> These people obviously know that their studies can't be replicated

I don't think that's fair to say at all. I can come up with several ways a study can non-maliciously arrive at a false conclusion:

1. The researchers may accidentally leak information to the participants regarding the hypothesis being studied. When using human subjects, it's very difficult to avoid biasing them toward results they may think you're looking for.

2. Researchers may not sufficiently blind themselves during the experiment, causing them to have undue influence on the outcome, even if they're not consciously trying to exert it.

3. Some unknown confounding effect may be at play during the experiment that wasn't properly accounted and controlled for.

Those are just three I could think of. When dealing with behavioral sciences, I can't imagine how difficult it must be to design a test protocol that eliminates all the messiness of the meatbags being studied.