engleski

Assessing personality at work: Social desirability and item response latency as possible ways of coping with intentional response distortion

Many studies showed that it is possible to distort responses on a personality questionnaire in the desirable direction and that distortion is present in the context of personnel selection (e.g., Griffith, Chmielowski, & Yoshita, 2007). Although contradictory evidences exist regarding the influence of response distortion on the validity coefficients of personality questionnaires (Ones, Dilchert, Viswesvaran, & Judge, 2007), most of researchers agree that intentional response distortion influences selection decisions (Rose, Stecher, Miller, & Levin, 1998). Two possible ways of identifying applicants who distort their responses on personality questionnaires are social desirability scales and item response latencies. Two studies examining usefullnes of these methods will be presented. Study 1 Recently, Paulhus (2002) proposed a new social desirability model. According to that model, deliberate self-presentation in favorable light is not one- (unfavorable-favorable) but two-dimensional, and should be measured with two impression management scales differing in content. The first is the Communal Management scale which measures respondent’s tendency for denying socially deviant impulses and claiming saint-like attributes and the second, Agentic Management scale, indicates the level of exaggeration of one’s intellectual and social competency. The aim of the first study was to test whether new impression management scales (Paulhus, 2006) represent valid indicators of intentional response distortion on personality scales. A within-subject design study was conducted during which 213 participants completed a five factor personality questionnaire and social desirability inventory, both in honest and “fake-good” conditions. The results showed that both scales were sensitive to situational demands and, therefore could serve as response set measure. However, both impression management scales scores gained under “fake-good” condition weakly correlated with direct indicators of response distortion on personality scales, defined as difference scores between “faked” and honest responses on each trait. Moreover, correction for response distortion did not make “faked” personality scores more similar to those gained in honest condition. These results indicate that it is not justified to use results on these SD scales to correct results on personality scales, or exclude participants whose results indicate overly desirable responding. Study 2 Another possible way of identifying applicants who distort their responses on personality questionnaires are their item response latencies. If personality questionnaires are computerised, this is easily accessible information. However, studies examining item response latency of participants who distorted their answers resulted with contradictory findings. While some indicated that it takes longer to distort responses, other showed shorter latencies under “fake good” conditions. The aim of our study was to test the model of personality test item response dissimulation that was proposed by Holden and associates (1992). The model explains the processes beyond the differences in item response latencies between honest respondents and those who “fake” their responses. The basic assumptions of the model are that, while responding on a personality questionnaire, participants compare test items with their cognitive schema, and that there is an interaction between the schema and social desirability of a response. This interaction determines item response latency, and has an effect on within-individual and between-individuals level. The study was conducted in two motivational conditions: honest (N=66) and “fake good” responding (N=56). In both situations an on-line version of personality questionnaire for measurement of five broad personality factors and two social desirability scales for measurement of self-deception and impression management were used. Out of seven used scales, assumptions of the model on within-individual level were confirmed for the two (extraversion and impression management scales) and on between-individuals level for one scale (impression management scale). The results partially confirm the model of personality questionnaire item response dissimulation and indicate a need for further studies.

socially desirable responding; response distortion; personality questionnaire; item response latency

nije evidentirano