Current thinking in education emphasizes a transition from traditional course design to enhance flexibility in instructional materials and methods to better meet learner needs and enhance learning outcomes. New “blended learning” approaches combine face-to-face learning in traditional classrooms with internet-based learning. Young (2002) notes that blended learning is an established trend in higher education, and predicts this format will eventually become the mainstream. Accelerating economic change and industrial disruption emphasizes the need to provide students with flexible problem solving skills which can be adapted and applied to future conditions which are difficult to anticipate. Problem-Based Learning (PBL) and other educational strategies have been found to significantly improve learning outcomes and to promote the development of effective problem solving attitudes and abilities (Schoenfeld and Herman, 1982; Brown and Kane, 1988; Lin, 1992; Huang, 2005; Hsieh, 2007; Huang, 2011). This research seeks to develop a concept of blended problem-based learning (blended PBL), which is then applied in university biotechnology courses to assess the impact on learning attitudes and attitudes and abilities for problem solving.

Recently, computer-based learning environments such as Massive Open Online Course (MOOC) and Intelligent Tutoring System (ITS) become more prevalent. One important characteristic of these computer-based learning environments is that they can capture in their log files what students are doing while trying to learn new knowledge without interrupting their learning processes. Since log files allow for re-constructing how students used computer-based learning contents, analyzing the log files can enable us to quantitatively study the learning behaviors of students and the usefulness of computer-based learning contents. As a result, Learning Analytics (LA) and Educational Data Mining (EDM) are emerging as new exciting research fields. One active research topic in EDM is predicting whether students can correctly solve a problem because many computer-based learning environments, especially for mathematics and science, use problem solving as a primary means for assessing student learning. As Koedinger and Aleven (2007) pointed out, in order to maximize student learning outcomes, it is critical to balance giving and withholding instructional supports and guidance in the computer-based learning environment. Students may not exert enough cognitive efforts and fail to acquire a schema from learning tasks if they receive instructional supports prematurely (Kapur, 2008; Schmidt and Bjork 1992). Academically weaker students, on the other hand, will fail to learn unless they are provided with appropriate instructional supports and guidance in time. This issue would become more important in the computer-based learning environment where students learn primarily on their own and there is no teacher who can regulate the learning process of students. Most computer-based learning environments rely on either simple heuristics (e.g., giving hints or feedback after students fail to solve a problem a certain number of times) or discretion of students in determining when instructional supports and guidance are going to be provided Obviously, simple heuristics would not be able to maximize the learning outcome of students because it does not take into account the difficulty of learning tasks and the ability of students. Similarly, providing instructional supports on the demand of students may not improve their learning outcome because novice students do not possess metacognitive abilities and prior knowledge required for determining the right moment to ask for help (Clark and Mayer, 2003; Lawless and Brown, 1997). To maximize the learning outcome of students, computer-based learning environments should be able to make a more intelligent decision based on the difficulty of learning tasks and the ability of students, which requires a quantitative model of student performance.