UBC Theses and Dissertations
Conversational computer terminals in psychological testing Longbottom, Ronald Arthur
The concept of CCTT (conversational computer terminal testing) is investigated as a practical and advantageous solution to the problem of administrating a statistically-complex testing model, which will minimize the time losses and delays of conventional psychological testing. In the process of personnel selection, larger organizations often invite certain of their applicants to undergo psychological testing. This usually involves inconvenience for the applicant, additional expense for the organization, and time losses for both. The testing procedure will typically be of a fixed treatment, group testing nature whereas for many applicants an adaptive testing procedure would be more appropriate. Although testing mechanisms for adaptive treatments exist, they tend to be awkward and limited. More sophisticated theoretical models have been developed but tend to lack appropriate means of administration. In this investigation a Fortran IV CCTT probability/sequential model is developed and comparisons are made between CCTT and conventional test administration of verbal and numerical analogy tests to 180 subjects. The CCTT model presents an applicant with a test item, scores the response, and statistically computes and predicts group membership within pre-definable accuracy levels. The three possible classifications and associated treatments are : high scorer, administer next test; low scorer, terminate testing; unclassified, continue present test. Findings of the investigation indicate that there are no significant differences in overall test scores between the two methods of test administration. The CCTT model is able to classify individuals with an accuracy level exceeding $0% on bo verbal and numerical tests after only 25% of the test items have been administered, with a resultant reduction in testing time of 65%. Additionally, CCTT offers the benefits of individualized demand testing, automatic scoring and recording, complex and tailored decision branching, and immediate interpretation of results. Suggestions for further research in this and associated areas are provided.
Item Citations and Data