This is a valid point. At this point I'm merely trying to establish an agreement on what science is from a basic perspective of man without bias or leaning towards either the quixotic or the mundane.
I don't want answers like "Oh, Science is wonderful, the answers to all man's problems!" or "Science is evil, and the source of all man's problems!" I want something like "Science is a formal excursion of mankind to figure stuff out." If we can get to that point we can quickly move on. If we are stumbling on either of the other two we have more work to do.
In essence, science is an approach to figuring things out. It is a process. That process consists, again in general, of making observations, forming testable explanations for the observations made, coming up with a prediction from the hypotheses and testing that prediction via observation. If the prediction is valid, then make another prediction and test it also. If it is not valid, form a new hypothesis and make a prediction from that and test it. Rinse and repeat.
The crucial difference between the scientific method and, say, philosophy, is that science requires testability before full adoption of an idea (although ideas can be accepted tentatively prior to testing). Also, dispute resolution (if two hypotheses are proposed) via finding predictions that are different and then performing an observation to see which one is correct, thereby eliminating the wrong idea.
Another aspect of the scientific method is that all ideas are subject to testing. There are no 'sacred' ideas. Now, any idea that has undergone extensive testing over many observations and a lot of people looking at it is unlikely to be simply thrown out, but if counter information is found and accumulates enough, the viewpoints will change.
Now, in practice, it is never quite this simple. Observations perform both the role of initial observation and of testing. Hypotheses are often modified slightly to accommodate observations as opposed to having a massive overhaul. There is also the issue of how observation and theory interact, which most observations relying on previously tested theory.
Another aspect of the history of science should be noted: since older hypotheses have been tested, the new hypotheses tend to give the same answers in those topics where the old hypothesis worked. As an example, Newtonian physics has been replaced by Quantum Mechanics and General Relativity as fundamental theories. But Newtonian physics gives answers in many situations that are incredibly accurate and easier to reach than the other subjects. The actual predictions in many (but not all) cases are very close. So many of the previous conclusions based on Newtonian physics are still valid in the modern context.
This issue arises in the study of evolution, for example. That species have changed over geological time is a fact that has been known since the early 1800's if not the 1700's. This is a well tested set of observations. The question of *mechanism* for such changes was a big dispute until Darwin figured out *one* of the main ways species change over time. Later, the Modern Synthesis joined the study of genetics (which was unknown to Darwin) with the questions of this mechanism. Later still the issue of Punctuated Equilibria was brought up. But, in each modification, the *tested* conclusions of previous descriptions have been kept...they were already tested.
Finally, the word 'science' is often used to designate the collection of tested results that have accumulated (as opposed to the method for accumulating them). These previously tested ideas are crucial for developing and understanding new observations.
Now, there are many areas that look scientific from the outside, but either are not or that do the science poorly. Any subject that starts with the conclusion (as opposed to a previously well-tested position) and only accepts ideas consistent with that conclusion is not scientific. Any viewpoint that keeps its ideas when the predictions made repeatedly fail to hold up under testing isn't science. There are a great many subjects that use, as standard, very low confidence thresholds for their testing (say, p<.05 is common in many areas). This leads to many 'false positives' and is a real issue.