Computerised Tailored Testing: Structured and Calibrated Item Banks for Summative and Formative Evaluation
Dieudonnè A. Leclercq · European Journal of Education · 1980
Educational Evaluation in Progress An educational evaluation is always part of a regulation process, and regulation processes in education have already been fully discussed elsewhere (D. Leclercq, I976, I978a and 1978b). Evaluation serves objectives that can be obvious, or that can be less evident, such as abilities certification (and thus selection), achievement prediction (and thus orientation), and individual proceedings analysis (formative purpose). The first two functions of evaluation (selection and orientation) have been developed in education since the beginning of this century in the field of psychometrics. The key-concepts are those of traditional test theory, i.e., test standardization, validity, reliability, and accuracy of scores. Psychometrics corresponds to summative evaluation, that seeks one stable and representative score. The third function of evaluation (formation) seeks analytical and causal information. The mastery learning approach has been settled on the principles of early diagnosis and immediate remedial treatment. Behaviourists, such as B. F. Skinner, have shown that a frequent and quick feedback can have a positive effect on the acquisition of abilities. Mastery learning illustrates the educationists' awareness of the dangerous 'normal curve myth', so that 'edumetrics' can now be developed. More recently, other methods have appeared, enabling self-regulation (self-diagnosis, self-correction, etc.) and self-assessment (in which the student expresses, by means of subjective probabilities, how confident he is of the correctness of his answer). A number of theoretical and practical works dealing with this problem have been published (D. Leclercq, 1975 and I977a, b and c). Not only do these methods open up new prospects for educational assessment, but they also offer the advantage of placing the student at the centre of his own learning, of having him assume the responsibility for his competence, and of getting him involved in his own development. Meanwhile, psychometrics has also evolved. The traditional test theory has been thoroughly revised in the light of the 'generalisability theory' (Cronbach et al., I972, and Tourneur & Cardinet, I978). The concepts of individual ability (A) and item difficulty (D) have been re-examined in the light of the Rasch model, in which the probability of a right answer to an item depends only on the individual's ability (A) and on the difficulty of the item (D). This probability is noted P(i I A, D), where P stands for 'probability', i for 'success', and the sign I means 'given'. The two parameters (A) and (D) are expressed in the same units and plotted on the same (horizontal) scale of measurement.