The fields where nobody can become properly expert

Some questions are structurally hostile to skill, and the honest response is to lower expectations on everyone.

There are domains where the feedback conditions cannot be met, no matter how diligent the people working in them are. Predictions about complex systems many years out are the clearest case: by the time anything can be checked, the question has changed, the surrounding conditions have moved, and dozens of other causes have entered the picture. A practitioner in such a field may work carefully for decades and still never receive a single clean correction. This is not a moral failing of anyone involved. It is a property of the subject matter.

The right response is not to dismiss these fields. Careful thinking about hard questions is worth far more than careless thinking about them, and the people doing it are usually the first to say how uncertain the ground is. The right response is to lower your expectations uniformly, including for the most impressive-sounding voice available, and to be suspicious specifically of confidence, since confidence is the one thing the domain cannot have earned. In slow-feedback territory, someone who is sure stands out for the wrong reason.

It also helps to separate the two things such a person offers. One is a framework: a way of breaking the problem up, the considerations that matter, the failure modes others have hit. That can be genuinely valuable and is transferable to you. The other is a conclusion, a specific claim about how things will go. The framework can be excellent while the conclusion is worth very little. Take the first and hold the second loosely, and say so out loud, because the person offering it usually knows this better than their audience does.