What counts as a competent experiment or a successful replication is settled by agreement among scientists, not read off nature. That agreement cannot be separated from the results themselves. Disputed science stays disputed precisely while no shared standard of competence exists.
To know whether an experiment was done properly, you need to know the right answer, but to know the right answer, you need a properly done experiment. When results conflict, evidence alone cannot break this circle. Someone has to judge which experiments to trust.
A researcher who strongly believes a new theory reads ambiguous results as confirmation. Those results then strengthen the belief, which shapes how the next results are read. The loop can feel like progress when it is really a self-sealing circle.
Scientific disputes are often closed not by a knockdown fact but by persuasion, ridicule, political maneuvering, and recruiting powerful allies. Sometimes they end simply because the minority side dies off or is outnumbered. The winning side is then remembered as having been proved right all along.
Science is powerful but does not know the limits of its own strength or ignorance, and its errors come from well-meaning clumsiness rather than fraud. Because its mistakes are human mistakes, it should be neither worshipped as infallible nor blamed as wicked. Treating a complex, uncertain system as fully reliable, as if it were as safe as everyday technology, is itself irresponsible.