Scientific Explanation
Next, we consider the basic structure of the most comprehensive and effective deployment of inductive reasoning in human history.
Since its development during the Renaissance, modern science has contributed significantly to our ability to perceive, understand, and manipulate the natural world.
Taken generally as a way of acquiring human knowledge, science is a procedure for the invention and evaluation of hypotheses that may be used to explain why things happen as they do.
Unlike dogmatic appeals to the absolute, unchallengeable truth of unsupported assertions (as, for example, when a parent tells a child, "Because I say so, that's why."),
scientific explanations are always tentative proposals, offered in hopes of capturing the best outlook on the matter but subject to evaluation, modification, or even overturn in light of further evidence.
The most productive model for the structure of a scientific explanation is that of a valid deductive argument whose conclusion is the event to be explained.
Some of the premises of this argument will be factual statements of the antecedent circumstances, while the others will be the
scientific hypotheses offered as a way of linking those circumstances to the outcome stated by the conclusion.
Scientific predictions have exactly the same structure; the only difference between the explanation and the prediction of an event is whether or not it has already occurred.
On this deductive-nomological model for scientific explanation, the conclusion of the argument must be true (that is, the event must occur) if all of the premises are true.
Those of its premises that state the antecedent circumstances will naturally be true so long as we have our facts straight.
But the truth of the hypotheses, which try to capture the lawlike relationship between those circumstances and the event to be explained, will always remain open to question.
So the quality of the explanation as a whole typically rests upon the extent to which these hypotheses are reliable.
This reliability can never be established with absolute certainty.
It is sometimes possible to eliminate bad hypotheses by using them as the premises of a deductive argument predicting that
particular consequences will follow from a particular set of circumstances and then showing that the predicted event does not, in fact, occur.
(This amounts to the use of Modus Tollens to show that since the consequent is false, some part of the antecedent must also be false.)
But if the events turn out as predicted, that only tends to confirm the hypotheses; it cannot prove their truth.
(Since that would amount to reliance on the fallacy of affirming the consequent.)
Empirical evidence typically underdetermines scientific explanation, leaving us with multiple hypotheses, any one of which would account for the facts.
Although it always remains impossible in principle to prove the truth of a scientific hypothesis, it is possible to compare the distinct hypotheses involved in rival explanations of the same event.
Here are several criteria that can bu used in making these relative judgments about the reliability of any hypothesis offered as part of a scientific explanation:
- Relevance.
At the very least, it must be possible to use the hypothesis as one of the premises of a valid deductive argument whose conclusion is the event to be explained.
In addition, we normally expect that the best explanation of any event will involve hypotheses whose relationship to that event is readily apparent.
Although it could turn out that an apparent case of food poisoning was actually caused by wearing uncomfortable shoes, for example, that hypothesis is less relevant than one concerned with the consumption of victuals.
- Testability.
There must be some way of acquiring evidence that would tend to confirm or discomfirm the hypothesis.
In fact, good hypotheses are always falsifiable in the sense that it is possible to state specific conditions under which the hypothesis would be decisively overturned.
(If, for example, we fed the tainted coleslaw from Lesson 23 to a hundred people and none of them got sick, this outcome would prove the falsity of Nurse Hayes's hypothesis that the students' indigestion was caused by eating the coleslaw.)
This is what is wrong with dogmatic hypotheses: there are no circumstances under which they could be proven false, and that is a measure of their total vacuity.
Good scientific hypotheses are always "on the line," subject to falsification by the appearance of counter-evidence.
- Compatibility.
The hypothesis should fit well with what we already believe about the natural order of things.
Since centuries of scientific investigation have provided us with a body of knowledge comprising many reliable hypotheses, we should expect that each newly added hypothesis will be consistent with them.
Of course, this criterion is not absolute; genuine scientific revolutions can involve the abandonment or significant modification of previously accepted hypotheses, and it is important to allow for that possibility.
- Predictive power.
A good hypothesis isn't just a way of explaining events of this one sort, but will be applicable to many other kinds of circumstances as well.
Newton's "Law of Universal Gravitation," for example, has a lot of predictive power, since it can be used to explain great portions of both celestial and terrestrial motion.
The extreme case of poor predictive power is an ad hoc hypothesis, which can be used to explain only a single occurrence of a particular event.
