Recently, other blogs have attempted to define faith and scientific knowledge in ways that demonstrate the two are not the same. (See "You don't need faith to believe the principle of evolution") An example of how scientific principles are developed occurred to me just a few minutes ago, which I thought I'd share.
We begin with a proposition, principle, theory, assumption, or even a belief if you like, since we'll assume the speaker believes it: "Whatever one puts into a refrigerator will be there when one opens it again."
This statement would be an excellent hypothesis for a scientific investigation because it is readily testable. One can open the door, and see that the items are either there or not. The hypothesis is readily generalizable as well, one can test with a variety of items, a variety of refrigerators, and a variety of individuals. Several experiments can be done to test for all sorts of conditions and combinations.
Now let's say the hypothesis is shown to be wrong in at least one experiment. The refrigerator is opened and a specific item is missing. Possible hypotheses might include the item turning invisible or disappearing on its own, but these would require significant alterations of well established principles. After doing some investigative work, we might determine that a flatmate removed the item. When more items go missing, we might discover that every time it was another human agent who removed the item and that they always removed it between our placing the item in the refrigerator and opening it again (assuming, for sake of argument, that we have extraordinarily honest or sloppy flatmates). Eventually we will amend our hypothesis: "Whatever one puts into a refrigerator will be there when one opens it again unless an entity has removed it during the interval."
Now suppose we place green leaf lettuce in the refrigerator and forget about it for a month. When we open the vegetable drawer again, we will find black spotted and wilting lettuce instead of the delicious and crisp lettuce that we placed there. More experiments will demonstrate that many items left unattended for long lengths of time will slowly undergo change and that the rate of change is dependent on the temperature setting and type of item. We can amend our hypothesis again: "Whatever one puts into a refrigerator will be there when one opens it again unless an agent has removed it during the interval. Items are subject to continual degradation depending on conditions including temperature, the type of item, and the air-tightness of the item's container."
This theorizing sounds painfully obvious to adults, but remember that most humans learn this theory of conservation at a young age. You can actual watch children progress through stages when first they realize that objects are permanent (do not disappear when they cannot be seen) and later that the volume of a liquid does not change with the shape of its container. (Many adults still have trouble thinking in terms of conservation as the term is used in the physical sciences, but that's another matter.)
The progression of our hypothesis demonstrates why scientific theories are subject to continual revision and why a good hypothesis is one that makes predictions that can be tested. If our hypothesis had been "whatever is found in the refrigerator will be a subset of items placed therein" it would have lacked explanatory power. That is, it fails to explain the hows and whys of the changes taking place. If our original hypothesis was "Whenever we open the refrigerator, what will be inside will be what Zeus wills to be inside", we would never have professed beyond that formulation unless we were given unmitigated access to Zeus's will and then began to investigate that (but, of course, Zeus's will is beyond human comprehension). Even if our flatmate had taken some food, we could still argue that that had been Zeus's will and not need to add additional clauses. Or we could argue with equal validity that our flatmate had violated Zeus's will and then feel justified in taking punitive measures.
This is the problem many of us have with beliefs held on faith alone. The religious begin with a proposition that they accept on faith (Yeshua was the legitimate messiah, Paul was divinely inspired, Mohammad was divinely inspired, the Hadith of the Cloak is valid, Zeus causes lightning, Kuan Yin refused to enter Nirvana). When required, evidence is selected to prove the statement of faith and contrary evidence is explained away (Zeus willed Melissa to take the last pita. Zeus only wills milk to spoil after a few weeks.). Very rarely will the proposition itself be modified (e.g., the emergence of Deists who believed Jesus was wise but not a miracle-worker).
When some IDers/Creationists argue against evolution, they often display a lack of understanding about the difference between a hypothesis and a statement of faith. No living biologist would expect everything Darwin wrote about evolution to be true. Scientists do not end debates by quoting him (though they make look toward his writings in search of inspiration or to wonder at how much he predicted on weak evidence was actually proven with the developments of genetics), nor are there active schools of philosophers debating the proper interpretation of his writings. If a fossil or living creature was found that was half-dog/half-grass and did not fit into our current biological classification morphologically or genetically, then the current theory of evolution would be thrown for a loop. Its possible that explanations could be found that explained the odd hybrid and preserved much of the current theory, but it would not be without considerable effort, testing, and modification of our knowledge about the process we call evolution.
On the other hand, religious faith is not subject to the same kind of revision (though it certainly changes, those changes often resemble other cultural changes and not changes in scientific knowledge). If a religion predicted that it was the only valid method for approaching an absolute source of morality, one might make the prediction that its adherents should be more moral by their own standards than adherents of other religions. If that observation was not born out in the data, the faithful could invent all sorts of justifications that may or may not have been elements of the original faith. From Christianity, one frequently hears that even Christians are sinners and therefore won't necessarily be more moral, but still insist that they have personally felt the holy spirit's touch at important moments or that atheists are by nature immoral.
The same is true for historical events that the religious accept on faith. That there is no independent evidence corroborating the story would not shatter their faith. They would argue (quite logically) that one cannot prove a negative in such a situation. But they might hypocritically deny similar stories originating in other religions have occurred as described on the basis of "common sense" or their faith.
This is the point where those who approach the world from a non-religious perspective reach impasses when we argue with the religious. A proposition held on faith and not subject to revision or review based on evidence is a proposition that is difficult to disprove. One can argue that the proposition is unlikely or unnecessary to explain the data (as most "evangelical" atheists are content to do), but is not the same thing and the faithful know it. When the interpretation of Zeus's Will is subject to extreme disagreements, it becomes even more difficult for the non-religious to suggests tests of those principles as there is likely to be some group that disagrees with the interpretation selected for testing.
In the scientific community, when two groups support theories that contradict one another, the outcome is determined by the strength of the theory in predicting new evidence. More than occasionally scientific communities will adopt theories that turn out to be erroneous, but it is most often those same scientific communities who discover the faults in the theory. Among the faithful, beliefs that contradict each other are typically determined by the strength of each side in converting others, military or colonial campaigns, trade radiating from key economic or cultural centers, and other means of cultural dispersion. Religious faiths have had difficulty in spreading when there is no social or economic advantage to their adoption. (To be fair, the adoption of science as a methodology frequently spreads in a similar fashion, but because science is based on logic applied to the natural world, there is nothing preventing scientific methodology from being developed twice except time and the dwindling numbers of pre-industrial societies. It is less likely that a religion identical to Christianity would develop independent of old-world denominations.)
Where the faithful (sometimes) find fault in scientific principles is that they are always an incomplete picture. In the real world, we would have to make an assumption that a human agent stole our food because human agents are not always honest. When it comes to matters like this, scientists must assume what Dr. House is fond of arguing: patients lie, but symptoms never do. This is why scientists in many fields often seek a variety of lines of inquiry into a problem to determine independently that some principle is true. For example, we would use video recordings to watch the refrigerator or an analysis of the kitchen trash can's contents in addition to conducting interviews of flatmates. We cannot date fossils through direct observation, but by using a wide variety of independent dating methods and finding that the majority converge at similar dates, one can establish the likeliest time frame.