Earlier quoted context omitted.
Designing good studies is hard. That doesn’t mean we should just roll over and be satisfied with bad studies. The same applies to most studies about human diet and nutrition.
That is why I asked for some way to test for what GP complained about, and the only answer I got was "there is no way". Do you have a way to design a study so the complaint "well, the participants could be faking it" isn't valid? I'm legitimately asking, so far all I've gotten was "no way".
There is far more incentive to conduct shoddy research in psychology (since as a corrolary to being hard to produce good evidence for, it's hard to produce good evidence against, which non-reproducability is often not treated as), than there is to only proceed with the studies you actually have sufficient known-good data to work with.
We have spent billions of dollars on single technology projects in order to answer questions about physics, like the LHC, because without those we often cannot actually get the data needed to answer certain questions.
Nothing equivalent is done in social psychology. There are no massive, multi-use infrastructures being built to engage with it.
While that is not the fault of social scientists, but of the devaluation of their field by the society around them, it still does mean they often do not have the necessary tools to do their jobs correctly.