隨著企業加速推動數位轉型與生成式人工智慧（Generative AI）應用，AI素養已逐漸成為企業人才培育之核心能力。然而，在企業培訓情境中，講師往往難以即時掌握學習者之情緒狀態與學習參與歷程，進而影響教學調整與學習支援成效。為此，本研究建置整合臉部情緒偵測、學習分析儀表板與生成式AI回饋之智慧學習參與支援系統（SLESS），探討其於企業AI素養培訓情境中對學習成效與參與歷程之影響。 本研究採準實驗設計，招募30位企業在職人員並分派至實驗組與控制組。兩組皆使用相同平台、教材與課程，其中僅實驗組額外導入情緒偵測、儀表板回饋與AI輔助功能。研究工具包括AI素養理解測驗（AILT）、學習動機量表（IMMS）、科技接受度量表（TAM）、參與度指標（ES）、情緒序列分析（GSEQ）及開放式問卷等資料。研究結果顯示，兩組學員於AI素養理解測驗皆呈現進步趨勢，其中實驗組後測表現及學習成長幅度顯著優於控制組，顯示SLESS有助於提升企業學員之AI素養理解。在學習動機方面，兩組之前後測皆顯著提升，但組間差異未達統計顯著水準。科技接受度結果顯示，實驗組對系統整體持正向評價。系統歷程資料與情緒序列分析顯示，學員整體維持高度學習參與，且有效學習歷程多呈現穩定專注、持續思考與問題解決特徵；部分負向情緒表徵未必代表低參與或低成效，而可能反映較高之認知投入。此外，即時回饋與學習歷程視覺化有助於促進學習者之自我覺察、自我監控及持續參與。 綜合而言，本研究顯示情緒偵測、學習分析儀表板與生成式AI回饋之整合應用，不僅有助於提升企業AI素養培訓之學習成效，更能支援學習參與歷程、自我調節及教學決策，並形成企業AI素養培訓情境中的智慧學習支援機制。;As enterprises accelerate digital transformation and the adoption of Generative Artificial Intelligence (Generative AI)， AI literacy has become a core competency for workforce development. However， in corporate training environments， instructors often face difficulties in monitoring learners’ emotional states and learner engagement portfolio in real time， which may limit timely instructional adjustment and learning support. To address this issue， this study developed the Smart Learning Engagement Support System (SLESS)， which integrates facial emotion detection， a learning analytics dashboard， and Generative AI-based feedback， and investigated its effects on learning outcomes and learner engagement portfolio in corporate AI literacy training. This study adopted a quasi-experimental design involving 30 corporate employees who were assigned to either the experimental group or the control group. Both groups used the same learning platform， instructional materials， and course content， while only the experimental group received additional emotion detection， dashboard feedback， and AI-assisted support functions. Research instruments included the AI Literacy Test (AILT)， the Instructional Materials Motivation Survey (IMMS)， the Technology Acceptance Model (TAM) questionnaire， Engagement Score (ES)， Lag Sequential Analysis (GSEQ)， and open-ended questionnaires. The results showed that both groups demonstrated improvement in AI literacy performance， while the experimental group achieved significantly better post-test performance and greater learning gains than the control group， indicating that SLESS contributed to enhancing learners’ AI literacy understanding. Regarding learning motivation， both groups showed significant improvement， although no significant between-group difference was observed. TAM results further indicated that participants in the experimental group held positive perceptions toward the system. System log data and GSEQ analysis revealed that learners generally maintained a high level of engagement throughout the training process. Effective learner engagement portfolios were characterized by sustained attention， continuous thinking， cognitive challenges， and problem-solving behaviors. Moreover， emotional states traditionally regarded as negative did not necessarily indicate low engagement or poor performance; instead， they might reflect higher levels of cognitive involvement. The findings also suggest that real-time feedback and learning process visualization support learners’ self-awareness， self-monitoring， and sustained engagement. Overall， the integration of emotion detection， a learning analytics dashboard， and Generative AI-based feedback not only enhanced learning outcomes in corporate AI literacy training but also supported learner engagement portfolio， self-regulation， and instructional decision-making. The findings further suggest the formation of a smart learning support mechanism for corporate AI literacy training.
