If I may elaborate and round off what I see are some sharp edges :--
In science, all we ask is that the proposition be consistent with all of the evidence we've gathered so far and that it be parsimonious - not add more assumptions than the minimum necessary.
Yes. But beware that "all the evidence we've gathered" is not an invitation to reason in circles. In science, we note an observation X that isn't presently predicted by any model. After some thought and discussion, we formulate a hypothesis H that attributes X to some supposed cause A. X is not proof of H. It may make certain sense to explain X in terms of what we think about A, but that's only because we formulated H arbitrarily and suppositionally to say that. To say that X therefore proves H is to reason circularly. X is not a sufficient body of information in science to constitute proof of H. We must gather more evidence of a different kind to test the hypothesis and break the circle.
We don't know that A causes X, and we probably can't observe A directly. But we may know that A must invariably result in observation Y, which is different than X. Y may be a strong enough proxy for A. We therefore deduce that where we fail to observe Y, we can deduce that A did not operate. Where we do observe Y, we can deduce that A operated. If we can observe X and Y together, and they vary together, this is when we can begin to say that we have enough information for scientific proof for H. The notion that causes, when they operate, leave behind more kinds of deducible potential observations than just the observation we want to explain is the heart and soul of what we call the hypothetico-deductive method of scientific reasoning. Where we can't directly observe the operation of some hypothesized cause, we turn to what we can deduce also follows the cause.
What we see often in reasoning about supernatural causes is that the hypothetical attribution tries to be the proof. If a person observes something that's difficult for him to explain by ordinary means at his disposal, he may formulate a hypothesis attributing the observation to supernatural causes. The contours of the hypothesis are intricately carved to match the contours of the observations and circumstances, leading to the illusion that the hypothesis forms a richly developed theory.
Here in this thread we're examining what happens when science recognizes one of those contours and says, "But we deduce that if this were the causation, then this other observation would hold. It does not, therefore the causation fails a scientific standard of proof." (More on this later.) Far too often the response from the claimant is to whittle out a new
ad hoc curlicue in his hypothesis that deftly avoids the scientific deduction while maintaining the gist of the claim. Accumulate enough of these speculative notches, rabbets, and
fleurs-de-lis and you have enough material to sell it as a book in those out-of-the-way bookshops that always smell so very nice when you step into them. Scientifically, however, the latticework of
ad hoc refinements to the hypothesis soon collapse under their own
prima facie weight, to be ground underfoot by the handsomely-sandaled foot of William of Occam.
[Loss Leader: "Objection, relevance."}
Yes, yes, I promised to talk about circularity. If we proposed to talk about all that's wrong with reasoning to support the supernatural we'd have, well, the forum
in toto.
When a claimant is asked why he believes in the supernatural, he'll cite the former problematic observation. In other words, the observation that the hypothesis was formulated to explain is improperly given as the evidence that the hypothesis is true. It doesn't stand out as the obvious circularity it is, often only because it is a widely-communicated hypothesis, or one that was made in antiquity and repeated often enough to have become familiar.
The acquisition of
additional, carefully-selected evidence is what makes a hypothesis scientifically tenable. Only when the observation
quod erat demonstrandum and a consilient sum of additional evidence deducible from the proffered causes are amassed and considered together do you meet a scientific standard of proof.
What Loss Leader needs us to understand is about this is that not all that can be deduced according to H in terms of possible observables from cause A, say the set Z, Q, W, R, and V, will actually be observed in ways that correlate well to X. What we require from science is that the speculative assumptions we employ to prop up A and H in the face of those shortcomings are as few as possible, and that the few that remain are reasonable on their face. Science doesn't lie about them or sweep them under the carpet. It simply lists them as the conditions under which the theory thus proven can be considered predictive.
Historical proof is less rigorous.
History and science become bedfellows when we ask science to explain happenstance events -- i.e., those which cannot be repeated on command or predicted.
We can't rewind the clock and replay the Battle of Cowpens. But we can know enough about what happened to conclude that Roland Emmerich didn't quite get it right. And we can't replay the crash of TWA 800. But we can know enough about it to say that it wasn't shot down by terrorists.
Back in the days when airplane cockpits had so-called steam gauges, we used to be able to recover their readings at the moment of impact using a phenomenon called needle slap. An airspeed indicator that goes from traveling at hundreds of miles per hour to a dead stop in only a few feet incurs such a high g-force that the needles hit the gauge faceplate hard enough to mar the paint at the place along the arc that the needle is currently reading. This and countless other tidbits of knowledge amassed via decades upon decades of systematic scientific testing and observation carried out independent of any particular crash lets us loosely fit the bits and pieces of what we
can know from a crash site to a particular likely pattern of causation.
Similarly, historians can read what survives regarding 18th-century British infantry tactics and contemporary records of what was observed and what can be known via letters and journals about the people involved and come to a reasonably reliable picture of what happened on some momentous day at some important place.
What the study of happenstance events makes us think about is the difference between a standard of evidence and a standard of proof. The way I conceive it, standard of evidence describes what kinds of evidence you will regard, while standard of proof describes how much of it you need before a proposition can be relied upon.
