Circumstantial evidence
Circumstantial evidence
/ˌsɜːkəmˈstænʃᵊl ˈɛvɪdᵊns/ (n.)
Evidence that suggests a fact — or is consistent with it — but doesn’t directly prove it. Circumstantial evidence, if reliable, makes a proposition more or less likely. We might say it changes the prior probability of the fact being true without directly attesting to it.
|
Crime & Punishment
|
“Direct” versus “circumstantial”
Compare circumstantial evidence with direct evidence that addresses the allegation without the need for further inference.[1]
Take the hypothetical allegation:
Colonel Mustard murdered Reverend Green with a lead pipe in the library.
A till receipt in Colonel Mustard’s wallet for a recently purchased length of lead pipe is circumstantial evidence to support this allegation. It is consistent with it, but does not settle it. A further inference is needed — that Colonel Mustard used his lead pipe to murder Reverend Green for which — yet — we have no evidence. Even if the evidence is fully reliable, the inference may yet be false. the best we can say is that, all else being equal, the till receipt is consistent with the allegation. It may increase the probability of the allegation being true — but by itself, not by much.
Not enough to secure a conviction beyond reasonable doubt, but enough to make Colonel Mustard a suspect. As we will see, once you are a suspect, things can change quickly.
But then Mrs. Peacock tells Inspector Clouseau, “I saw Colonel Mustard strike Reverend Green with a lead pipe in the library. He fell down, stone dead”. This is direct evidence. If we believe Mrs. Peacock, this evidence will settle the matter. It directly attests to cause and effect. The Colonel’s till receipt corroborates it, but is hardly needed to make out the case. He is already a goner.
You might say that direct evidence establishes causation while circumstantial evidence establishes correlation. One attests that the event did happen, the other that it is more likely to have happened.
An accumulation
With circumstantial evidence there may be major and minor significance — major pieces that get you most of the way to a proving your hypothesis, and minor, supplementary pieces that just confirm and validate it: that “put the matter beyond doubt”.
As such these minor pieces — especially if there are a lot of them — may suffer less scrutiny: they are corroborations; if they turn out to be mistaken they do not disprove anything; they just fail to ice a cake that could well be iced equally well by something else.
So it is, perhaps, with Lucy Letby. All evidence against her was circumstantial. Not just circumstantial but weakly circumstantial — consistent with the prosecution theory rather than tending to exclude of alternatives. This is prime territory for confirmation bias. The “killer fact” — the one piece of accumulated evidence that seems to support the prosecution theory to the exclusion of all others is the notorious nursing rota, on which she was on duty for every suspicious incident.
Are the remaining evidence weaker? Well, imagine the prosecution case without that linking evidence. Say we dispel the idea there was a single perpetrator. No nurse had an unusual confluence if shifts.All of the rest of the evidence — blotched skin, dislodged tubes, air in vessels and so on , was still present, and is presented not to a serial murder trial, but to a series of individual trials against different nurses. Now how seriously would we inspect the chain of custody of samples? The suitability of the insulin test? The value of the “consistency” of the blotching with air embolus? Without this evidence now there is no case to answer. The fact that Nurse Letby was on duty is of no moment because, well, someone had to be on duty, and it being Nurse Letby rather than Nurse Smith or Jones tells us nothing about the incident in question.
The blotching becomes simply “odd blotching” that could mean anything. The insulin poisoning becomes “an unusual insulin reading”. Analysis of insulin tests are vulnerable to base rate neglect after all. Especially tests that are not in the first place suitable for forensic purposes. This is no grounds for a murder prosecution. There remains an obvious reasonable doubt. A probable doubt. This was a premature sick baby who died of natural causes.
All of this changes if there is a highly improbable cluster of incidents and one nurse is a highly improbable common factor in all of them.
Now the blotchiness is not founding evidence but confirmation, with all the implications of that term: if it fits the hypothesis, it adds weight. If it does not it is irrelevant. Only were It to contradict the hypothesis — unlikely — might it have adverse probative value, but that requires a defence advocate to know about it, and be prepared to argue an alternative contradictory hypothesis (which, remember, is not the defence’s responsibility).
So we must come back to the “clinching” evidence — the ward rota — and ask whether it is the smoking gun it appears to be.
Vulnerability to tunnel vision
To connect to the crime, circumstantial evidence requires inference beyond the evidence in itself. A till receipt for the purchase of lead pipe means nothing without the inference that Colonel Mustard used the lead pipe to murder Reverend Green. The various cognitive biases making up prosecutor’s tunnel vision impact on that process of inference — a process which does not itself form part of the evidence. Theory dependence, Confirmation bias expert overreach base rate neglect and hindsight bias all tend to lead to wrong inferences out of innocent information.
See also
- ↑ In this discussion we will “control” for witness credibility by assuming all witnesses are speaking the unimpeachable truth.