解難能力是21 世紀人們需要具備的核心特質之一。20 世紀，由於研究方法的 限制及缺乏合適的測評工具，較少學者會研究受試者的解難歷程。隨著資訊科技的 進步，人們發展出記錄大樣本和大量解難歷程行為的方式log 文檔。本文以 PISA 2012 電子解難測試之一道公開試題（機械人吸塵機）的學生log 文檔作為研 究數據，探索出三個用於分析的表現指標：學生解難的時間、解難步驟的數目，以 及觀看測試素材的次數，以這些指標比較七個高表現地區和三個低表現地區的狀 況，發現：(1)低表現地區學生的解難時間比高表現地區長；(2)低表現地區與高表 現地區相比，解難步驟較少的學生百分比和觀看素材次數較少的學生百分比均較 高。之後，透過決策樹方法分析出五種學生解難模式，並進行跨地區比較和對應分 析，發現：(1)高表現地區學生多為「低頻答對」或「高頻答對」型，低表現地區 學生多為「低頻答錯」型；(2)學生整體解難水準和對問題解決持開放態度的水 準，與這五種解難模式分別在高表現和低表現地區有不同對應關係。最後，本文根 據研究結果為教育工作者提供關於解難教學，以及研究者相關後續研究的建議。

Problem solving is one of the core skills of the 21th century. In the 20th century, due to the limitation of research methodology and adequacy of assessment tools, very few studies focused research on the students’ problem-solving processes. With the rapid development of computer science and information technology, log files have been deployed to document students’ behavioral processes in large-scale sample surveys. This research studied the student log files of a released item (ROBOT CLEANER) in the PISA 2012 digital problem-solving assessment. Three behavioral indicators had been developed for analyses and comparison between the seven high-performing and the three low-performing economies: response time, number of problem-solving steps, and number of times watching the simulation. The comparison results showed that: (1) the overall response time of students from the low-performing economies was longer than students from the high-performing economies; (2) Comparing between the low- and high-performing economies, the proportion of students of low-performing economies who had fewer number of problem-solving steps was higher. The same was found regarding the number of times watching the simulation. Through the use of decision tree, five problem- solving patterns were uncovered. The correspondence analysis results showed: (1) “Lowfrequency Answer-correct” and “High-frequency Answer-correct” were the two dominant problem-solving patterns found in the seven high-performing economies, whereas it was “Low-frequency Answer-wrong” found in low-performing economies; (2) students’ levels of overall problem-solving performance and openness to problem solving were having corresponding relationships with the five problem-solving patterns. This research provided recommendations for improving problem-based teaching and learning, and suggestions for pursuing follow-up studies.