Reflection-based deep learning, essentially a type of deep learning, aims to promote deep learning by developing reflective learning ability. In order to test "the promoting effects of reflection on deep learning", according to the process model of reflection-based deep learning, we designed the experimental scheme. Firstly, through the pretest questionnaire, we determined the experimental subjects according to their learning capacity level; then through the specific teaching interventions, we guided the experimental class to carry out their reflective learning activities throughout the whole class; finally through variance analysis of the written test and the work creation, we confirmed that the model of reflection-based deep learning had certain feasibility, furthermore the reflection activities really can effectively promote students’ deep learning. Keywords—reflection; reflective learning; deep learning; reflection-based deep learning I. BACKGROUND OF EXPERIMENTAL RESEARCH As early as the mid of 1950s, Marton and Saljo were the first to start the experimental study of deep learning [1]. Around the same period, many other researchers developed the research on deep learning from different perspectives. Through the analysis of deep learning scales developed by Biggs, Entwistle and Ramsden, Nelson Laird and his colleagues found that deep learning can be deconstructed into three component parts: high-order learning, integrative learning, or reflective learning [2]. In other words, all of them are the concrete forms of deep learning. Reflection-based deep learning is mainly about "how to promote deep learning by reflection". Based on lots of literature analysis, we had demonstrated the mutual promotion between deep learning and reflective learning from the theoretical level, and through incorporating the ideas about reflection into the general process of deep learning, constructed the process model of reflection-based deep learning as shown in Figure 1[3]. Now we will put these theories into practice, which is to combine with the relevant courses, design the experimental scheme and develop the teaching activities. Fig. 1. The process model of reflection-based deep learning II. EXPERIMENTAL DESIGN AND IMPLEMENTATION A. Experimental design Deep learning aims to achieve overall learning objectives and develop high-order thinking ability [4]. So is reflectionbased deep learning, which is a special kind of deep learning. As promoting learning is the common character of deep learning and general learning, we can assess deep learning from the traditional point of cognition, motor skills and emotion. Moreover, to develop high-order thinking ability is its particularity, so we also need to measure learnings’ thinking structure level and identify their cognitive development level according to Biggs’ SOLO(Structure of the Observed Learning Outcome) taxonomy. In order to examine the promoting effects of reflection on deep learning, a quasi-experimental study is designed, which has both the experimental group and the compared group at the pre-test and the post-test. And put forward the following three hypotheses: firstly, reflection can help attain the intellectual target of deep learning, especially the high-level cognition of deep knowledge and complex concept; secondly, reflection can promote learnings’ understanding of the problem, and then International Conference on Social Science and Higher Education (ICSSHE 2015) © 2015. The authors Published by Atlantis Press 101 help them get complex learning outcomes and high-level thinking ability; thirdly, reflection can effectively facilitate the learners’ grasp of information technology skills and improve their problem-solving ability and practical-creating ability. B. Experimental implementation In the preparation stage, through the discussion with cooperative teacher, we determine the teaching contents and adjust the teaching plan accordingly. Then in the implementation phase, we observe the two groups’ teaching for eight weeks and properly modify the experimental material. It is composed by pre-test phase, treatment phase, and post-test phase. 1) Pre-test phase In order to ensure the experimental effect, we should choose two classes as similar as possible. According to the outcomes of questionnaire survey and variance analysis, we determine to take class 4 and class 5 as the experimental subjects. And randomly class 4 is selected as experimental class to adopt the model of deep learning with reflection, while class 5 is selected as compared class to adopt the model of deep learning without reflection. 2) Treatment phase Knowledge content is the object for learners to recognize and practice [5]. Spiro considers that knowledge can be divided into well-structured knowledge and ill-structured knowledge, accordingly learning can be called primary learning and advanced leaning. Deep learning aims to solve the ill-structured problem of the real social situation and complex technology environment through the integration of multi-dimensional knowledge and the construction of new knowledge and meaning [6]. So the involved knowledge should be complex, open and ill-structured, and also easy to integrate, construct and contextualize. Reflection-based deep learning intents to discover, inquire and solve problem by the reflection and regulation of learning process and outcome, and then promote the development of high-level thinking and the realization of deep leaning. So we can say that the two key points of reflection-based deep learning are reflection and problemsolving. Hence we choose the chapter "the acquisition and processing of image" from information technology course in senior middle as the teaching content of our experiment. And based on the process model of reflection-based deep learning and the teaching objectives of the experimental section, we design the experimental program and put it into practice [7]. The concrete process of reflection is as following: a) Reflection in the inductive stage Before the new class, we will guide learners to review the old knowledge by creating the problem situationhow to make a poster with texts and pictures. And then elicit the new teaching content by asking the question-how to deal with the pictures. Meanwhile, require learners to fill in "the pre-course self-test form". b) Reflection in the main stage The reflection in class mainly follows the "ConflictReflection-Generation" mode, as shown in Figure 2 when lecturing on "the classification and comparison of graphs and pictures", that is, to create the cognitive conflicts by showing two different kinds of images, and then require learners to generate solution to eliminate them through reflecting old knowledge. Besides, this reflection can also follows the "Variant-Reflection-Migration" mode, as shown in Figure 3, to regulate the process of variant practice through reflection and promote the exploration and solving of new problems, for example when asking learners to design and create their own work by Photoshop. Meanwhile, require them to fill in the record form about work designing and creating, which aims to guide learners to reflect, extract the relevant knowledge and skills, and regulate their work creation. Fig. 2. The Reflection following "Conflict-Reflection-Generation" mode Fig. 3. The Reflection following "Variant-Reflection-Migration" mode c) Reflection in the evaluated stage After every class, learners are required to reflect on their classroom learning process and result by fill in a record form