Evidence!

[R_Daneel]

A lot of threads seem to be bogged down on this concept. It might be a good idea to devote one to this problem. What is evidence? What kind of evidence sould one consider and accept? When does the evidence become overwhelming? I am pretty sure there is a lot of subjectivity about it. Something that is accepted as evidence by one person, could be unacceptable by others?

Is it possible to reach some kind of concensus about this? (Personally I doubt it, but it might be interesting to explore).

 

[James_S_Saint]

It is much like definitions. No one wants to settle on it because that would end the arguments.

 

[Sair]

I think it really depends upon what one is trying to verify, and the circumstances that require the verification.

I was asked recently on another thread whether, when someone told me what they were feeling, I would take their word for it or insist that they give me ‘evidence’ of what they were feeling. Clearly the latter suggestion is absurd. In everyday conversation, a person’s assertion that they are feeling happy, down, excited, etc is all the evidence we need, coupled with visual cues from their facial expressions and body language, which are also important factors in human communication. We might question their claim to be happy if their mouth was downturned, their shoulders slumped, their tone of voice flat - we might suspect they were not being honest about how they felt, and would perhaps, if they were a close friend, want to question them and find out if there was anything we could do to help them. On the other hand, we probably wouldn’t suggest they get a brain scan to find out what was ‘really’ going on in their head. In such circumstances, our normal, everyday means of communication are sufficient.

Historical evidence is another matter. If we want to find out what people of the past were thinking or feeling, we might rely on an old diary, for example. Mostly, the thoughts and feelings of past generations of people are lost to us, and the documentation we have of past events is often quite limited. The important things when it comes to using historical evidence are reliability and corroboration. To assess reliability, we need to decide how closely the document was written after the event it deals with, whether the writer was really in a position to know about the events they describe, and what their bias might have been. And if multiple sources agree on certain aspects of the events they deal with, we can be reasonably sure that certain things happened in the past. There are, of course, some things that we’ll never know beyond reasonable doubt, and that’s why historical fiction can be so fascinating!

The nature of evidence becomes more serious when we are dealing with a legal case, where someone’s liberty (or life) and the demands of justice are at stake. The standard of evidence is ‘beyond reasonable doubt’ - the requirement that any reasonable person could feel satisfied of the verdict. Again, corroboration and reliability are issues, and some forms of evidence are necessarily more reliable, more accurate than others - DNA evidence, for example, is more accurate than fingerprint evidence. A surveillance video showing someone’s face may be more reliable than an eyewitness identification. If multiple sources of evidence - witness testimony, fingerprinting, DNA, exhibits, circumstances - all support a particular verdict, it is highly likely, if not absolutely certain, to be the right one.

Scientific evidence is another matter again. The scientific method sets out clear parameters by which evidence can be examined and hypotheses tested. Yet again, corroboration and reliability are important factors. Corroboration occurs if multiple sources of data support a given hypothesis. Reliability can be established through repeated experiment. If a reliable piece of evidence is discovered that doesn’t fit the hypothesis, the hypothesis is revised, or, in extreme cases, may be abandoned altogether. True scientific theories remain open to falsification - they may never be 100% ‘proven’ (especially as in the case of evolutionary theory, since it deals with past events for which there are necessarily no living witnesses), but the more evidence is gathered which supports the theory, and the longer a theory lasts without falsification (and so far no-one has ever discovered JBS Haldane’s fossil rabbits in the Precambrian), the stronger it becomes as a description of some aspect of how the world works.

In all of the above, corroboration and reliability are key factors. If all the evidence we have adds up to support a particular conclusion, and if the evidence is sound (a genuine diary rather than a forgery, or a genuine fossil rather than a carefully constructed hoax, for example), then it’s likely that our conclusions are correct.

 

[NowAgnostic]

A lot of threads seem to be bogged down on this concept. It might be a good idea to devote one to this problem. What is evidence? What kind of evidence sould one consider and accept? When does the evidence become overwhelming? I am pretty sure there is a lot of subjectivity about it. Something that is accepted as evidence by one person, could be unacceptable by others?

