Well, the facts of the example you gave include flipping a switch, and a correlated light bulb glowing. The conclusion that a "light fairy" is the underlying cause is IN ABSOLUTELY NO WAY supported by the two starting pieces of information. The correct conclusion is simply that a correlation exists between the state of the switch and the state of the light. In no way do I need to know HOW they are correlated or if there is causation. Making up theories about such things is not a valid, which is why yours is a poor example.
I've pointed this out several times now, and you continue to make up these non-sequitur examples, completely unable to refute my examples of drawing conclusions without any knowledge of causation:
Flipping a light switch and a light bulb going on
Using a remote to turn on a TV
Driving a car
Reading the time on a watch
You don't need to know how the physics of time work, nor how the watch works. Clearly your watch is not the causative agent of time either. And yet you're still able to draw perfectly valid conclusions about the time WITHOUT knowing those things. Clearly you are still able to drive a car WITHOUT knowing how it works. Clearly you don't need to open the tank of every toilet to figure out that pushing the handle will cause it to flush.
The idea that we can't draw any conclusions without understanding the underlying mechanisms of systems is just foolish.
I never said we cannot draw conclusions, I just insist they are not necessarily valid unless we know more about the underlying causes than simple observation tells us. I think we essentially agree on this, we are just saying the same thing in different ways.
No! I have never once said that about evolution or anything else! Where are you getting this from? The point has and still is: the results are still just as valid regardless of the path it took to get those results. If bacteria are exposed to antibiotics and don't die, it can be validly concluded that they are resistant, regardless of the mutations or mechanisms used to survive that antibiotic, because natural selection just doesn't care.
My mistake. We are getting bogged down in your insistence that conclusions are possible, which they are, and my insistence that only valid conclusions matter. If we do not understand what is happening it is entirely possible to draw the wrong conclusions, and even miss the correlations entirely.
It is enitrely possible to look at data, not completely understand it, and draw the wrong correlation and conclusion.
False Correlations
NO! You're still not understanding the difference. You're trying to draw a conclusion about the causation based on the correlation, which IS NOT VALID. That DOES NOT MEAN that no valid conclusions can be drawn about the data. It ONLY means that conclusions regarding the causation cannot be necessarily drawn. Conclusions regarding the correlation can readily drawn.
For example: Let's say an inheritable disease runs in your family, and you know you have a 50% chance of having it. To be sure, you go see a doctor, who does a blood test, and it comes back positive. It is known that the test doesn't pick up all the people who have the disease, but when it does come back positive, it is accurate 100% of the time. Now you know nothing about the disease process. In fact, the protein that was tested may have absolutely nothing to do with the symptoms of the disease, but is just a random finding that's always associated with it. Should you then say "I can't draw a valid conclusion whether or not I have the disease because I don't know the causative agents or the process"? NO! And this has nothing to do with your non sequitur conclusions about the disease's onset, or how hard each symptom will hit you, or ANYTHING ELSE.
Here is where we have a problem.
I have been wracking my brain the last couple fo days to remember where I read bout the phenomena called voodoo correlation. (To be honest, I was trying to remember the term.)
Brain imaging studies under fire : Nature News
While this is not proof of anything whatsoever, it does indicate the danger of making correlations based on insufficient data. Let us go back to your example of the light and the switch.
Every time you walk into a room you flip the switch to turn on the light, and the light comes on. You therefore conclude that flipping the switch turns on the light. What you do not know is that I have installed a pressure mat directly under the switch that actually turns on the light and all that the only reason the switch is there is I have not yet had time to remove it, or to wire anything else into it. Your action of standing next to the switch actually turns the light on, but your habit is to flip the switch, and that is reinforced because the light always comes on as you are flipping the switch.
Observed data leads you to an entirely false, yet seemingly valid, conclusion.
We know what effect the pencil will have, yes. Why? Because it has a specific, non-random effect. The effect certainly changes based on the orientation of the pencil, but for any given orientation, there is one resultant effect on the greater shape, the "big picture". Thus, the big picture is affected in non-random ways based on the filter, despite the non-deterministic location of the remaining pixels.
So too is evolution: The pixels of your fractal are the genotypes, constantly worming around in a random manner. The pencil is natural selection, completely knocking out genotypes which lack survivability for that filter, and leaving specific non-random genotypes to remain. So for any given environmental pressure, a non-random big picture emerges despite variability within it.
Why do the random genotypes suddenly become non random? A predictable effect does not make the result non random.
If I plant the seeds of a Macintosh apple most people would expect me to get a tree that produces Macintosh apples, but I will not do so. The only way to actually get Macintosh apples is to graft a branch of a Macintosh tree onto another apple tree. The non random result will me give me Macintosh apples, but the actually randomness of the process prohibits me from knowing exactly how that branch will grow.
You insist on looking at the big picture, which is survival, and insisting that the process of survival is irrelevant. If evolution is not random, why do some flatfish have both eyes on the right side, and some on the left? They both obviously survive, your non random result, but the process is totally random.
Wait a minute now. Did you or did you not say that things "will average out in the long run" despite local trends? Well? Will things average out based on the given probabilities or won't they?
Do casinos sometimes lose money at all their games combined some years? Is it really a random outcome that casinos generally come out ahead?
Yes I did, but a poker game does not last forever. Both players have a limited starting stake, thus making it possible for random luck to make one the winner. Otherwise two equally qualified players would always end up in a draw, like playing Tic Tac Toe.
The reason casinos come out ahead is they stack the odds as much as possible in their favor. The best example is roulette, where a win pays 35 to 1 even though there are actually 37 spaces the ball can land in. Another tactic they have recently developed is not paying the big jackpots on their progressive slot machines because they claim there was a computer error, another reason not to play slots. In other games they just take a share of the pot, thus ensuring that they always make money.
Which is why most casinos frown on card counters, and make the job of counting cards as difficult as possible for those who do so, despite the fact that it is entirely legal to do so. Blackjack is the only game in a casino where a smart player can actually turn the odds in his favor.
Is the overall result of the house winning random? In the truest sense of the word, yes. I would also point out that how much they clear on a given day is entirely dependent on how lucky their guests are.