What is Social desirability bias?
Also called Faking good, Impression management
Social desirability bias is the tendency of respondents to answer questionnaires in the way they believe will be viewed favourably rather than accurately. It inflates scores on desirable traits, compresses the differences between people, and is strongest exactly when something depends on the answer — such as a job application.
Why it is worse in selection than in development
The same questionnaire behaves differently in the two settings. A person completing a personality inventory for their own development has little reason to distort; a person completing it as the last step before an offer has every reason to.
This is one of the strongest arguments against carrying a development instrument straight into a selection process without re-examining it. The norms, and often the item behaviour, were established in the low-stakes condition.
What actually reduces it
Forced-choice formats, where the respondent picks between equally desirable statements, reduce the room to inflate everything at once, at the cost of a more complex scoring model. Validity scales — items almost nobody can truthfully endorse, and pairs of items that ought to be answered consistently — detect it rather than prevent it.
Warning candidates that response patterns are checked has a measurable dampening effect and is also the honest thing to do. What does not work is assuming that an anonymous-feeling interface removes the incentive; the incentive is the job.
Not the same as acquiescence bias
Acquiescence is the tendency to agree with statements regardless of content, and it is countered by reverse-keyed items. Social desirability is direction-aware: the respondent is aiming at a particular impression, not simply agreeing.
How AssessAll handles it
The DISC Prism engine runs three validity checks on every sitting: consistency pairs scored for divergence, low-frequency items almost nobody truthfully endorses, and instructed-response items that ask the reader to select a specific option (lib/scoring/engines/prism.ts). Two divergent pairs, two infrequency endorsements or a single missed instruction raises a flag, and a flag lowers the confidence band the report is read with. It never auto-rejects anyone.
Read next
Related terms
More terms beginning with S
Check a selection process against the four-fifths rule
Free, no signup, computed in your browser — with the remedy, not just the verdict.
Last reviewed