Is it possible to reach some kind of concensus about this? (Personally I doubt it, but it might be interesting to explore).

And this shows why it is necessary to understand statistics and Bayes’ theorem before one can talk of “evidence”. Statisticians have actually devoted themselves to the problem. We have data (D), when does it constitute “evidence” for a model or a hypothesis (M)? There are two ways this can be interpreted. First, Bayes’ theorem says that

P(M|D) = P(D|M)*P(M)/P(D) where P(D) = P(D|M)*P(M)+P(D|~M)*P(~M)

A possible source of confusion here is that P(D) is called the “evidence” in Bayes’ theorem but that is not the connotation of the word here. P(D|M) is the “likelihood”, what is the probability of observing data D given the model M, and P(M) is the “prior probability”, what was the probability of the model before observing the data? The two meanings of “evidence” are:

1. The posterior probability P(M|D) is greater than the prior probability P(M). This doesn’t mean that you would necessarily accept the model, since P(M|D) could still be pretty low, but the data provides some evidence in support, it’s more probable after observing the data than before.
2. The likelihood P(D|M) is greater then the likelihood without the model P(D|~M). What this means is that if you were “on the fence” previously speaking, in other words if P(M) = 0.5, then if P(D|M) > P(D|~M) then P(M|D) > 0.5 from Bayes’ theorem and thus now your model is more likely true than not.

One sometimes also hears the term “model evidence” which is an unfortunate ambiguity; what this refers to is how the likelihood P(D|M) must be computed if the model has parameters P which are optimized to best fit the data.

P(D|M) = (integral) P(D|M,P) P(P) dP or better explained here:

Anyway, this is what “evidence” means.

 

[Areopagite]

I would like to ask the agnostics and atheists whether “evidence” is necessary for the constitution of every kind of knowledge.

And, NowAgnostic, I’ll admit: I have no idea what you just said. If there is any chance you can come down to us mortals and explain your mystical calculations, that would be much appreciated.

 

[Dameedna]

I would like to ask the agnostics and atheists whether “evidence” is necessary for the constitution of every kind of knowledge.

Every kind of knowlege? I doubt we’ve even discovered all the kinds of knowlege that is possible, so I seriously doubt it.

I think the question is a bit too general, but given my above statment, I’d say by default no. We simply couldn’t tell that yet.

 

[Areopagite]

Every kind of knowlege? I doubt we’ve even discovered all the kinds of knowlege that is possible, so I seriously doubt it.

I think the question is a bit too general, but given my above statment, I’d say by default no. We simply couldn’t tell that yet.

It was general because I wanted to know if some agnostics/atheists think that “evidence” is intrinsic to the definition of knowledge. You, I assume, Dameedna, claim that knowledge does not intrinsically pertain to evidence, right? Or if you do, then the answer to my question would be yes.

 

[R_Daneel]

First, a few remarks to NowAgnostic’s excellent post. Most people have no idea about the intricacies of probability theory, so I will bring up a simple example.

Suppose you have two boxes, with red and white balls in them - for example, box #1 contains 10 red balls and 90 while balls, while box #2 contains 90 red balls and 10 white ones. You close your eyes, and reach into one of them and pull out a while ball. The question is: “what was the probability that you chose from box #1?”

Here the hypothesis is that you chose box #1. Does the result of selecting a white ball support the hypothesis? The chance that one ball selected from box #1 will be white is 90%. The chance of one ball selected from box #2 will be white is 10%. Since the boxes are equally possible, the chance that box #1 was used is: The chance (of selecting a white ball from box #1) divided by the chance (of selecting a white ball in general). And that shows that selecting the white ball from box #1 is much more likely than from getting the white ball from box #2. However, this does not exculde the possibility that the ball came from the second box, it just makes this event unlikely. Keep in mind that experiments like this can never “prove” the hypothesis. However if box #1 would not contain any white balls, then the result of having a white ball would disprove the hypothesis that it came from box #1.

