Of course we should try to learn from past research.

But sometimes the lesson to be learned is that what we were is a bad approach, and we need to rethink things. We can't simply run experiments and get results. We also need to study whether our experiments are telling us what we think that they are telling us. Because if we don't know what our experiments are actually telling us, we're going to misunderstand the results.

https://people.cs.uchicago.edu/~ravenben/cargocult.html is worth a read on this. Among other examples, it includes an explanation of what actually has to be done in order for a rat running a maze to actually test the rat's memory. And how it is that we know that this is what has to be done.

The fact that psychologists ignored that experimental procedure marked psychology as a cargo cult science. (Some 40 years later, the replication crisis forced psychologists to at least talk about the issues that Feynman raised...)

The fact that we continue testing animal models of Alzheimer's that we have proven to be a bad model of humans, likewise shows that Alzheimer's research has become a cargo cult science.

>The fact that we continue testing animal models of Alzheimer's that we have proven to be a bad model of humans, likewise shows that Alzheimer's research has become a cargo cult science.

Partially. It also, more primarily, shows that medical funders and ethics committees demand very strong evidence in one animal model before they allow you to move a step "up the ladder" towards humans. What you can test on mice, you haven't managed to get approval to test in nonhuman primates. What you can test in nonhuman primates, you might get approval to test in humans. What you can test in humans, might finally prove useful for treating humans.

But the broad assumption is that if something fails in mice, we can know it would have been unethical to try it on humans first.