The law has standards of evidence. Hoo boy, does it ever. In the U.S. we start with
Federal Rules of Evidence and go from there, right down to individual rulings
in limine that affect individual cases. I'm told that whether a case actually makes it to trial depends significantly on what is to be considered "admissible," or what meets the standard of evidence for that issue. Science has standards of evidence too, involving criteria such as reproducibility, objectivity, and measurability. The wisdom behind these criteria doesn't differ much from the rationale behind legal rules of evidence. We don't allow certain kinds of testimony in court because it's prejudicial, which correlates to scientific standards of objectivity. We don't allow hearsay as evidence of the matter asserted because it cannot be meaningfully put to the test. This is similar to reproducibility.
Science and the law butt heads sometimes on standards of evidence. Repeated testing in contrived high-g rigs have confirmed the reliability of the needle slap test. We can go into a crash site confident that if the faceplate can be carefully recovered, the slap mark on the airspeed indicator will be the airplane's speed at impact. However, little such preparatory work preceded evidence that fingerprints from a database can be assertively matched to others obtained from a crime scene, or that bite marks on skin can be conclusively matched to a given set of human teeth. Much of what has been presented in court as "forensic scientific evidence" doesn't at all meet a scientific standard of evidence.
[Loss Leader: "Objection, relevance"]
Okay, right. The anecdote is the ill-favored stepchild in the family of evidence. You have to examine it in terms both of standards of evidence and standards of proof. The reason anecdotes -- even lots of congruent ones -- don't make evidence is because they are generally untestable. They are the hearsay of the evidence world. As such they may incorporate selective recounting of fact,
ad hoc interpretations, and so forth. When science is trying to discover a general rule, anecdotes are inadmissible.
But science isn't always trying to find a general rule. Sometimes it's just trying to find out why one Boeing 747 crashed. As such, it must consider eyewitness testimony -- happenstance observation in the form of an anecdote. But this is because science here is operating at a lower standard of
proof. It's operating at more of an historical standard of proof. History notes general trends, but doesn't shoehorn individual occurrences into them. History and forensic science are more concerned with what happened here and now, agreeing that it may never be possible to decide that to a scientific degree of confidence.
On the one hand, the Battle of Cowpens was a classic double envelopment. It was a tactic well known in the 18th century among leaders of armies. It can be found in contemporary writings on how to conduct battles. The Colonial commander simply executed a textbook maneuver. But how did he know it was the right thing to do? Because the British commander at Cowpens was known to be brash and aggressive. The double-envelopment tactic requires the attacker to overextend his line, something the Americans felt this particular British commander would do. By the same token, overextending the line was something the British would not have done but for the individual characteristics of one person. The lesson here is that the relationship between standards of evidence and proof depend highly on the deductive and inductive properties of the thing you're tying to study. The battle was simultaneously representative and particular. There is no one-size-fits-all standard.
Here in this thread we're looking for a general-case explanation. We want to know how any soul allegedly communicates with its host organism. The desire is for a theory that will hold in all or most cases, not one that answers the particular fact pattern alleged in one anecdote. We're not trying to explain the equivalent of just the Battle of Cowpens, but rather some equivalent to the general doctrines of infantry engagements.
The standard of evidence in history accommodates the impracticality of obtaining new evidence to test hypotheses. As such, the standard of proof in history is more lenient. And because it is, the conclusions that history accepts remain more tentative. What JoeMorgue fears is that because different disciplines console themselves with different standards of evidence and different standards of proof when deciding different kinds of questions, all such standards must be arbitrary. That's not the case. Any conclusion in any field that is meant to stand as the general case must be based on standards of evidence and of proof that rise to that level of rigor. Anecdotes just don't do that on either dimension.
I won't delve into legal proof here because it really does seem to be a different animal altogether.
And then there's personal proof - whatever you choose to believe based on your own experience.
That goes right to the various definitions of evidence and how that relates to the standard of proof.
We say that evidence is information that tends to make a proposition more probable or less probable. The question then becomes how we reckon that probability. In science, the standard of proof is generally a statistically computable probability that there is only a 5% chance the hypothesis is wrong. Granted that can be considered an arbitrary standard, but what's more important is that we require the relevant probability to be computable. That burbles back to the standards of evidence. Evidence is admissible only if we can measure the relevant observations so as to afford a statistically computable level of confidence in the answer.
History not so much. We have a fair number of documents in hand from the 18th century that prescribe good practice in leading infantry. And from those we can come to believe that Col. Tarleton botched the attack according to what another British commander may have done. But the probability that our conclusion is correct is difficult to quantify. You can't measure propositions like that in terms of numbers. No toehold therefore for statistics.
The law often uses the "reasonable person" standard. I'm estopped from going into too much detail by my comments above. But the gist of it is that we empanel twelve persons who we agree will collectively act as a reasonable person to weigh competing claims of fact and estimate which version of the facts a reasonable person would consider most probable. Perhaps not thoroughly objective, but practicable.
The personal standard is just the estimate of probability that allows the greatest proportion of subjective factors. Science tries very hard to approach the mathematical standard of proof. It seeks to measure the residual inductive leap and force it to be very small.. The personal standard is just at the other end of the objectivity spectrum from math. Too often a claimant wants us to just keep lowering the standard of proof -- the objectivity in the estimate of probability -- until his proffered evidence clears the bar.
And that's really what we want to point our fingers at. The tendency of information to affect the probability of a proposition is most hopefully employed to create belief where none -- or a contrary belief -- previously existed. We want evidence to be convincing to people who don't already hold the belief and who may even be predisposed to reject it. The only evidence that has the power to do that is evidence whose effect on probability tends to be objective. So while a variety of standards of proof exist, a claimant doesn't arbitrarily get to decide what the standard should be for his question.
