There are all these hard-to-evaluate pros and cons in the outcome-prediction process, but at the end of the day the pros seem to add up to a bit more than the cons.

An interesting topic and I disagree.

The problem is that some people see the question as symmetric: either we bail out the banks or we don't. No reason to prefer a priori one course of action over the other. But my whole point is that we should prefer the simpler action a priori, and require much more concrete evidence in favor of the complex plan before we choose it.

I read your example and thought {X, X + bail out banks}. But since you are a bailout advocate I suppose you wouldn't have brought up the example if you thought the complexity weighed in against your position. So I infer that you mean {bail out banks, do a more complex thing Y}. For many instances of Y that people are likely to propose I expect I would agree with your judgement, with the complexity being a significant factor.