[jcahill]

That wouldn't work like you're hoping. You're assuming a (representative) respondent pool that's all of these:

1. capable of switching off preference falsification for a survey

2. introspectively active (and accurately so) over a long enough period to notice that they behave differentially in public and private contexts

3. interested in exposing this degree of candor for a survey, i.e. motivated to "show the researchers" (show themselves, really) their belly after collapsing across all the reasons they'd rather not.

Humans are primitive. Questionnaires have to be designed around that.