Fallacies of evidence and causality
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Hasty generalization (rash generalization, insufficient sample, jumping to conclusions)
In a hasty generalization a conclusion or generalization is based on too small a sample size, sometimes on only one or a handful of examples. Of course, it is not always easy to determine how much data is needed to reach a valid conclusion. Most of the time it is more valid to assign a probability rather than make an absolute statement. Scientists and pollsters use statistics to determine the confidence limits and margins of error.
"That 'psychic' was using cold reading. They must all be frauds."
"Women are all lousy. My first three wives were real losers."
"After treatment with the drug, one-third of the mice were cured, one-third died, and the third mouse escaped."
There was a news story that the major export from one port was scrap material. It was soon being reported that scrap was the major export from the hole country..
Biased data (unrepresentative data)
Here the sample size may be large enough, but it is not representative of the whole. If you asked the shoppers in an organic food store what they thought of pesticide use, their replies would not be representative of the population as a whole. Here are some examples from news stories:
A survey of college students found that music was the most important thing in their lives. It was soon discovered that it was a survey of music majors.
Another survey found that Hawaii was the favorite vacation destination for Americans. But the survey used data from travel agencies, and not all vacationers use travel agencies. According to most surveys, Florida is the favorite vacation destination for Americans.
Cherry picking is deliberately picking out the data or scientific studies that support your view, while ignoring the data or studies that oppose your view.
People tend to remember what confirms their beliefs, and forget what does not. For example, they will remember the times they made a correct prediction, and forget the times when they were wrong. Or they will remember the rude teenagers, or the lousy woman driver, reinforcing their prejudices and stereotypes. People also tend to look for information that supports their beliefs, while ignoring or downplaying whatever contradicts it.
These are events that happen to someone, or that are told to them, with no attempt at any scientific analysis. Some are simply stories, events that did not really happen or that have been significantly altered (for example urban legends). They are useful for adding interest, and in illustrating points, but they should not be used to make generalizations.
"This is third day in a row that we've had a record high temperature. Global warming must be real."
The gambler's fallacy is the belief that the possibility of a completely random event can be influenced by other events. For example, if a tossed coin has turned up heads three times in a row, the gambler believes that their is a reduced chance that it will be heads the next time. But the chance of it being heads is still 50%. This fallacy leads people to believe that they are on a hot streak, and cannot lose, or on a cold streak and just cannot win. Conversely, they may believe that their luck is about to change. They may also believe that a slot machine that has not paid off for a while is about to pay out big, or that certain numbers are more likely to be picked in the state lottery.
A variation is the two bomb fallacy. According to this thinking, since the chance of there being one bomb on an airplane is very small, the chances of their being two bombs is virtually zero. Therefore, if they bring a fake bomb aboard they have greatly reduced the chance of real bomb being on board. But in reality bringing one bomb on board has no effect.
If the incidences of an event are randomly distributed, then they will tend to form clusters. These can occur based on time (the hot streaks and cold streaks of the gambler's fallacy) or over space. The most famous of the spatial clusters are the so called cancer clusters. But cancer clusters are what you expect, based on a random distribution. And if you cannot find a cluster based on total cancers, you can look for certain types of cancer. Unfortunately, a whole cottage industry has grown up around looking for cancer clusters. This does not mean that there is nothing wrong, just that the number of cases needed to demonstrate a problem is extremely high.
Post hoc fallacy (post hoc ergo propter hoc, after this because of this)
In a post hoc fallacy it is claimed that because some event b occurred after event a, then a caused b, with no evidence given to show causation.
"We didn't have this strange weather before they started flying the space shuttle."
"A back cat walked across my path on my way to work. An hour later I got the worst paper cut I've ever had."
From Elizabeth Whelan (see The DDT ban myth) " Why was there an increase in malaria in Ceylon [now called Sri Lanka] after 1964? It is clear that the effects of Silent Spring was not limited to the United States. Following the publication of this book, the use of DDT was discontinued in Ceylon."
In a correlation, as a goes up b also goes up, and when a falls b also falls. (In a negative correlation as a goes up b goes down.) Usually it is possible to do a mathematical analysis and determine a correlation coefficient. Correlations are more impressive than post hoc arguments, but they do not prove causation. The changes may be do to simple chance, or some other factor may be the cause. It is also possible that the two factors are reinforcing each other, or that b is causing a.
Leaving out a evidence that would weaken or even cause the average reader to dismiss the claim. This is so common that it requires its own page.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Written by Jim Norton
Visit my practical skepticism page
Visit my anti-environmental myths home page.
The text on these pages may be freely copied, distributed and posted as long as my name, this statement and the URL (http://info-pollution.com/evidence.htm) are included.