The Correct Application of Probabiilty to Evidence
Which I would have titled 'The Correct Application or Probability to Evidentiary Assessment - stupid character limits...
Interrupting our previously scheduled program because I found a new shiny to play with for the moment - yes, I'm easily distracted, why do you ask?
Anywho, blame Leo for getting me on this topic last night in the SB.
So, we were discussing a Chesterton quote which he's read but couldn't find and I've only heard his paraphrase which goes something like this: 'I can't understand why someone believes his grandmother when she says she's seen the mailman but not when she says she has seen an angel' (Leo did a better job, I think, paraphrasing - I'm paraphrasing his paraphrase - okay, kiddies, this is NOT how you do this in debate!). I'll let Leo discuss his own understanding but here's how I interpret it: 'If Granny is credible for X under the exact same circumstances as Y then she is also credible for Y and the mere fact that the listener does not like Y is not valid grounds for dismissing Y'. (And you thought the first paraphrase was bad!).
In other words, if circumstances warrant believing Granny saw the mailman they also warrant believing she saw an angel and the mere content (mailman v angel) is not a reasonable reason to deny she saw either one. Sauce for the goose is also sauce for the gander where credibility is at issue.
So what factors do we correctly examine to determine credibility if not content? Reliability (can correctly report), conditions (can correctly observe) and reputation (will correctly report) are the main factors. If Granny has dementia then she isn't credible when she says the mailman came to the door or that an angel visited her - she cannot trust her own perceptions and neither can we. Both may be true, false or some combination but we can't know which is or is not true and probability is no help here. The mailman may be more probable - or not - it doesn't matter because Granny's ability to correctly report what she saw is compromised to the extent that we cannot trust it. If the mailman is later accused of murdering the neighbor at the time Granny says she saw him he's out of luck using Granny as a witness because it would never be admitted.
If conditions are poor Granny's ability to correctly observe is compromised and so is her credibility. It's dark and stormy out, the guy is two blocks away and Granny's eye sight isn't the best so did she really see the mailman or was it Bob the beat cop? Both are plausible but we can't know which because Granny couldn't have seen the person clearly. When she says the guy was in a blue uniform but had wings did she see an angel or a blurry vision of either the mailman or the cop? We cannot know and again, probability is no help. She could have seen another person entirely - the conditions are poor enough to compromise the credibility - inserting probability is the same as making a guess. We want to know what she saw - guessing doesn't accomplish that and really, without other evidence (a witness or picture), we cannot know.
Reputation. This, kiddies, is why it matters. If Granny is sharp as a tack, has great eye sight and saw the figure in broad daylight up close, but has a reputation for lying, her credibility is just as shot as when she had dementia or conditions were bad. Granny may have indeed seen the mailman and no doubt his attorney will call her to the stand (in desperation) but the prosecutor will eat her for lunch. "Isn't it true you told Bob the beat cop you were having supper at Brad Pitt's house that day?" "Um, well, I might have mentioned that to Bob but I really saw the mailman after I got home from Brad's..." Bob has already testified and the prosecutor will prove Granny doesn't even know where Brad Pitt lives - so how many jurors should believer her? None - she is not credible. She may be telling the truth about the mailman but because her relationship with the truth is so tenuous we can't trust it. When she later claims she saw an angel then instead, she is still not credible (okay, less so - mounting lies reduces credibility) and although she's finally telling the truth, we cannot know it and so rightfully dismiss the claim. Probability? Hopelessly useless - how the dickens would you determine when she is most likely or least likely to lie as those determinations are strictly about her - how do you get into her head well enough to even make a decent guess at how to assign values to your variables? You don't - not if you want your own credibility to remain intact.
So, when is probability appropriate to use? In those cases where we have the option to consider the evidence or not. That option is not available if you are debating the topic (side does not matter) or charged with a duty to find the truth (officers of the court, law enforcement, scientists, investigators, doctors - et al). If you are merely satisfying your own curiosity you have no obligation to examine anything and in that case you have every right to decide X is improbable and not worth consideration for your own purposes. The only other time probability enters in is arguable but I'd grant that if could be reasonably used to assess what priority you are going to give a given piece of evidence in the timing of your examination. It's legit to want to examine the most probable accounts first - but it would not be legit to rule out examining an account merely on probability. For that you must assess credibility - and probability tells us little to nothing about that.