- Simplicity.
The best hypotheses are rarely intricate in structure.
Again, it would be possible to overemphasize this criterion, since an adquate complex hypothesis is clearly preferable to a simple but inadequate one.
Still, there is a profound elegance and dignity in the simple character of many scientific hypotheses (universal gravitation is a good example again).
What is often called the scientific method is nothing more than a step-by-step procedure for the conduct of scientific research:
- State the Problem.
It is important to begin with a clear statement of what phenomenon is to be investigated.
This isn't always as easy as it sounds; faced with a real-life perplexity, we sometimes become distracted by issues that aren't really to the point.
But careful adherence to the rest of the method will be useless if we don't focus on the matter of greatest genuine concern.
- Invent Preliminary Hypotheses.
Next, we spin out as many possible explanations for the phenomenon as we can.
At this stage of the process, there is no reason to limit the range of our creativity by dismissing anything as irrelevant.
We'll have opportunity to weed out bad hypotheses later, but a possibility overlooked at this point may be lost forever.
- Collect Additional Information.
We next try to observe the phenomenon in its natural context from every angle.
Again, at this early stage, the premium is on breadth of investigation rather than on a prematurely narrowed focus.
The goal of this wide-ranging review of the facts is to gain some insight into the relative likelihood of our preliminary notions.
- Formulate a Hypothesis.
Now we're ready to focus on a specific hypothesis, using the information we've gathered to devise a detailed (though still tentative) explanation of the phenomenon under investigation.
This marks a significant shift in our procedure, returning to the narrow focus of our original definition of the problem.
- Deduce Further Consequences.
Since a good hypothesis has predictive power that reaches far beyond its function in any particular explanation, we now consider its additional consequences.
If the hypothesis used to explain this phenomenon were actually true, what else would follow from it?
Again, our work at this stage should be narrowly focussed: exactly what should happen if we have identified a correct hypothesis?
- Test the Consequences.
Now we look at the facts again, to see whether or not these consequences actually occur.
If we can set up a concrete situation in which our hypothesis, if correct, would lead to striking results, then if they do not occur as expected, we'll know that we were wrong and need to go back to step 4 and come up with another hypothesis.
- Apply the Hypothesis.
If everything checks out, however, we are ready to apply our new explanation to the original problem for which it was developed.
Of course, there's still no guarantee—it may work out everywhere else and yet not deal effectively with this case.
But we can always go again.
In general, we repeat this procedure as often as necessary, going back to start over at step 4, or step 2, or even step 1 until we arrive at a satisfactory solution to the problem.
Notice that although scientific method is properly considered to rely upon empirical knowledge, several steps in the process do not appeal to any direct observation of the facts.
Steps 1, 2, and 4 in particular often involve creative leaps of intution that are not warranted by the evidence.
Novel ways of defining a problem and radically different hypotheses about solving it typically arise from insight and imagination rather than from observation.
But even in such cases, the scientists differs from dogmatists in their determination to put every such hypothesis to the test in steps 5 and 6.
Usually experimentation confirms a hypothesis by eliminating its likely alternatives.
Thus, the most powerful confirmation occurs when we are able to devise a "crucial experiment,"
a set of circumstances from which rival hypotheses predict distinct results: when we perform such a test, one hypothesis nearly always emerges as the most likely.
But this situation tends to arise only within the context of a well-developed general theory with which only a few alternative hypotheses would be relevant.
As we've noted before, scientific revolutions can occur only when we step outside the bounds of such a restrictive experimental framework.
It is worth noticing that even apparently factual, "objective" investigations of the natural world often rely upon the theoretical background of a set of accepted scientific hypotheses.
Thus, for example, biological taxonomy and descriptions of historical events usually employ formal systems of
classification that embody significant hypotheses about presumed similarities of nature or origin.
Here, as in all of our efforts to engage in sound reasoning, it is vital to recognize and uncover the implicit foundations of what we believe.
The Philosophy Pages by Garth Kemerling are licensed under a Creative Commons Attribution-ShareAlike 3.0 Unported License.
Permissions beyond the scope of this license may be available at http://www.philosophypages.com/referral/contact.htm.
©1997, 2011 Garth Kemerling.
Last modified 12 November 2011.
Questions, comments, and suggestions may be sent to:
the Contact Page.
