• Quick note - the problem with Youtube videos not embedding on the forum appears to have been fixed, thanks to ZiprHead. If you do still see problems let me know.

Suggestion for a new dowsing testing method

saizai

Graduate Poster
S
Joined
Jul 29, 2005
Messages
1,374
Bayesian network model.

Have the testee blindfolded and be led around a grid that contains variously (predetermined at random) water or not. Have them give a clear definition of what counts as a "yes twitch" of their equipment.

Note every time that it does so and in what grid location it does so.

Then run a Bayesian model of it seeing if there is any predictive power of the twitching for the location.

The math is a bit complex but the testing procedure itself is eminently simple.
 
This is exactly the test Randi uses. Water (or whatever) is hidden in some locations and not in others. The usual test will just hide containers under buckets, but I believe at least one test has been conducted where pipes were buried. No blindfolding is needed since all clues to where the water is are removed (just use identical buckets) and no complex maths is needed, you simply see which percentage of locations they got correct.
 
Cuddles said:
This is exactly the test Randi uses. Water (or whatever) is hidden in some locations and not in others. The usual test will just hide containers under buckets, but I believe at least one test has been conducted where pipes were buried. No blindfolding is needed since all clues to where the water is are removed (just use identical buckets) and no complex maths is needed, you simply see which percentage of locations they got correct.
Click to expand...

Agreed.

Another cool and important feature of Mr. Randi's test method (used e.g. here http://video.google.com/videoplay?docid=7461912885649996034&q=james+randi&hl=en ) is the Base Line Test or Open Test.

Especially maatorc should notice, how "relaxed" and confident the dowsers were after the Base Line/Open Test.

Even Mr. Mike Guska, known around here as edge, was confident in his power/ability after the Base Line/Open Test during his actual prelim years ago. Or Mr. Borer: http://www.internationalskeptics.com/forums/showthread.php?postid=1178020#post1178020

We all know the results: Not one dowser in those tests performed better than mathematical chance would suggest it. Badabing badaboom.
 
The difference in my proposal is that it is (potentially, a lot) more sensitive; it does not require a positive determination from the dowser, merely a readout from their 'instrument'.

This changes a binary dataset to a larger one. More data -> more sensitive, if there's anything to be detected.

Blindfolding would help ensure that the operator doesn't know how far they are from ANY given target, so you eliminate an artificial "closeness-to-some-target" correlation. And they can concentrate more on just what they feel through the instrument.
 
saizai said:
The difference in my proposal is that it is (potentially, a lot) more sensitive; it does not require a positive determination from the dowser, merely a readout from their 'instrument'.

This changes a binary dataset to a larger one. More data -> more sensitive, if there's anything to be detected.

Blindfolding would help ensure that the operator doesn't know how far they are from ANY given target, so you eliminate an artificial "closeness-to-some-target" correlation. And they can concentrate more on just what they feel through the instrument.
Click to expand...

Saizai, perhaps you should take your ideas directly to randi@randi.org

I find the test method executed in the video quite satisfying.
If my uncle wasn't in his 80s, his wife hadn't passed away last year (with him suffering clearly) and his health wouldn't have deteriorated significantly, I'd have proposed the same test already.
 
GK - I'm not interested in dowsing myself; just openly proposing something that came to mind for the benefit of future applicants and general discussion.

What does your uncle have to do with it?
 
saizai said:
GK - I'm not interested in dowsing myself; just openly proposing something that came to mind for the benefit of future applicants and general discussion.

What does your uncle have to do with it?
Click to expand...

I mentioned in another thread that one member of my "gene pool" claims water dowsing abilities; backed up by at least two dozen villagers.

Since I like him very much because he is a very skilled stonecutter (you should see his sandstone projects), for his generation he is a very educated and wise man plus he's a very stand-up guy who never gave me any crap about anything, if there was a problem he got me straight on - it poses quite a hurdle for me to deal with.
I wish I knew a way to get him to do Mr. Randi's test. Unfortunately, he may simply be too old for it (mobility issues). This bugs the crap out of me. I do not consider him a real dowser; he most likely had enough knowledge of the area to have led people to believe he really "dowsed" for the water.



Your proposal surely came out of good spirits. For now, I think the current method used in the videos and the tests done by JREF and e.g GWUP seems to work quite nicely.

If you'd come with a method to definitively, once and for all test for "Gawd", go ahead. Sally'll be all ears. :)
 
GK - Have you read my Atheism thread?

I don't think there is ANY way to test for a God not willing to cooperate. And if said God were willing, I'm sure we would already see obvious evidence. (After all, how hard is it to begat another son to go around doing miracles once every couple generations?) Thus, one must assume that any existant Gods must prefer to work in mysterious and subtle ways, or that they are unable to produce obvious evidence.
 
saizai said:
GK - Have you read my Atheism thread?

I don't think there is ANY way to test for a God not willing to cooperate. And if said God were willing, I'm sure we would already see obvious evidence. (After all, how hard is it to begat another son to go around doing miracles once every couple generations?) Thus, one must assume that any existant Gods must prefer to work in mysterious and subtle ways, or that they are unable to produce obvious evidence.
Click to expand...

