There’s much to unpack in your comments. Others have mentioned Peirce, abduction, inference to the best explanation, etc. I agree these perspectives are illuminating and would like to consider them in the context of the “gaps” mentioned in the present essay, specifically as they relate to decision-making and creativity/discovery. (Apparently, Popper was a fan of Peirce.)

Similarly on your reference to Quine, interesting from at least two vantage points: Illuminating the need to weigh hypotheses in a rigorous way (a common appeal from Bayesians), as well as the sociological aspects, the need to orient yourself to a falsification ethic (Popper’s view, anticipating Quine).

Personally, I believe explanations provide the foundation for science (following Deutsch), including the foundation for experimentation and measurement. For example, the problems with null hypothesis testing and p-hacking may be understood as consequences of explanationless testing.

That said, I agree that the psychological and sociological aspects are important (e.g. Kuhn, Lakatos) for making sense of the prevailing approach to AI and how it might move (or not move) in the future.