Because there's nothing else - there's no alternative; because science has an inbuilt methodology to eliminate human bias, partiality and preference as far as possible and to ensure that its productions are, depending on context, testable, repeatable and shareable.
That's why.
Does it? Shortly after I finished my teacher training, I seem to remember that certain long-favoured educational principles were shown to have been falsified by the scientist in question (Burt(?) iirc). Then, of course, we had Kinsey and his erroneous material regarding sexuality - something that was accepted by the scientific community for some time. That's just a couple of examples which - whilst the errors have now been uncovered - were allowed to influence public practice for various periods of time.
An other problem, of course, is that when such false material is accepted, 'testable, repeatable and shareable' becomes obsolete. How long does it take to iron out 'tested, repeatable and shareable' outcomes that have been based on such erroneous material?
Finally, for now, is whilst a lot of scientific material is based on observation and other important and valid ideas, they are often based on an element of guesswork. Think, for instance, of the idea of carbon-dating. This is fairly accurate up to the last 50,00 years, but beyond that period of time is pretty useless. With a planet that is possibly as old as 4.5 billion years old, 50K is a pretty insignificant period of time.
