Cognitive Mechanism Of Scientific Discovery Learning And Software Modeling Of Its Learning Environments

The simulation based scientific discovery learning environment is one type of constructivistic learning environment and has been used wildly in science education. However, the majority of these learning environments is domain-dependent and lack of support for learning activity. Therefore, this thesis intends to build a domain-independent and generic scientific discovery learning environment. To achieve this target, three sub-problems need to be solved: (1) what is the cognitive mechanism of scientific discovery learning? (2) How should the internal data model of scientific discovery learning environment be organized? (3) How should the functions and data views of software be designed to support scientific discovery learning?Firstly, based on the research framework of information process theory about problem solving, this thesis classifies scientific discovery learning as a well defined problem solving activity in a special domain, which is essentially a kind of induction reasoning. The task environment of scientific discovery learning consists of experiment model and scientific theory model, and the induction logic indwells the hierarchy of concepts and relations. On the layer of information process, scientific discovery learning can be viewed as a dual search between hypothesis space and experiment space with two strategies-theory driven or data driven and three core sub-processes-search hypothesis space, search experiment space and evaluate evidence. To help learners overcome difficulties during the process of learning, providing domain knowledge and meta-knowledge to them are necessary. In addition, self-adjusting and reflection are important.Secondly, the system model of a simulation based learning environment can be divided into domain knowledge model, learner model and activity model. Domain knowledge model includes simulation model and conceptual model. To realize a domain-independent simulation, the simulation model is implemented in a frame based knowledge representation plus simulation engine way; and the conceptual model is constructed in a traditional way, which is often used in intelligent tutoring systems. The learner model is responsible to record the cognitive process of a learner, and implemented by mapping the learner's hypothesis space and experiment space. According to the normative procedure scientific discovery, the activity model provides appropriate interaction spaces such as problem interaction space, hypothesis interaction space, experiment interaction space and conclusion interaction space. Thirdly, on the issue of human-machine interface design in scientific discovery learning, this thesis summarizes existed studies and point out the design of interface should emphasize five aspects: showing domain knowledge, formulating hypothesises, designing experiments, processing data and self-adjusting.Finally, based on the research described above, we introduce the prototype software GSDLE, which is a generic scientific discovery learning environment for continuous systems.