How Evidence Works
To begin with a familiar use of evidence, consider the reasoning in criminal investigations frequently
portrayed in books, movies, and television. Such reasoning has been performed by Sherlock Holmes,
the detectives in Agatha Christie novels, the investigators in TV shows such as CSI and Law &
Order, and many other fictional characters. Reasoning to identify the criminals responsible for illegal
actions is also performed by real-life investigators and prosecutors, as in the famous case of O. J.
Simpson, a football player and movie star whose ex-wife was killed in 1994. Los Angeles detectives
collected many kinds of evidence, such as Simpson's bloodstained glove, that led many people to
conclude that he was guilty. Nevertheless, a jury in 1995 acquitted Simpson on the grounds that the
prosecution had not shown beyond a reasonable doubt that he had killed his ex-wife. The jurors were
legitimately influenced by evidence that racist members of the Los Angeles Police Department had
fabricated some of the evidence. But it also appears that some of the jurors were motivated to find
Simpson not guilty because of his achievements in football and movies.
Such motivations aside, here is how legal reasoning is supposed to work. Detectives and forensic
investigators of a crime collect all the available relevant evidence, such as fingerprints. The best
evidence is gleaned by carefully conducted observations, as when investigators thoroughly go over
the undisturbed crime scene using techniques such as dusting for prints, collecting hairs, and taking
photographs. Evidence can then be supplemented by scientific tools for analyzing blood and DNA.
Contrast these kinds of evidence with information unlikely to have any connection with the actual
crime, such as a psychic who reports seeing a killing in a dream.
On the basis of evidence and information about the victim, investigators form hypotheses about
who committed the crime, and evaluate these hypotheses according to how well they explain the full
range of evidence. A hypothesis is a guess about what might have caused something to happen. For
example, the hypothesis that Simpson killed his ex-wife provides an explanation of why her blood
was found on his glove. The explanation here is causal: the event of Simpson's stabbing her could
have produced blood that got onto his glove. The job of the defense is to propose alternative
explanations, in this case that the blood on Simpson's glove was planted there by police officers, and
that Nicole Simpson was killed by drug dealers rather than by O. J. The jury is supposed to
impartially determine whether the hypothesis that the accused committed the crime is the best
explanation of the full range of evidence, beyond a reasonable doubt. Philosophers call this kind of
reasoning inference to the best explanation.
Such reasoning is commonplace in everyday life. You use it whenever you are puzzled by the
behavior of someone you know, as when a normally good-natured friend treats you in a hostile matter.
In such cases you naturally seek explanations—for example, your friend is depressed because of
troubles at work or school. An alternative hypothesis might be that you inadvertently said something
that your friend found insulting. You need, then, to collect additional evidence that might tell you
whether work stress or a perceived insult is the best explanation of your friend's hostile behavior. We
use similar reasoning in dealing with mechanical problems. When your car won't start and you have
to take it for repairs, the mechanic's job is to find the underlying breakdown that is the best
explanation of what's wrong. Mechanics carry out a number of tests to try to figure out whether it is
the battery, the ignition, or some other component that is preventing your car from starting.
Similarly, when you go to the doctor with a medical complaint—say, a pain in your stomach—your
doctor collects additional evidence by probing your abdomen and possibly ordering tests such as
blood work and X-rays. Your doctor's diagnosis is an inference to the best explanation about what
underlying disease is responsible for the full range of evidence, including both your reported
symptoms and the test results. The television show House portrays an obnoxious but brilliant doctor
who every week has to find an unusual diagnosis for a patient suffering from an unusual range of
symptoms. Dr. House is carrying out the same kind of reasoning as would Sherlock Holmes and your
automobile mechanic: collecting evidence and trying to find out the best explanation for it.
Legal and medical hypotheses often involve multiple layers of explanations. Detectives looking for
evidence that a suspect is guilty of a crime collect observations, such as fingerprints on the murder
weapon, that are explained by the hypothesis that the suspect did it. But they also investigate possible
motives that would explain why the suspect did it: perhaps the suspect was angry at the victim
because of a previous fight. Similarly, a doctor looking for the best explanation of your stomach
symptoms will try to ascertain not only the condition that caused them, but also what might have
caused your condition. For example, your having eaten some exotic food might explain how you got a
gastrointestinal infection that is the cause of your stomach pain.
2.1 Structure of inference to the best explanation, with a higher hypothesis explaining a hypothesis that competes to explain the evidence.
The solid lines indicate explanatory relations, whereas the dotted lines show competition between alternative explanations.
Figure 2.1 depicts the structure of how hypotheses such as those about diseases serve to explain
observed evidence and are themselves explained by higher-level hypotheses. The general case is on
the left, and a very simple medical example is on the right, with solid lines indicating explanatory
relations and dotted lines indicating competition between hypotheses. In the general case, hypothesis
1 is highly coherent because it explains two pieces of evidence and is explained by a higher
hypothesis 2, which makes hypothesis 1 superior to a competing hypothesis 3 that explains only one
piece of evidence. Similarly, in the stomach example on the right of figure 2.1, the hypothesis that the
ache is caused by a bacterial infection wins out as the best explanation both because it explains more
evidence than does the competing ulcer explanation, and because it can be explained by the
hypothesis of having eaten bad food. Choosing the best explanation requires not just counting the
pieces of evidence explained, but also evaluating which of the competing hypotheses have most
overall coherence with all the available information.
Evidence and Inference in Science
Of course,