> [!tldr] If a hypothesis, and how it will be tested, are Being specific about what your are measuring and your hypothesis **going in** gives you the best perspective. If your hypothesis is vague, you will tend to look for ways it could be phrased and interpretations of the data that might support it. Otherwise you won't question your incoming assumptions enough when you get the data, instead you'll tend to try to make the data fit your assumptions. Scientists given too many degrees of freedom can always find some sort of chance correlation in data, especially in the age of [[Big Data]] - [[A 1-in-a-billion kind of day happens to 8 people every day.|Rare isn't rare.]]. This feels related to how [[Black and White Goals]] can be effective. [[Failure is most useful when you give your best effort.]] If your goals aren't black and white, that (in a way) translates to not giving it # Drug Studies In 1990 or 2000 (can't remember which) the "hit rate" for successful drug trials for all sorts of conditions fell off a cliff. It wasn't that we stopped having good ideas and figuring things out - it was that we started **having researchers explain in advance how they would test their hypothesis**. This took away the ability (or, perhaps, [[Incentives|Incentive]]) to use the greyness of the hypothesis to find ways where the experiment succeeded all along. **** # More - [[Black and White Goals]] ## Source - [[Inside the Box]]