There's a fuzzy gray line here. Whether formalized or not, the way I see it, pilots will de facto be penalized for making decisions that cost an airline a ton of money, especially if it turns out to not have been an actual emergency.

I have no idea what legislation is in place regarding this topic and protecting pilots. But even with really good protections, there's no way that it won't affect your social standing with the higher ups if you divert and cost the company exorbitantly.

So pilots do have to make judgment calls about how much they're willing to risk their lives and the lives of their passengers versus a perceived risk, and worse, this judgment call exists in a weird fuzzy area where they're "allowed" to make the safe bet, but in practice, they will be punished for it.

Software doesn't (generally) have this luxury. It decides whether to land or continue flying based on some defined set of conditions.

It can still decide to risk the lives of the passengers. But that decision has to be approved by the airline, and can't live as an informal "you will lose brownie points with management" sword over the necks of the flight crew.

Unless we start to incorporate LLMs and probabilistic models into autopilots... but no-one would ever do that, right? ....