But that is not really what I had in mind when presented the OP. I was more inclined to think about what evidence means.

First, there is physical evidence. Something that can be analyzed my people, whether they have a pre-conception or not. Something that is objective. That is the best type of evidence there is.

Then there is the second-hand, or hearsay evidence. Much less reliable, since people make mistakes, have faulty recollection of the events in question, or are simply inclined to lie. This can be countermanded by having many witnessess, who independently describe the same event more or less similarly. The more independent witnessess we have, the more reliable the testimony can be. But it can never reach the reliability of physical evidence. (If the witnesses are not independent, that would lessen their credibility)

Then comes the really farfecthed, not even second-hand evidence. He said, that she said, that he said… where the chain is very long. This kind of evidence is flimsy, pretty much worthless.

Now to Aeropagite’s remark.

The definition of knowledge is information about some object or event. Without some kind of evidence it is called an assertion, nothing more. Suppose I would say that there are some very strange flowers in my backyard, something that differ from every kind of flowers ever described in botany. Would you accept my testimony without evidence? Why?

 

[NowAgnostic]

I would like to ask the agnostics and atheists whether “evidence” is necessary for the constitution of every kind of knowledge.

Knowledge regarding necessary facts, no. It can be deduced from first principles.
Knowledge regarding contingent facts, yes. It cannot be deduced from first principles but has to be observed or inferred.

And, NowAgnostic, I’ll admit: I have no idea what you just said. If there is any chance you can come down to us mortals and explain your mystical calculations, that would be much appreciated.

Well basically, for something to count as evidence (in one sense of the word) in favor of a hypothesis the likelihood of observing the data must be greater assuming the hypothesis is true than assuming the hypothesis is false. It is not enough to state the likelihood is low assuming the hypothesis is false; one must also show the likelihood is greater assuming the hypothesis is true. This point is not appreciated by many.

 

[James_S_Saint]

P(M|D) = P(D|M)*P(M)/P(D) where P(D) = P(D|M)*P(M)+P(D|~M)*P(~M)…
.
Anyway, this is what “evidence” means.

Note that in that formula the “likelihood” is equated to the probability of the truth of the evidence. It is a circular issue.

There are better formulas in the works pertaining to points of coherence to the ratio of a single hypothesis to the many alternate options. Such formulae are far more complex, but consider true probability rather than merely saying that evidence is what we think is probably true.

As someone already pointed out in another thread, when examining for the probability of ID, the numbers become staggeringly in favor of ID and thus such formulas won’t be used by the QM/Atheist/Secular crowd. When they find a formula that shows what it is “supposed to”, they will adopt that one.

 

[Areopagite]

The definition of knowledge is information about some object or event. Without some kind of evidence it is called an assertion, nothing more.

Interesting.

However, I think I see a problem here. You have said that “knowledge is information (about something).” However, at least in how the word “information” is used, information can sometimes be false. You can be given false information. However, can knowledge be false? You can believe that something is true when in fact it’s false. But can you know something is true when in fact it’s false? I don’t think so. (For what it’s worth, I’m not making this up … Plato talked about all this). Thus, equating knowledge with information (which is what you’re essentially doing) seems to be incorrect. Or if you qualify it and say that “knowledge is TRUE information about something” … that doesn’t seem particularly helpful. The question is, what do you mean by information then? What is valid information?

Suppose I would say that there are some very strange flowers in my backyard, something that differ from every kind of flowers ever described in botany. Would you accept my testimony without evidence? Why?

If you are a leading expert on botany and a seemingly trustworthy person, I would probably accept it. Even if you weren’t, but you got some botanists to investigate it and they agreed with you, then I would also accept it, while at the same time never looking at the flowers.

Knowledge regarding necessary facts, no. It can be deduced from first principles.

Right, so some knowledge (i.e. necessary facts) does not require evidence (do you disagree R Daneel?).

Knowledge regarding contingent facts, yes. It cannot be deduced from first principles but has to be observed or inferred.

