I'd also be worried about the model being biased. For example, perhaps the model is "better" because it uses correlated features it shouldnt be using (due to fa

I'd also be worried about the model being biased. For example, perhaps the model is "better" because it uses correlated features it shouldnt be using (due to fairness laws, etc.)