"Johnny take a walk with your sister in the rain
Let her talk about the things you can't explain
To touch is to heal and to hurt is to steal
If you wanna kiss the sky, better learn how to kneel
(On your knees, boy)
...
It's alright, it's alright, it's all right
She moves in mysterious ways
It's alright, it's alright, it's all right
She moves in mysterious ways."
U2



(Do line 4 & 5 refer to Aer Lingus?) :D



(Saizai, I do not at all care for "theistalicious" threads outside the Challenge Forum. Doesn't mean I dislike you. See my sig for raisins.)
 
saizai said:
The difference in my proposal is that it is (potentially, a lot) more sensitive; it does not require a positive determination from the dowser, merely a readout from their 'instrument'.
Click to expand...
I see a couple of initial problems with this. I assume that the readouts would be used to give us the unary potentials. How do we get these? I can't think of any dowsers that will tell you "I'll give you an 79% local node belief that water is located here."

Secondly and more importantly, what are the pairwise potentials? How do we determine them? If the water is placed independently at each node and the dowser only detects water at a the local node(as they claim to be able to do) there will be no correlation between predictions and the model will be the same as the one Randi uses.
 
Jekyll - Every twitch adds to the count of all targets, but with extra info on distance from each.

So for each target you have a set of distances.

You compare water targets vs sand targets to see if there's a beyond-chance pattern.


GK - What "theistalicious"ness are you talking about? I'm not a theist...
 
saizai said:
Bayesian network model.

Have the testee blindfolded and be led around a grid that contains variously (predetermined at random) water or not. Have them give a clear definition of what counts as a "yes twitch" of their equipment.

Note every time that it does so and in what grid location it does so.

Then run a Bayesian model of it seeing if there is any predictive power of the twitching for the location.

The math is a bit complex but the testing procedure itself is eminently simple.
Click to expand...

I don't quote understand how the reading of the observer would eliminta the "judging" element in the procedure. How could a dowser be prevented from saying: "You just read me wrong!"?
 
Rasmus said:
I don't quote understand how the reading of the observer would eliminta the "judging" element in the procedure. How could a dowser be prevented from saying: "You just read me wrong!"?
Click to expand...

Very easy: have someone film the whole thing, making it so that only their instrument is in view (not the targets), and a timestamp.

Then have them point out every timestamp at which there is an 'event'. They could even rate the events on a scale of magnitude.

THEN you unblind things and run the calculations.
 
saizai said:
...
GK - What "theistalicious"ness are you talking about? I'm not a theist...
Click to expand...

I did not name you theist.


I have not read your atheism thread because I do not at all care for "theistalicious" threads outside the Challenge Forum.
 
saizai said:
Jekyll - Every twitch adds to the count of all targets, but with extra info on distance from each.

So for each target you have a set of distances.

You compare water targets vs sand targets to see if there's a beyond-chance pattern.
Click to expand...

Why bother?

This smacks of serious amounts of needless obfuscation. The easiest way to control for distance is that you place the buckets sufficiently far apart that the dowser doesn't get any false positives in the open trials.

Problem solved.

On the other hand you still need to find a mapping from twitches to a local belief and choose your distance function. Is it inverse square, Inverse cube or exponential?

I'm sorry to be so blunt but the challengers deserve to be tested for what they say that they can do as quickly and as simply as possible. They do not need to be tested for everything that they might be able to do, but aren't aware of.

That's not to say your ideas don't have merit, in a long series of research trials you could do this work and slowly build up and fit your distance function, but this is work that should be done after dowsing is shown to exist and is not necessary in the first pass at testing.
 
Shy God

Changed my mind about posting here.
 
Last edited:
Isn't the point of the test as it presently is constituted that the results will be immediately and unambiguously clear to both parties without the need of further analysis or interpretation? Isn't that why the Randi challenge doesn't have or require a judge?
 
saizai said:
Bayesian network model.

Have the testee blindfolded and be led around a grid that contains variously (predetermined at random) water or not. Have them give a clear definition of what counts as a "yes twitch" of their equipment.

Note every time that it does so and in what grid location it does so.

Then run a Bayesian model of it seeing if there is any predictive power of the twitching for the location.

The math is a bit complex but the testing procedure itself is eminently simple.
Click to expand...

I'm afraid that I don't see the advantage of this protocol, and I see several disadvantages.

One key aspect you may not be realiziing. The JREF does not test to see "whether dowsers are more accurate than blind chance." As you are no doubt well aware, that's impossible, since you can't prove a negative (the best that could be said is that they are not observably more accurate in the test that was run). It tests, instead, to see whether dowser-claimants can do what they claim to do.

In particular, if a dowser claims to be able to find buried gold bars with 100% accuracy, then we don't need the complexities of Bayesian analysis to test him. Bury several gold bars and ask him to find them. Technically speaking, if he misses one (or has one false hit), he's failed in his claim. But we can afford to be somewhat generous and insist that he get, say, 80% correct. If he finds 80% of the bars -- and "finds" no more than 20% of the distractors -- then he wins.

Of course, if someone came in claiming only 80% accuracy, we would expect a lower rate -- say, 75% -- and we would probably want to bury more bars. But again, why do all the math that almost no one understands?
 
You must log in or register to reply here.

Back
Top Bottom