Against the Grain

Even the philosophy of science can be replaced with a theory of learning, which will replace "The Structure of Scientific Revolutions." On one hand, science is not a specific, timeless logical inference, and on the other hand, it is not a collection of incomparable paradigms. On the contrary, it is an evolutionary development, meaning a learning system over time, which adapts to reality on one hand, and on the other hand, increases in its level of complexity. Mammals are not the paradigm that comes after reptiles, but are built upon them, and it was not possible to skip directly from amphibians to mammals, and their adaptation to reality is not arbitrary - but also not strict. All this is in accordance with deep learning, where there are many layers to each paradigm

Occam's razor is minimalist learning, tight around all positive examples - and Popper is maximalist learning, tight around all negative examples. The truth is in the middle, according to SVM (source)

Because of Occam's razor and Popper, science tends to have a minimalist bias compared to reality. In other words - there will be an underestimation of culture in archaeology (compared to the Bible, and this is a good example of a case where there happens to be a text), an underestimation of the abilities of children and animals, or of the progress of early humans (it will always turn out in history that things happened before the current estimate), an underestimation of the complexity of processes, and laws of nature, etc. That is, perhaps didactically it is beneficial for science to progress this way, but from a Bayesian perspective, there is a bias from the start to think that everything is simpler and more basic than it really is.

Therefore, the philosophy of science is a philosophy of learning about the world, and not of knowing reality (epistemology). That is, science is what is learned about the world and not the best estimate and hypothesis about the world. Because in learning, stages are needed and each stage needs to be minimal, while reality is almost never tight to the minimal, but between the minimal hypothesis that fits the data and the maximal hypothesis, and there are always missing data that complicate things.

Therefore, it's not paradigm shifts - but learning paradigms in science. The new paradigms are built on the previous ones. It was not possible to reach Einstein before Newton, and Newton has a place as a necessary stage in learning. Paradigms do not replace one another like fashions but are built on each other, expanding or opposing for example, and in any case, the reasonable thing to assume is that every scientific explanation given is partial and too simple compared to the truth, and so is every hypothesis. And so is every legal or psychological explanation for someone's behavior.

It's always a first approximation of learning, and then the truth is in the third or tenth approximation. That is, law or psychology are not human sciences in the sense of knowledge, but in the sense of learning about humans. Learning is based on what is known and not knowledge itself. Therefore, there is a big difference between law and truth, because of procedural learning methods - and procedural learning. And so in psychology, it is learning of self-perception, and perceptions are certainly limited because many people have limited perceptual ability intellectually, limited to images for example. And therefore, it is mostly image-based learning.

Therefore, if the first archaeological evidence for something is from a certain year - it is likely that it existed long before then. And therefore, there is always a tendency in science to see the past as more primitive than it actually was. Philosophy of science needs to understand that science is a learning system, and not a system of knowing or perceiving reality, and therefore it has its biases as a learning system. Just as art has a bias towards over-meaning of past works, because it is an interpretive system, and so do literature and religion, through the "assumption of genius" and the myth of the "exceptional person" or "divine revelation". So too, science has a tendency to underestimate past capabilities, because material evidence must remain in order to claim something, and because of Occam's bias. It is likely that the simplest explanation is not the correct explanation, because it is likely that we do not have all the data.

And if there is an explanation where the simplicity bias reaches absurdity, it is the philosophical explanation. In fact, this might be the definition of philosophy. The learning that has the greatest bias towards generalization. And in this sense, art is the learning that has the greatest bias towards lack of generalization. And the truth - is in the middle.