This is a really valuable essay. It is a feast when it comes to food for thought.
I have never before seen the analysis of belief in terms of Stoic psychology:
impression > assent > impulse
That triad will be worth my looking into in detail.
As someone with a background in Machine Learning and Computational Learning Theory, I have, since the 1980s, looked at the process of learning, creativity and knowledge acquisition in terms of a diad: the tradeoff between faith and skepticism.
In my experience, all Machine Learning algorithms that I am aware of are a two-step process that switches between "generate and test".
Formally, there is a space of real-world events, either seen or possible, and a space of theories, hypotheses, models, what have you. I will use the term "model". Typically, a Machine Learning system will select a model, usually based on some preliminary observations. This is hardly ever deterministic, nor is it purely random, but a mixture. For example, neural nets use backpropagation to select the new model, but the randomness is in the selection of training data to recompute the model. In something like genetic algorithms, a random population of new models is chosen using the current population as a basis.
This is the "Generate" step. Philosophically, this selection is a leap of faith. There are reasons for the choices, but there is some guesswork too. But the selection of the new model is explicitly made in the belief that it is a better model than the one you have now. The faith is in the belief that this new model has predictive value superior to the model you have currently.
It is at this point that "Test" kicks in. Philosophically, this is the application of skepticism. The faith is put to the test in the real world, and it is weighed on the scales of skepticism by the evidence. It hardly ever happens that the model is an exact match to the data. So the process of skeptical evaluation is not just a pass/fail judgment. It must needs state where the model failed, because this provides insight into the next leap of faith.
It is interesting to compare this diad to the triad of the Stoic. I suspect that the Stoic model splits out the evaluation of the evidence from the generation of the model and its being put to the test.
Most religions come a cropper not because of belief (or faith) per se, but because of a misapplication of skepticism. What happens is that, in the presence of disconfirming evidence, the religious person will reject the evidence instead of abandoning their faith. There actually are good, rational reasons to reject evidence that contradicts your faith, due to the uncertain, probabilistic nature of living in the real world. Sometimes we can't believe our lying eyes. But too often this conservative notion of clinging to our faith is too inflexible: we hold on until the point arrives where the evidence is just too much and a crisis of faith occurs. The Taoists have a metaphor for this: instead of being like the mighty oak in the forest, which stands until the storm knocks it over, it is better to be like bamboo and shift as the winds move us.
In the case of our shipowner, the problem is not the faith, for, after all, the past has validated his belief. The error instead is in the lack of skepticism - his belief was not sufficiently put to the test and the possible dangers were discounted. This lack of due diligence was the cause of the moral failure.
It is also possible, it should be noted, to have the opposite error: that of insufficient faith, which leads to an inflexible dogmatism that does not adjust to changing circumstances. But that is a separate issue.
