On the Development of CAI Course Wares by Group Data

目前在CAI中采用的课程决策方法，无论是线性类型（斯金纳）还是分支类型（克劳德），都缺乏灵活性。即使在分支类型中，学习者也只是遵循程序员决定和提出的几个分支。这意味着在具有固定分支的CAI中，没有什么比程序员对材料的教学能力更受期待的了。更糟糕的是，只要课程决策的灵活性是有限的，它基本上是不可能准备一个系统，以应付各种个体差异，因为软件的大小不可能是无限的。因此，智能CAI的研究取得了很大的进展。然而，这些研究被认为是成功的条件下，子程序的答案是相当清楚的。但在现实中，它们在许多情况下并没有得到澄清。我们的研究采取了不同的方法。我们假设个人合适的学习课程是通过分析群体数据和个体差异从现象学角度决定的。我们得出的结论是，通过将个人数据放入群体数据中，多变量分析在许多获取重要信息的尝试中是有效的。我们发现，在这些情感领域，如理解图片和因素分析，这些是直接适用的。在新材料的智力学习中，学习者的数据本身几乎没有偏差。因此，学习课程应该根据教师所假设的材料的逻辑结构与个人学习者对放置在组数据中的材料的假设之间的相互作用来决定。最后，经过大量的讨论，我们得到的信息，制定算法的课程决策的个人学习者。

The method of course decision, presently adopted at CAI for practical use, lacks flexibility regardless of either linear types (Skinner) or branching types (Crouder). Even in branching types, the learner simply follows several branchings decided and presented by the programmer. This means that in CAI with fixed branchings nothing more expected than the programmer'steaching ability of the material. What is worse, as long as the flexibility of course decision is limited, it is basically impossible to prepare a system coping with various individual differences because the size of softwares cannot be limitless. For these reasons, studies of intelligent CAI have been making great progress. These studies, however, are seen successful under the condition that the answers to sub-procedures are quite clear. But in reality, they are not clarified in many cases. Our study took a different approach. We assumed that a proper learning course for an individual is decided phenomenologically, by analysing the group dataand individual differences.We came to the conclusion that multivariate analyses were effective in many attempts to obtain significant information, by placing an individual's data in the group data. We found that in such affective domains as apprehending pictures and factor analyis, these were directly applicable. In the case of intellectual learning of new materials, the leaners' data itself has little deviation. Thus, learning courses should be decided in terms of interaction between the logical structures of the material assumed by the teacher, and the individual learner sassumption about the material placed in the group data. Finally, after much discussion, we obtained information for making algorisms of course decision for individual learners.