Going for a consistent set of principles seems to be really about:
- “Scientifically” imagining scenarios where we are clear about right and wrong courses of action, for example torturing an innocent person is wrong, giving a meal to a starving person is right.
- Narrowing this down by deciding where we stand on more difficult choices.
- Determining a feature-set of each group (right actions and wrong actions) from which we generate principles which would apply to newly imagined scenarios
So a bottom-up method to design principles which then work top-down.