I’m a statistician (which is even better than a scientist ) and have some experience using the scientific method. The image of the rational scientist strictly following the evidence wherever it may lead is an ideal. But it is not the fact of practice. Data doesn’t speak for itself, we interpret it based on our preconceptions, which are based on the larger existing available set of data, prevailing theory, and our own dispositions. To illustrate this I want to bring up two concepts: Bayesian Epistemology and Thomas Kuhn’s Structure of Scientific Revolutions.
In statistics there are two different general ideologies about probability, Frequentist and Bayesian. Under the Frequentist paradigm, probability is defined simply as the number of times an event occurs in a number of trials, ie if I flip a coin 1,000 times and it comes up heads 550 times, the probability of it coming up heads is 55%. Under the Bayesian paradigm, the results of experiment are used to supplement an existing theoretical probability (called a “prior”). So I might believe based on what I know about coins that they usually come up heads 50% of the time, so after getting that result in the experiment I’d weight that result against my prior expectation, and maybe say this coin has a 52% chance of coming up heads, maybe it’s kind of a trick coin or maybe it was just dumb luck that it had 550 heads. The result of experiment adds information but doesn’t supplant theory.
Frequentism is great if I can just keep flipping the coin a thousand, a million, or a billion times. Bayesianism is great if you have limited data. Unfortunately we can’t flip Jesus a thousand times and see how many times he comes back from the dead so we need a Bayesian approach, which relies heavily on our prior expectations. Now, the example I used is mathy but a Bayesian framework is starting to become used in the ‘soft sciences’ like history.
Which leads into Kuhn, whose landmark book The Structure of Scientific Revolutions is available all over the place and can be read about on wiki.
It’s the book that coined the phrase “paradigm shift”. I can’t do the book justice, it’s not that long and is mandatory reading for anyone interested in science. I introduce it here to highlight the fact that all models are wrong, but some models are useful, and that all scientists have to deal with inexplicable data, called ‘anomalies’. Ptolemy’s geocentric model of the solar system worked. It was convoluted, but it worked. Newton’s physics works even better. Einstein’s model explains some things Newton’s model just had to shrug at, but it also has anomalies like dark matter and dark energy. All this is to say that the image of rationally pursuing the data where it leads is complicated, and doesn’t tell the whole story. There are anomalies in all worldviews because humans are imperfect and lack even the capability to understand the world perfectly.