Yeah, I think your comment adds some interesting ideas. Following the original chess comparisons, we could also say that new players will learn more and faster

Yeah, I think your comment adds some interesting ideas. Following the original chess comparisons, we could also say that new players will learn more and faster when given some models of what are good and bad moves, than if they need to learn from scratch with nothing else. Eventually they will outgrow that and their knowledge will exceed what can fit into a "mental model", but models are not useless. But still with chess, there's also the so called Kotov syndrome, which is "a situation when a player thinks very hard for a long time in a complicated position but does not find a clear path, then, running low on time, quickly makes a poor move, often a blunder". Overthinking, getting too deep into a certain path, can also be a problem. Intelligent people sometimes make obvious mistakes due to getting lost on deep thoughts. And this is a problem related to deep knowledge, not just models. Nothing is foolproof. What I think is productive here is to distinguish between different types of models: if a model can help you acknowledge certain problems or patterns, that can be pretty useful; but if it tries to be the all-encompassing explanation to something, then it's probably wrong and might lead you the wrong way more often than not, constraining your vision instead of expanding it. You need experience to make good decisions, not just be given a model, but experience is also not enough or not always there, and models can help fill some of those gaps too.