How we change what others think, feel, believe and do
Reducing faking in tests
Whilst faking is a perennial issue, designers of selection tests have a number of tools and techniques available that can be used to counteract or at least detect this.
At the start of a test, the instructions given to candidates can include warnings of the consequences of detected faking and request honesty. Instructions may also ask candidates to answer quickly, with their first thoughts rather than pondering. Holden et al (2001, p160) indicate that lying takes time. This is also supported by Ekman’s (1985) general study of lying.
It is also possible to include ‘trick’ questions, where a faking response is easily identified and hence raises suspicions about all other responses. For example in an assessment of a given set of skills, a multiple-choice question may have no right answer. Earlier assertions may later be probed in more detail.
Instruments that use self-reporting may give false readings when they are used by candidates who have insufficient self-insight to be able to be answer fully. If information is collected from multiple sources then this problem may be reduced, for example through the use of ‘assessment centres’ where multiple methods and assessors give a range of data and viewpoints which can be cross-checked.
Test takers who use the ‘central response tendency’ and opt for ‘safe’ central options may be identified by asking different questions for which a consistent response would include high and low responses.
Where individuals have a high need for approval, they may tend towards positive ‘agree’ and ‘yes’ responses. This may be countered and detected by reversing some questions (reversing also breaks up habituating patterns of similar responses). This tendency towards seeking approval may also be detected by including a ‘social desirability’ scale within the questions to enable isolation of this.
Assessing the same attribute with multiple questions can also show whether the candidate is averaging across questions (‘I’ve been a bit negative, I think I shall be positive for a while now.’), although obvious care needs to be taken to ensure that similar questions are interpreted in the intended way. Analysis of sequential patterns of positive and negative responses across responses may also identify uncertainty or deliberate averaging.
Normative items ask the candidate to rate their level of agreement with statements, and can give a good measure of psychological characteristics (Kline, 1993). However the question of faking has led to an ipsative approach being used in many contexts, where the test-taker is forced to make a choice from a fixed number of options. Ipsative questions either offer choice between items from very different areas (one question I recall from such a test is ‘Which do you prefer, a poem or a gun?’), or a polar choice from the same scale, which may have a yes/no response.
However, as Johnson et al (1988) has pointed out, ipsative forced-choice approaches are highly problematic. The very notion that you can ‘force’ someone to do something denudes them of free will and the very real problems of respondents either second-guessing or making a random choice from a set of items amongst which they have no clear preference. Martin et al (1995) have shown that test takers with a good insight into job needs can provide realistic faked responses. Ipsative methods still persist, in particular where sound alternatives are not available, for example the Zuckerman, Eysenck & Eysenck (1978) scale of sensation-seeking is still in used, despite the report by Ridgeway & Russell (1980) on unacceptably low reliabilities for the various sub-scales.
Faking may also be reduced by use of item opacity, where the respondent does not know ‘right or wrong’ answers. For example use of Biodata approaches, where traits and historic activities have been correlated with requirements of the job in question, can offer very opaque questions (such as the WW2 discovery of the correlation between childhood flying of model aeroplanes and good pilots).
Including the candidate
Including the candidate in the assessment process can also help to reduce faking, socialising them into providing honest responses. This may be implemented, for example, in assessment centres, where they may be involved in discussions about psychometric outcomes.
Ekman, P. (1985). Telling Lies, New York: Norton
Holden R.R., Wood, L.L. and Tomashewski, L. (2001). “Do response time variations counteract the effect of faking on personality inventory validity?” Journal of Personality and Social Psychology, 81, 1, pp. 160-69
Johnson, C.E., Wood, R. & Blinkhorn, S.F. (1988). “Spuriouser and spuriouser: The use of ipsative personality tests.” Journal of Occupational Psychology. 61, pp. 153-162
Kline, L. (1993) A Social Scientist in Industry. Aldershot: Gower.
Martin, B.A. Bowen, C.C. and Hunt, S.T. (1995). “How effective are people at faking on personality questionnaires.” Personality and Individual Differences, 32. pp. 237-46
Ridgeway, D. & Russell, J. (1980). “Reliability and validity of the Sensation Seeking Scale: Psychometric problems.” Journal of Consultative Clinical Psychology 48, pp. 662-664
And the big