H.S.,
I disagree with some of your claims, but I agree with others. Here is what I concluded by myself so far:
(1) Machines can never have ethics in the normal sense because they lack two abilities that all animals have: the ability to feel pain, and susceptibility to death.
(2) Any application of formal systems to ethics must use some form of uncertainty (e.g., probability, likelihood, fuzziness) to compensate for modeling real life, which is fuzzy and error-prone. In other words, the formal system must use some form of heuristics, which inherently means that it cannot be pushed to extremes such as applying to every situation or creating very long chains of reasoning. There will always be exceptions, and there will always be factors not considered.
(3) The study of ethics arises naturally from wisdom. Unfortunately, wisdom is undefined, although I believe I'm on the track of creating an excellent definition of wisdom that can even allow wisdom to be measured.
(4) Many aphorisms about ethics are seriously flawed, even by definition, and can be shown to be flawed via an appropriate model. One such aphorism is "The end justifies the means," which disregards which people and how many of those people benefit, and in which way.
(5) Much of the study of ethics will revolve around simulation, such as simulating behavior using heuristics from psychology.
(6) Animals (including humans) are networked in different ways--they have different dependencies--so those dependencies must be modeled to get a good understanding of overall cause-and-effect (requiring simulation again). In other words, the animals involved are not independent variables, but dependent variables.
(7) Time evolution must be modeled, especially if-then conditions. Therefore steady-state (long-time) solutions will be of high importance. An example of such an if-then situation is "If you kill me, then my hostages will die because nobody else knows where they're hidden."
(
Lack of knowledge will likely also need to be considered, similar to how Dempster-Shafer Theory handles it (
https://en.wikipedia.org/wiki/Dempster%E2%80%93Shafer_theory).
Yesterday while killing time at a bus stop I began piecing together a formal model of ethics, based on math and directed graphs and simulations. I might be persuaded to create a thread in the Projects section on my progress on this. However, I am very busy on much more important matters, so unless someone wants to collaborate on a mathematical ethics project with me, it won't be worth my time to work on this additional, low-priority project just so a lot of non-contributors can be dully amused for a few minutes each day. It might also be useful to have someone throw out a number of aphorisms and situations to use to test my model and to force it to consider more types of situations.