Yesterday, we discussed the nature of the scientific method and the intrinsic bias to prove one’s hypotheses correct. Given that fallible and inherently biased humans do science, one mechanism used to help overcome this bias is to, whenever possible, design experiments to disprove one’s hypothesis. Typically, a hypothesis is based either on a limited number of specific observations or on a large number of general observations. Therefore, trying to disprove a hypothesis helps to determine under what conditions the hypothesis is true. The harder it is to disprove the hypothesis, the higher the chances it is a reasonably accurate model and therefore more “rugged.”
Another, related, reason for disproving a hypothesis is the ‘correlation versus causation’ problem. Let’s say that someone observes that when event B occurs, event C also occurs. A preliminary hypothesis might be “B causes C.” Further investigation shows that there is another event, A, that also occurs immediately prior to B and C. Ultimately, we discover that A causes B and C. Therefore, B doesn’t cause C but occurs under the same conditions as C. In scientific jargon, we say that B and C correlate with each other, but A causes them. It turns out that many mistakes in science have occurred because the scientist assumed that B causes C rather than correlating with C and both being caused by A.
One historic example of this is that of spoiled meat producing maggots. The belief since Greek times was that if you left meat out to spoil, it would spontaneously create life in the form of maggots. This idea, appropriately enough, was called ‘spontaneous generation.’ Italian biologist Francesco Redi in 1668 disproved this theory by putting gauze over the meat, which kept flies from landing on it and laying their eggs, which hatched into maggots. Meat with gauze had no maggots. Meat without gauze had maggots. Thus the spoiling of meat did not cause maggots. The laying of eggs by flies in meat caused maggots and also contributed to the spoiling of the meat. A 2000+ year understanding of nature was overturned by someone who understood the difference between causation and correlation and how to experimentally distinguish between the two.
To emphasize the point, let’s look at a whimsical example of choosing good hypotheses and designing good experiments. A string of strange disappearances has occurred near Waller Creek (which runs through the UT campus). Victims were either alone or with a group and would suddenly just vanish. One UT student, a fan of the paranormal, suggests the disappearances are due to a rare creature called a one-eyed, one-horned flying purple people eater. Of course, society laughs. His response, “Prove I’m wrong.” The budding conspiracy theorist has a point. Which is easier to prove, the existence of something or the nonexistence? To prove nonexistence, one must explore the height, width, breadth, and depth of the natural world and show that in no situation does the object exist. To prove existence, one must merely produce the goods. Not being able to produce the goods is not a disproof but a failure to prove. There is a difference.
In either case, the real problem is the statement of the hypothesis. A slightly better hypothesis might have to do with the existence of such a creature on modern planet Earth (though this is still potentially a tall order). In order to make the hypothesis workable, the student would need to also propose what sorts of evidence/tests would satisfactorily test or ‘falsify’ the hypothesis. So you can see that it is important to choose both an appropriate hypothesis and design an appropriate experiment for its testing.
A possible appropriate response to this student is that if he is to make a truth claim such as the disappearances are caused by a one-eyed, one-horned flying purple people eater, then the burden of proof is on him, the maker of the claim. To tell others to prove him wrong is irresponsible and not intellectually honest, especially as the request to disprove is impossible unless you are an omnipotent person with full knowledge of the universe past, present, and future. The trouble is that polemics tend to make better sound bites than truth: it is a better sound bite to say “Prove me wrong” than it is to explain that it’s his job to prove his own point. His snappy response is that I’m blame shifting, when the truth is that he is doing the blame shifting, then accusing me of it when I try to shift it back.
The problem of good hypotheses and good experiments is why science has the process of peer review, as a check on bad science. It is not foolproof, but it is better than nothing.
SDG
No comments:
Post a Comment