Search

The last few years have seen an increase in open and student-centred learning. Changing patterns in the requirements and constraints associated with training and education are encouraging a greater emphasis on self-study packages. However, the inevitable variations between students in terms of their background experience, level of expertise, learning objectives and study styles can lead to problems of specifying efficient study regimes, and in the management of such learning systems. Addressing these issues, an earlier paper in PLET (ETTI) described a prototype version of an adaptable knowledge-based educational Advisor capable of planning a route through a generalized topic-based curriculum domain and offering individualized learning advice to the learner. This paper presents an interim report on the more sophisticated second-level design currently being developed. A set of software tools in Smalltalk has been constructed allowing progressive implementation and refinement of the prototype and the current state of the system is described.

Despite widespread and increasing use of CAL teaching programs and self-learning materials, little serious evaluation or analysis of such systems has taken place. The consequential lack of a clear theoretical understanding of the cognitive principles involved prevents the formation of clear design criteria capable of ensuring maximal efficiency of the teaching material. This study sets out to explore three key variables which may influence the learning of statistics material: structure of presented material, mode (aural or visual) of presented material, and previous mathematical experience. Although the latter two were not found to be significant factors in learning, a detailed analysis of structure revealed many interesting results regarding its influence on the quantity and quality of student recall. Finally, the relevance of these findings to the design of learning materials is discussed.

Students normally perform processing tasks as part of their learning experience, for example by answering questions embedded in teaching texts. This experiment compared the quantitative and qualitative effects of types of processing which used semantic ('application' and 'transposition') and non-semantic ('rating') tasks.Three experimental groups studied the same core material, each receiving one of the styles of embedded tasks. Both semantic groups produced higher scores on subsequent tests of free-recall, cued-recall and comprehension than did the non-semantic group. The data also allowed comment on the interactions between quantity and quality of recall and mathematical background of students. Since one of the fortes of computer-assisted learning (CAL) is its ability to provide individualized teaching relating type of task to learner characteristics, these findings are very relevant to the design of CAL materials.

During the last decade there has been a resurgence of support for the notions of open and student-centred learning, extending far beyond the concepts originally pioneered by distance-learning institutions such as the Open University. Hence current and projected requirements and constraints in the fields of training and education are placing a greater emphasis on programmes of self-study, and many types of suitable materials are now available for such schemes, ranging from conventional books to sophisticated systems involving computers and interactive videodisc. However, the increasingly wide variations between students in terms of background experience and level of expertise can cause difficulties when specifying efficient study regimes. The solution proposed by this paper is the design of an adaptable knowledge-based educational Advisor, capable of planning and executing suitable routes for each individual learner through a generalized topic-space in order to achieve a specified set of didactic goals. The plan is generated by the computer according to a logical unfoldment of topics determined from the inter-topic links and is continually reviewed and revised according to learning outcomes. Execution of the study-route involves a further search as each stage is reached, this time within a database of study materials, in order to select those most appropriate for the student to achieve the learning goals. The individualized nature of the searches of the topic and materials spaces takes full account of data held about the student in a user model. The paper will first consider general design criteria for a system meeting such objectives and then present details of the implementation of a small working prototype Advisor developed in collaboration with Leeds University. Finally, our current work towards developing a more sophisticated design will be discussed.

A new hypothesis of spectrochemical excitation is proposed to explain the region of uncertain detectability, whereby it is possible to formulate equations that will cover the two concentration ranges found in spectrographic analysis. The first concentration range occurs when in replicate analysis on the same material all results are detected, the second concentration range when some results in the replicates are not detected. An example is quoted where the over-all correlation for these two levels is 0.9972 and χ2for three degrees of freedom is found to be 0.12.

When a dc arc is used for trace analysis on a spectrometer, calibration curves with a pronounced toe due to spectral background are found. For analysis of high purity materials it is most important that linear relationships should be established to ultra-trace levels when total impurity content is specified within the ppm level. A solution to this problem is outlined and the linear relationships obtained for copper, bismuth, and cadmium in high purity silver to 0.1 ppm are shown. An electronic circuit is described that enables these background corrections to be made.