Computerized versus paper-and-pencil assessment of socially desirable responding: Score congruence, completion time, and respondent preferences.

摘要

Computers' lightening fast calculation speeds and high precision are largely responsible for their influx into the testing arena. At first glance, transferring test or questionnaire items from a paper-and-pencil instrument to a computerized version may seem trivial, with the inference of score equivalence being intuitively obvious. However, equivalence of scores from the two modes cannot be assumed. Because interacting with a computer is a different physical and psychological experience, respondents sometimes answer differently across modes of administration. To treat results from computerized and paper versions as interchangeable, compelling evidence must be provided to assure comparability in the scaling, measurement precision, and validity of scores across modes.;The focus of this dissertation was on the comparability of results yielded by computerized and paper versions of one of the most commonly used measures of socially desirable responding (SDR) to questionnaires---the Balanced Inventory of Desirable Responding (BIDR, Paulhus, 1991). Research literature into mode effects on SDR has yielded conflicting results. The goal of this dissertation is to overcome limitations of previous research so that effects of mode on SDR scores, completion time, and respondent preferences could be clearly understood. Key improvements in this dissertation over previous studies included: two independent studies with a large number of participants (ns = 710 & 500), a within-subject design with counterbalancing, tightly controlled administration conditions (anonymity guaranteed, both forms administered individually in the same private cubicles), a computerized BIDR with review, skipping, and backtracking capabilities, rigorous analyses of central tendency, variability, reliability, and construct validity of BIDR subscale scores, and examination of questionnaire completion time and respondent preferences in addition to subscale scores.;Results for the psychometric properties of scores were very consistent within and between the two studies, providing strong support of comparability in the scaling, reliability, and construct validity of BIDR subscale scores across modes. Completion times were slightly longer for computer administration in both studies, but this difference had little practical significance. Finally, participants across both studies preferred the computerized version with respect to readability of items, inputting and changing responses, quality of the assessment experience itself, and credibility of information collected.