USING A COERCED CONFESSION WITNESS GUILT CALCULATOR

I was recently trying to convince people how Crosley Green could have been convicted on the testimony of Kim Hallock alone, and certainly after adding the claim by his own sister Sheila Green that he confessed. I performed an analysis which showed that such coerced confession witnesses who are not in jail, are significantly more credible than the standard categories of jailhouse confession, and coerced co-defendant. Interestingly, according to my scratch analysis, such a witness can reduce doubt of guilt from 20% to 13%, even if he is more than twice as likely to be lying. And in the extreme case of a sister who is unlikely to lie about a brother and who is also a criminal whom Crosley would be open with, Sheila Green would be able to reduce doubt of guilt from 17% to 3%, and remove reasonable doubt even if there is a 19% chance she is lying. To examine this process, I put up a rough calcultor at the following link:

http://cops2prison.org/confession_calculator/

Let’s assume 3% of murderers will actually confess a true story to stranger in jail, 25% of desperate inmates will invent a fake confession whether or not it is useful, and 15% of murderers will confess a true story to someone they know outside jail before they get arrested. Let’s suppose 10% of drug dealers out on bond with pending cases are sophisticated and desperate enough to lie that an accused murderer they know confessed to them.

Let’s suppose an individual murder suspect knows on average 1 drug dealer out on bond with a pending case at the time he is arrested. But the chance of him confessing specifically to the drug dealer rather than someone else is 1 in 3, so that he would have to know three drug dealers with pending cases to have a 15% chance of actually confessing to even one of them.

The actual chance of a murderer confessing to a drug dealer he knows who also has a pending case while the murderer is in jail is .15 chance of real confession * 0.333 chance it is to the drug dealer = 5%. If the chance an accused murderer is guilty is 80%, the chance of an accused murderer producing a real coerced friend confession is 4%.

If the accused murderer knows average one drug dealer with a pending case at any time (two a year), the chance of a fake coerced friend confession is 5/6 will have the timing * 10% will be sophisticated and desperate enough to lie = 8.3%.

(The timing problem is that the murder and the falsely claimed confession need to take place somewhat correlated with the timing of the drug dealer’s case. The drug dealer may not think he needs a fake confession the day after he is arrested. There may be times in the murder case where they are not needing a confession, which may coincide with the months the drug dealer needs to accept a deal to avoid trial.)

So if we have an accused murderer who knows an average of one drug dealer with a pending case not in jail, the chance there will be a coerced non-custody confession witness is around 8.3% fake + 4% real = 12.3% and the chance it is a fake confession is 67%. Notice we assumed an 80% chance the accused is guilty, but there is still only a 33% chance any drug dealer on bond who claims he confessed is telling the truth.

Using Bayes Theorem

p (guilty | friend) = p (friend|guilty) p (guilty) / p (friend)
0.865 = .133/.123 * .8

The presence of a coerced friend takes the chance the accused is guilty from 80% to .133/.123 * .8 = 86.5%! So even if a drug dealer is more than twice as likely to be lying, he can take the chance a person is not guilty from 20% down to 13.5%. Here is the same problem presented in the calculator.

You can easily see how two or three partially-correlated drug dealers who claim a person confessed could take chance of innocence from say 15% to 3%, and seal a conviction. Even if all three are individually more likely than not to be lying.

Let’s suppose you assume a 20% chance the star witness got the lineup wrong, or something is just not certain. Let’s suppose there are three non-jailhouse confession witnesses. There is a good chance a person who confessed to one would confess to more. There is also a chance of a feeding frenzy, where more people come forward to lie, after they see the first liar get a deal.

We have one confession witness, and an 86.5% chance he is guilty, when suddenly another confession witness comes forward. The chance the suspect confessed once is 33%. A suspect who confessed once has a 25% chance of confessing again, not 5%. Chance random drug dealer will show up with a copycat lie is 25%. 0.865 chance of guilt * 0.25 = 21.6% chance of second real confession, plus .25 = 46.6% chance of second confession, 54% chance it is a lie.

p (guilty | second) = p (second|guilty) p (guilty) / p (second)
.9281 = .50/.466 * .865

With a second confession witness, who is more likely than not to be lying, the doubt of guilt has now dropped to 7%, from 20% with no confession witnesses, even though they are both likely liars. As long as there is even a slight chance the confession witnesses are telling the truth, they raise the probability of guilt by a small percentage.

The trick, is that even if the confession witnesses are lying, that does not prove the suspect is not guilty. So any chance they are telling the truth, increases the chance of guilt by some tiny amount. If the chance of innocence is already relatively small, it gets smaller.

At this point it is very likely something is going on. Either the suspect is guilty and is a serial confesser. Or all the drug dealers in town see they can lie to get out of jail. So there is very likely to be a third confession witness. But the third confession witness will not actually tell us as much, because the feeding frenzy effect is so strong, and the fact we already knew there was some chance the suspect is guilty and talks a lot.

If the suspect is guilty then he wants to confess more, but he is out of people he actually knows who need a deal. And every drug dealer in town will have realized they can lie and say he confessed. But the damage is already done with the first confession witness who is twice as likely to be lying.

You may wonder how can people who are most likely to be lying, cut doubt of guilt in half. It helps to think of it as the confession was previously missing if a star witness makes him most likely guilty. A person who if he is guilty has a 5% chance of confessing, should have 5% of a confession if he is guilty. Zero confession is less than the expected 5% chance of a confession. If a person wants to get out of a drug charge and there is some chance he will lie to do it, the claimed confession fills that hole as well. The confession makes both things more possible, lie and truth, by curing what is missing from both at the same time, if they were true. It is an expected lie, and an expected confession at the same time.

The biggest difference in jail is that there are 20 times as many sophisticated sociopathic felons who desparately need a deal and are looking to lie (and know there is never any possible penalty even if they are caught lying or even admit they did), the feeding frenzy started before the suspect even got to the jail, and the suspect does not know them to be likely to tell even one a true confession.