This is in line with the "tree falls in a forest" example as well - the "paradox" is not in whether the tree falls, but whether we would consider it sound. The solution is to be more precise about defining "sound", as either "vibration of air molecules" or "vibration of someone's ear drum". Similarly, the "are viruses alive" question is pushing us for a more precise definition of "alive", which we can then apply to (for example) how we might classify self-sustaining chains of chemical reactions on alien planets.
As for the tree in the woods, again this is a toy example. It's a vastly simplified example that philosophers use to discuss epistemology because it provides a simple basis on which to argue about something. In math and engineering they use the spherical cow in a friction-less world. Obviously no one cares about a spherical cow but it's useful as an isolation tool so that you can really work on just the problem at hand.
I don't know a lot about string theory but if it cannot be tested and is still considered science this is indeed very troublesome for many theories of science. Perhaps this ties into the realism / anti-realism debate where philosophers debate whether there are actually strings or they are just convenient mathematical constructs with wide explanatory power. I personally am somewhat of a nihilist on this point and I don't think it's appropriate to bend the concept of reality to apply to things like strings and that this is basically just an incoherent exercise to begin with.
A non-string example of this debate that I know a little bit more about (but not much) is the debate over observability. I can observe the wall in front of me unaided with my own eyes and sense of touch.
We can also "observe" radioactive decay in a gas chamber by examining the condensation trails of particles traveling through the gas, but are we really observing them? This seems to be more indirect than the first example, so can we really know that there are particles traveling through the gas? If indirect evidence is not acceptable in science, what is the cutoff point that it becomes unacceptable. I could hear from a friend of a friend of a friend that they saw bigfoot and while everyone might trust everyone else this most certainly is not a scientific observation. This latter example is extreme but it demonstrates the importance that scientists understand the scientific method and the philosophy behind it.
If string theory can only be proven indirectly, at what level of indirection does it become inappropriate to say that the observations are evidence for the theory. It sounds as if we only have very indirect evidence for string theory which is probably why it is so controversial.
The lung cancer thing, I'm not sure I follow. Is this simply not statistics issue combined with a lack of knowledge about the human body?
I think an interesting distinction can be made between concepts that are linked more to the actual things covered by them and as we find out more about those things, the definition text gets updated, vs. concepts where entities can move in and out of its coverage as we learn more about them.
In case of life, you may say at the outset that dogs, humans, birds and trees etc. are alive, but rocks, clouds etc. aren't. Then your task essentially becomes like a machine learning algorithm's: you get a training set and then you have to make a model, a decision surface in some feature space that separates the yes from the no examples well enough, while still fulfilling certain smoothness etc. criteria. Of course we never do such things actually, but it's related.
When you get a new example, like a virus or prion, you have to decide on which side it falls.
The interesting question is: what happens if it turns out that one of your training examples were represented with some erroneous features? Or what if we discover some new features that could be relevant?
Do we still label those examples as animate/inanimate and update our decision surface accordingly, even if it makes the surface quite complicated? Or do we relabel them to so that we can keep the decision surface simple? Or do we keep its label and rather relabel some of its neighbors too, to make the decision surface simpler?
These are rhetorical questions, I'm trying to show how arbitrary the whole thing is. There is no "One True" label until we decide it and it's actually like an engineering trade-off decision between matching tradition/intuition and the simplicity of the definition. The former is like the training set error in machine learning, the latter is like regularization.
I think learning about CS and programming would be beneficial to philosophers as well as the other way around. Due to the required unambiguity of computer programs, many times programmers and CS people had to tackle issues like this. Philosophy is largely written in natural language, which results in a lot of ambiguity. For example, read the source code of a simple Quine (a program that prints out its own source code). You'll see code, code written in quote marks, things in quote marks escaped and nested into other other quote marks etc. I mean CS has developed concepts like currying, or variable scoping, which are very much related to philosophical issues. For example when you create and event handler, and say onMouseClick = function(){print x;} what do you mean by x, the current value of it, or the value when the event happens? Do you evaluate it now, or postpone it until the event? In natural language both sound the same, "you print x". But do you mean x as a symbol or do you actually mean the thing pointed at? Like asking "Do you think the president of the US will die in 2050?" can mean whether Obama will die in 2050, or the president incumbent in 2050.
The similarity of this to the "life" issue is whether we freeze the current meaning of "life", or we allow it to change.
So I think a lot of this confusion is just the result of not having to face these distinctions in natural language, while CS people have sorted out many of these things conceptually, in things like reification, reflection, virtualization etc. Of course it's not new, the "use-mention distinction" is well-known from earlier, but programming makes it really straightforward and obvious.
Who cares if we say "a virus is [not] alive"? Obviously that's predicated on the definition of alive. It doesn't change what a virus is, and certainly doesn't change our understanding of viruses. It seems like the only time that kind of thing matters is if you've got some existing issue with "alive" (like, say, a law granting rights to anything "alive") and now need to fight it out to retrofit something.
I'm thinking of "Dissolving the question"[1] and the infamous "If a tree falls in a forest a but nothing hears it, does it make a sound?" The fact that this generated any serious question/answer is absurd. Everyone agrees on what happens. They just like to argue over what "sound" really means. Again, this is only an issue if you're e.g. at court resolving a poorly written noise violation law. Yet "If a tree falls" isn't apparently used as a quick lesson in definitions, but apparently treated like it holds something interesting.
It might be fun to argue stuff. Like, is vim truly an IDE if I load it up right? Doesn't really change anything, but it might be fun. I get the feeling a lot of philosophy consists of people chasing this. If it's not to judge merits or validity, and is just passive, then perhaps that explains it.
On physics and using Bayes versus Scientific Method, I was thinking more about string theory (and I know nothing of this at all), where, since we can't test it at all, it's not proper science. Yet using a Bayesian model, we're free to add evidence apart from verifying experiments. If, for example, I come up with some simple rules for how the universe works, and out of those arise a huge amount of known equations, there may be no way to test if my rules are true and worth studying more on. It'd violate the scientific method, since there's no testing. Yet the mere fact that that "one little trick" explains a bunch of known physics is a huge amount of positive evidence by itself.
Again, thanks for taking the time to write out your reply; I did appreciate it.
1: http://lesswrong.com/lw/of/dissolving_the_question/