How can you determine the probabilty of a witness identifying a suspect?

Question

Many people who witness crimes become suspects. Many people who commit crimes invent perpetrators who do not exist. How can you evaluate the probability that a witness could be lying and potentially be the perpetrator? I will give two examples, one is a specific case.

1. Witness identifies a vague description of a single person of a racial/ethnic minority that frequently commits crimes in the area of the act in question.

2. In the case of Dr. Jeffery R. MacDonald, he identified four people, each with different racial characteristics, and included details about hair and articles of clothing. The description also included that they were drug users, and the nature of the crime suggested they also are involved with satan worship. The description of a woman was so specific that it lead to a unique person. In addition to what I imagine is a low probability of Jeffery R. MacDonald imagining a real person that exists (in the case that he is the perpetrator, not the witness), what is the probability of the details of this person's confession matching the details of the crime if they did not commit the crime, and then for each additional witness statement corroborating the vehicles she was in, who she was with, time she answered the phone, etc?

Answer 1

Essentially the probability the witness is right is not quantifiable, but his testimony shouldn't be trusted too much. Keep in mind:

• eyewitness evidence is not a very reliable kind of evidence. Eyewitnesses fabricate what they don't remember, sometimes without even realizing they have done so. Additionally, most people claim more confidence in their beliefs than is truly warranted. See this article, including the gem, "Some 75% of the wrongful convictions for rape and murder, including a number that led to people being scheduled for execution, were based on eyewitness testimony."
• An assertion with more details should logically be trusted less than an assertion with fewer details. P(A and B) is always less than or equal to P(A) and also less than or equal to P(B). See the conjunction fallacy.
• If you give a detailed enough description of a person, eventually only one person will fit that description, even if you chose the description totally at random. "A tall, black-haired man in his 30s with a red sports car and a broad face, overweight, balding" - there might be only one or a few people matching that description in any given town or neighborhood, despite the fact that I chose it at random.
• Jeffrey R. MacDonald was the prime suspect. If he committed the crime, and there was physical evidence he did, he would have had a strong reason to lie.
• MacDonald's claim that four people acting together murdered his family while chanting "acid is groovy, kill the pigs" is not believable. It reads like he needed a scapegoat, chose "acid heads" (in his first statement to the police about them) and then had to contrive a reason he would know they were acid heads. It's the kind of thing a cartoon caricature of a hippie would say.