭解雙方互動，因而不易發展有意義的社交關係，也造成合作問題解決能力不足。

資料科學(Data Science)與人工智慧蓬勃發展，並被廣泛運用在許多領域且有相當成效，也帶動了「智能化」的思維。依據資料科學之方法，將數位學習智能化，進行更深層的分析與決策，給予學生更即時、適性的輔導與支持，為當前數位學習之趨勢。

相對於一般學生，身心障礙學生對合作問題解決技能的學習與適性化學習輔導，有更高的需求。本研究參考資料科學之概念，規劃一整合多項學習理論，適用於身心障礙學生之「合作問題解決能力發展模式」，再針對身心障礙學生即時、適性化輔導的需求，設計「合作問題解決技能學習輔導模式」，依此模式，運用機器學習技術，開發學習監控、能力分析、決策、輔導、能力追蹤與優化等機制，並開發「數位遊戲式合作問題解決技能學習輔導平台」，提供合作問題解決學習情境案例與輔導，在學生的學習過程中，給予適性化的輔導，提高學習成效。

本研究以國小三年級至六年級身心障礙學生為對象，於此「數位遊戲式合作問題解決技能學習輔導平台」進行實驗，利用單一個案研究方法進行實驗與後續分析，檢驗此一合作問題解決學習輔導模式之可行性與有效性。實驗結果顯示學生合作問題解決能力明顯提升，且社會互動行為亦有所增加。同時，依據實驗結果，分析身心障礙學生之社會互動能力與異常行為對其合作問題解決能力之影響，以及合作問題解決技能間之相關性。

Abstract

Collaborative problems solving (CPS) skills are indispensable ability for modern citizens. Due to physical and mental characteristics, students with disabilities generally have low empathy, limited cognition, and slow and insufficient development of social interaction skills, which makes it difficult to initiate dialogue with others, maintain and understand interactions with others, and therefore it is difficult to develop meaningful social relationships. It also causes insufficient ability for CPS skills.

Data Science and artificial intelligence are booming, and are widely used in many fields with considerable results, and they have also promoted intelligent thinking. According to the method of data science, the digital learning is intelligentized, and deeper analysis and decision-making are carried out to give students more immediate and appropriate guidance and support. This is the current trend of digital learning. Due to the high heterogeneity of physically and mentally handicapped students, it is even more necessary to provide appropriate learning guidance in response to individualized differences.

This research reference the data science concept, designing a CPS Ability Learning Tutoring Model, using machine learning technology to develop learning monitoring, ability analysis, decision-making, tutoring and optimization mechanisms, and developing a Digital Game-Based Learning Tutoring Platform, to provide appropriate guidance and improve learning effectiveness. Finally, this experiment is to verify its effectiveness and feasibility of this CPS Ability Learning Tutoring Model.