The idea that science begins with hypotheses instead of data arose in the 19th century, first in the work of William Whewell, and was very influential on Charles Darwin. It is scientific orthodoxy today but is a flawed methodology. Here are three flaws:
- Positive bias: It’s virtually impossible to demonstrate a negative empirical conclusion because “absence of evidence is not evidence of absence”. So there’s a bias in favor of any positive hypothesis.
- Cherry-picking bias: Evidence against an hypothesis can easily be avoided, ignored, deprecated, etc., even unconsciously since the search is for positive evidence. Researchers are pressured to show results so there is pressure to ignore contrary evidence. This is a huge problem is medical science.
- Inconsistency bias: There is no requirement that hypotheses be self-consistent. Since a contradiction implies everything, a self-contradictory hypothesis explains everything and so is the ideal hypothesis. The prime example of this is evolution. What doesn’t it explain?
One supposed fix is the requirement that hypotheses be consistent with the science that is already generally accepted. The flaw in this is that what is accepted science may be flawed itself. Evolution is again the prime example.
Conclusion: Science as it is practiced today is flawed. Scientific methods today are flawed.
The fix to all these flaws would mean beginning with the widest collection and compilation of data, not hypotheses. Generalizations of data should come from the bottom-up, inductively. That is what Isaac Newton did and he was right to say “I frame no hypothesis”.