What are the reasons why a hypothesis needs to be testable?

Any science based on hypotheses inherits myriad flaws. These flaws include the idea that hypotheses, by design, drastically reduce the field and scope of exploration. In doing so, they fatally skew experimental results, so much so that they guarantee these results will translate poorly to individual, real world cases.

A separate but equally problematic assumption is modern science’s willingness to treat philosophical concepts (e.g. the idea of string theory, the idea of a big bang, even the idea of “energy”) as equal to actual, physical science. Using hypotheses which are testable limits this problem. These tests root science in the physical world and in doing so, make science more honest.

Of course, being able to test a hypothesis then introduces its own shortcomings. For instance, according to David Hume, we can never design a science which renders a certain outcome. By this, he means a test which perfectly marries ideas to the physical world; the mind to the body, if you will. At best, we can assume strongly correlated outcomes equate to highly predictive probabilities. And this does then allow us to continue to do science. But we can never actually observe “cause and effect.” Post hoc, ergo prompter hoc. We can only act as if we can see it.

Ignoring this limit leads to scientific arrogance. Nonetheless, well designed hypotheses will always be rooted in testable science.