I have this impression that if you ask respondents in a survey, questions such as "do you like the item X/ person X" or "which one of the two variations do you like more", or "what in particular would you like about the item X/ person X", etc, you would get a lot of bias and rationalizations. That is, "say vs think vs do" would be extremely disparate.
At the same time, I would expect negative questions such as "is there something you dislike about the item X" or "could you list top three shortcomings of the item X/ person X", etc, to render quite objective responses; I would expect strong consistency between what respondents would say compared to what they actually think or feel and then compared to what they would actually do.
(Of course, I exclude some extreme cases of social desirability.)
Given that I am speaking about my subjective impressions, based on my personal limited observations, I would like to ask, is there such a pattern indeed and if so, how could it be possibly explained?