| 摘要: | 本研究透過腦電圖分析學生解題過程中的深度認知歷程。學習者的認知歷程一直是教育研究的重要議題，深入了解認知歷程有助於教師掌握學生的學習狀態與成效。傳統上，相關研究多仰賴問卷等主觀資料，難以真實反映學生的認知狀態；而眼動儀僅能夠確認學員的視線落點推測其在題目敘述區進行閱讀，或在選項區進行決策，但難以進一步辨識其推理等深度歷程。為了得知學員的推理歷程差異，本研究在學生的答題歷程中記錄其腦電圖，透過資料導向的分群方法，客觀分析學員在解題過程的認知狀態，突破過去眼動儀無法得知推理歷程的限制，並將腦電波訊號分類為閱讀、推理與決策三種具代表性的認知狀態。為量化每位學生在各認知狀態的時間比例，進一步提出認知狀態比例指標( State Ratio Index, SRI )，並以此為特徵建構學員表現預測模型，探討不同認知策略與作答結果之關聯性。結果顯示，在大多數情況下學習表現較佳的學生於作答含有圖片資訊的題目時，其推理階段的佔比高於決策階段，而學習表現普通的學生則呈現相反趨勢。這表示學習表現較佳的學生在解題時較常透過圖文整合進行推理，而普通學生則可能較偏向直接對應問題與答案，導致決策階段的佔比較高。為驗證分群結果的合理性，本研究依據分群結果之標記繪製腦電波拓樸圖，發現額葉區 θ 波功率在推理階段高於決策與閱讀階段( p<0 .05 )，決策階段亦高於閱讀階段( p< 0.05 )；額葉區 β 波功率則在推理與決策階段均高於閱讀階段( p< 0.05 )。本研究採用腦電波訊號探討學習者於解題過程中的認知行為，其結果提供了一種客觀方法來探索不同解題階段大腦的認知狀態，並為未來個別化學習與教育評估的應用提供參考。;This study investigates students′ deep cognitive processes during problem-solving through electroencephalography (EEG) analysis. Understanding learners’ cognitive processes has long been a crucial issue in educational research, as it helps educators grasp students’ learning status and outcomes. Traditionally, related studies have relied on subjective measures such as questionnaires, which may not accurately reflect students’ true cognitive states. While eye-tracking devices can determine gaze locations—such as whether students are reading the question stem or making decisions among the options—they cannot effectively capture reasoning processes. To explore differences in students’ reasoning phases, this study recorded EEG data throughout the problem-solving process. By employing a data-driven clustering method, we objectively analyzed students’ cognitive states during problem-solving, overcoming the limitations of eye-tracking in identifying reasoning activities. EEG signals were classified into three representative cognitive states: reading, reasoning, and decision-making.
To quantify the proportion of time each student spent in these states, we proposed a novel metric called the State Ratio Index (SRI). This index was then used as a feature to construct a performance prediction model, aiming to examine the relationship between cognitive strategies and problem-solving outcomes. The results showed that, in most cases, high-performing students spent a larger proportion of time in the reasoning phase than in the decision-making phase when solving problems containing visual information. In contrast, average-performing students exhibited the opposite trend. This suggests that high-performing students tend to engage in more integrative reasoning using both textual and visual information, whereas average-performing students may rely more on direct matching between questions and answers, leading to a greater proportion of time spent in the decision-making phase.
To validate the clustering results, EEG topographical maps were generated based on the identified clusters. The findings revealed that theta power in the frontal region was significantly higher during reasoning than during decision-making and reading (p < 0.05), and also higher in decision-making than in reading (p < 0.05). Beta power in the frontal region was likewise significantly higher during both reasoning and decision-making phases com-pared to reading (p < 0.05). This study employs EEG signals to investigate learners’ cognitive behaviors during the problem-solving process. The results provide an objective method to explore the brain’s cognitive states at different problem-solving stages and offer a reference for future applications in personalized learning and educational assessment. |
