Investigation of scaffolding learning by the development of a server-client-based intelligent learning system.

我们开发了一个教授本科生自然演绎的智能辅导系统。该系统为学生提供下一步推理的推荐推理规则作为脚手架，并通过规则的应用得出推理结果。通过实验研究，我们证实，随着支持水平的增加，问题解决绩效提高；但学习收益在一定的支持水平上达到最大，然后随着极端学习支持的增加而下降。

We developed an intelligent tutoring system that teaches undergraduate students Natural Deduction. The system gives students, as scaffolding, recommended inference rules for the next inference step, and inferred results drawn by the applications of the rules. Through experimental investigation, we confirmed that as the levels of support increased, the problem solving performance was raised; but the learning gains reached the maximum at a certain support level, and then decreased with the further extreme learning supports.