But first, about last night. So I went to my friends house to watch a movie, and ended up falling asleep on the couch, and waking up too late to get home on the bus, with her going to bed. So I brewed some coffee and started reading up on philosophy related to this, and wrote you. After writing that example, I decided an apple really did sound good, so I went into her kitchen to see if she had anything. Abover her counter, she had one of those baskets that hangs attached from the ceiling in the air, which happened to have only apples in it. I plucked one from the basket in the air, and ate it. I realized I would have been really uncomfortable if the apples were just floating in the air, so I was grateful for the basket. ;) (coincidence???)
But you understandably want a more structured logical approach. I sounds like you agree that a meta-narrative CAN exist where you wouldn't see evidence for it because of your assumptions, but, well, you don't see any evidence for it. Let me instead attempt an attack on the concept of epiphenomenon.
So Wikipedia states:An epiphenomenon can be an effect of primary phenomena, but cannot affect a primary phenomenon. In philosophy of mind, epiphenomenalism is the view that mental phenomena are epiphenomena
Given this, we can see that there are two kinds of knowledge. Those derived from the primary phenomenon (reality) into the secondary phenomenon (mind .. ie knowledge which is derived from observed reality), and knowledge which goes from the secondary phenomenon into itself. Kant defines these two as analytical and synthetic propositions:
http://en.wikipedia.org/wiki/Immanuel_Kant#Kant.27s_theory_of_perceptionSo for example, "Mary's poodle weighs 20 pounds" is synthetic proposition because it gives information about reality (primary phenomenon) into secondary phenomenon (mind), but the "Mary's poodle is a dog" is an analytical proposition, because a poodle is by definition a dog, so no information is taken from reality that wasn't already in the secondary phenomenon.
The interesting thing, while I was reading about this last night, was how Kant bends over backwards to define mathematics as synthetic, when its so obviously is analytical. (all poodles are dogs, 'Mary's pet is a poodle, therefore Mary's pet is a dog' is obviously a mathematical statement from first-order logic) But lets model epiphenomenalism and talk about probability, shall we?
So we have our model of primary phenomenon, physical reality, which we will model with an array of numbers representing different things in different states. Then we have our epiphenomibot, which has a smaller array for representing bits of reality, and can accurately represent a subset of reality. It has two functions: observe reality (input synthetic proposition), where it sets its internal arrays to match some subset of the reality array, and think (input/create analytical proposition), where it rearranges its internal array in some random way. Conforming with epiphomenalism, its observations have a probability of 1 of corresponding to some subset of reality, (ie it accurately observes some subset of the bigger array). And also its random internal rearrangements have no correspondence with reality, they are random, daydreams if you will.
So this works fine and well with ephiphenomemalism, but doesn't model our reality. The clearest couter example of of this is in mathematics, which is clearly analytical, occuring inside our minds, but also has the power to predict real things like the movements of heavenly bodies, so there is clearly probabilistic correspondence with what a physicist predicts will happen based on calculations, and what actually does happen. This seemingly violates epiphenomenalism as defined in our model, so we have to revisit our model.
NOW two things have to happen. Our reality array has to have observable patterns in it, and our physicist epiphenomibot has his think function redefined so that he can deduce patterns from the subset of the reality he can observe, and use them to accurately predict things he has not yet actually observed. Once we have this in place, we need to check to make sure our model conforms to epiphenomenalism by making sure our robot isn't creating reality with his assumptions. This is done simply enough, because we created the reality array before hand, we need merely look at it and make sure the predictions are conforming to it and not changing it. Viola! Epiphenomealism modeled.
Now the last step is simply to assume that we to are epiphenomibots. For it to be known that our reality is conforming to epiphenomenalism, we need mere follow the last step from above, which is to say that our creator/programmer, who is also the all knower of the reality array of which we observe only a subset, must confirm that our predictions conform to his predefined reality rather than creating it.
Wait, WHO???
That's right, our creator. God. This is the logical problem with epiphenomenalism. If you look carefully, you'll notice that an epiphenomibot
by definition can only perceive a subset of reality, make predictions and see if they are accurate. It has no way of knowing whether its predictions were passive observations of primary phenomenon, because
it cannot access the primary phenomenon, it's can't see the original "reality array" in entirety or it would be omniscient, violating its definition. Yet this is the only way of proving that's its predictions aren't creating/effecting reality, changing the array as it goes. Therefore the only being that can KNOW epitphenomenalism is correct is omniscient, i.e. God. Does that mean its wrong? Not at all. For all I know there may be some programmer watching us both in amusement who knows our reality array. What I'm saying though is that you've got your faith, I've got mine. :P
(note: my argument rests on a conjecture, that it is possible to build an "array like object" which provides a "subset to itself" to an epiphenomibot, and then responds to predictions from the epiphenomibot as "right" or "wrong" in such a way that the epiphenomibot can never determine whether the array is static and pre-defined, or making things up as it goes along. I haven't proven this conjecture, and if its wrong, my whole argument is also wrong.)