And so what constitutes observation? (I’m sorry if I’m the most annoying person ever)

Well basically, for something to count as evidence (in one sense of the word) in favor of a hypothesis the likelihood of observing the data must be greater assuming the hypothesis is true than assuming the hypothesis is false. It is not enough to state the likelihood is low assuming the hypothesis is false; one must also show the likelihood is greater assuming the hypothesis is true. This point is not appreciated by many.

Thank you. This sounds pretty reasonable, I think.

 

[James_S_Saint]

However, I think I see a problem here. You have said that “knowledge is information (about something).” However, at least in how the word “information” is used, information can sometimes be false. You can be given false information. However, can knowledge be false? You can believe that something is true when in fact it’s false. But can you know something is true when in fact it’s false? I don’t think so. (For what it’s worth, I’m not making this up … Plato talked about all this). Thus, equating knowledge with information (which is what you’re essentially doing) seems to be incorrect. Or if you qualify it and say that “knowledge is TRUE information about something” … that doesn’t seem particularly helpful. The question is, what do you mean by information then? What is valid information?

Interesting point. If you are a solipsist, there is no such thing as false information. But for real people;

“Truth” (valid information) is merely a set of information that accurately models reality. But any model of reality is inherently limited. This is the whole foundation of Buddhism.

A fact is not really a truth, but an approximation void of detail. Language requires statements that require objects and verbs that are necessarily only approximations to Truth. Truth is only reflected in the Whole (Holy) and is difficult to comprehend at any one time and almost impossible to say.

 

[NowAgnostic]

Note that in that formula the “likelihood” is equated to the probability of the truth of the evidence. It is a circular issue.

No, the likelihood is the probability of observing the data given a model.

There are better formulas in the works pertaining to points of coherence to the ratio of a single hypothesis to the many alternate options.

No, there aren’t.

Such formulae are far more complex, but consider true probability rather than merely saying that evidence is what we think is probably true.

Evidence is not what we think is probably true.

As someone already pointed out in another thread, when examining for the probability of ID, the numbers become staggeringly in favor of ID

No, they don’t.

and thus such formulas won’t be used by the QM/Atheist/Secular crowd. When they find a formula that shows what it is “supposed to”, they will adopt that one.

Codswallop.

 

[R_Daneel]

Interesting.

However, I think I see a problem here. You have said that “knowledge is information (about something).” However, at least in how the word “information” is used, information can sometimes be false. You can be given false information. However, can knowledge be false? You can believe that something is true when in fact it’s false. But can you know something is true when in fact it’s false? I don’t think so. (For what it’s worth, I’m not making this up … Plato talked about all this). Thus, equating knowledge with information (which is what you’re essentially doing) seems to be incorrect. Or if you qualify it and say that “knowledge is TRUE information about something” … that doesn’t seem particularly helpful. The question is, what do you mean by information then? What is valid information?

Information and true information are synonyms. (A false information would be called a lie, or a misunderstanding). So there is no need for qualification. Certainly it is possible to incorrectly interpret the available information and reach an incorrect conclusion, but that is the problem of the evaluating process. Example: mistaking a mirage for late in a hot desert. The information is correct, its interpretation is not.

If you are a leading expert on botany and a seemingly trustworthy person, I would probably accept it. Even if you weren’t, but you got some botanists to investigate it and they agreed with you, then I would also accept it, while at the same time never looking at the flowers.

This is the inherent process of trusting second-hand information. Since the example I presented is competely neutral (you lose or gain nothing by accepting or rejecting the assertion) it is easy to trust the authority. You would lose nothing if they are wrong. You would not use the same process (hopefully) if I asserted that I own the Brooklyn Bridge and offer it to you at a reasonable price. You would (hopefully) demand to see the supporting paperwork to give you factual evidence (proof) for my claim. Would you not?

Right, so some knowledge (i.e. necessary facts) does not require evidence (do you disagree R Daneel?).

Of course I agree. But that is not what I was talking about. No one demands “evidence” for the Pythagoras theorem, they demand proof. There is no evidence in deductive sciences, only in the inductive ones. The title of the thread is “evidence”, not “proof”.

 

[James_S_Saint]

No, the likelihood is the probability of observing the data given a model.

“The probability of observing…”??

The evidence is dependent on the probability that you looked???

Well that could certainly explain a lot.

"Only look over here where the light is. There is **probably **nothing over there or we would have seen it."

 

[NowAgnostic]

“The probability of observing…”??

The evidence is dependent on the probability that you looked???

Well that could certainly explain a lot.

*“Only look over here where the light is. There is **probably ***nothing over there or we would have seen it.”

Let me know if you’re up for a serious discussion on this.

 

[James_S_Saint]

Let me know if you’re up for a serious discussion on this.

To calculate any probability, you have to know both the number of instances being considered and the total number of possibilities along with their comparative significances (if they are not equal).

If evidence is going to be associated with probability, which it should be, the total number of possibilities concerning the evidence must be known and evaluated with respect to each other. Thus the evidence is not the result of any probability, but rather of the possibilities that lead to the probability of the evidence being valid.

All but one possibility must be eliminated in order to achieve certainty. If any more than one cannot be eliminated, their relationship must be compared. Such a comparison is also an issue of possibilities. If one of the remaining possibilities has 2 possible options for each one of the other, then the first has a higher possibility by a ratio of 2:1 and the evidence must reflect that.

The end result is that the evidence is “weighed” so as to reflect a probability of one possibility over another. Neither is rejected. But each possibility had to be measured in relation to all others before the calculation could even begin.

 

[Areopagite]

Information and true information are synonyms. (A false information would be called a lie, or a misunderstanding).

My apologies, but what you say here doesn’t make sense … technically (I am not out of line to demand precision and exactitude in this discussion, am I?).

If Information is the same thing as true information, and yet there can be false information … then there can be true false information. That doesn’t make any sense.

Certainly it is possible to incorrectly interpret the available information and reach an incorrect conclusion, but that is the problem of the evaluating process. Example: mistaking a mirage for late in a hot desert. The information is correct, its interpretation is not.

And yet you said that there is such a thing as false information. Most people would claim that false information can exist. You can say, “I was given false information about you at the time of murder” etc. Information can be true or false, and hence information is not synonymous with true information.

So once again, a qualification or something is needed when you essentially equate knowledge with information.

This is the inherent process of trusting second-hand information. Since the example I presented is competely neutral (you lose or gain nothing by accepting or rejecting the assertion) it is easy to trust the authority. You would lose nothing if they are wrong.

This is kind of the language Pascal’s wager uses … somewhat. If you believe in God, but God doesn’t exist, you lose nothing. If you don’t believe in God, and God does exist, you may lose everything.

Is this a fair comparison to what you’re saying (albeit it may be a bit different, though not in any way that matters as far as I can see).

Am I missing something here?

You would not use the same process (hopefully) if I asserted that I own the Brooklyn Bridge and offer it to you at a reasonable price. You would (hopefully) demand to see the supporting paperwork to give you factual evidence (proof) for my claim. Would you not?

Of course, I could be very paranoid, and think that the paperwork was forged or something. Where does it end? I’ll let you answer that first.

Also, I may be convinced that you own the bridge for other reasons … like you were on TV being interviewed about your experience owning the Brooklyn Bridge or something. Paperwork would not necessarily be required in this case.

Of course I agree. But that is not what I was talking about. No one demands “evidence” for the Pythagoras theorem, they demand proof. There is no evidence in deductive sciences, only in the inductive ones. The title of the thread is “evidence”, not “proof”.

I’m glad you agree. I was just guarding against the error that all knowledge in general requires evidence. I am trying to very carefully see what people think the nature of evidence is by seeing how they think it relates to knowledge. I’m glad we are in agreement that some knowledge does not require evidence.

Now, just in case, I’ll repeat my next question (put particularly to NowAgnostic, but anyone is welcome to join in):

What constitutes observation?