[extropy-chat] Bayes, crackpots and psi

[Damien Broderick]

At 08:18 AM 12/20/2004 +0000, BillK wrote:

>Ertel is well-known to Randi. He has been trying for years to convince
>people that bits of astrology have a scientific basis (and failed).

You mean the Gauquelin `Mars' effect? (Unknown to any traditional astrology 
system.) The one where CSICOP was caught fiddling the data in a lame 
attempt to falsify their own successful replication? It is to laugh. (I 
recall Eliezer some years ago citing the sTARbaby scandal as an example of 
how NOT to do debunking.)

http://cura.free.fr/xv/14starbb.html

>Now he has gone back to the 1970's (pull a rabbit out of a hat style).

What hat? What rabbit? Oh, you mean a haptic protocol. Haptic, Bill, not hatic.

>His students pick a number one to five, put their hand into a bag, and
>pull out a table tennis ball (or a pearl), show it to the recorder and
>put it back in the bag. The possibilities for elementary
>sleight-of-hand (or other trickery) should be obvious.

Mention some examples that Ertel hasn't guarded against. There probably are 
some. Suppose the students palmed a ball somehow and brought it out again 
next time? Then there'd be a visible excess of instances where the same 
number was called twice in a row, the second time correctly. Actually, as 
Ertel's paper shows, there are slightly *fewer* cases of this happening.

I'm sure there are other fail points. Luckily, skeptics don't have to 
enumerate them because it's *obvious* that they must be in play, since we 
know without looking that this stuff is such *bullshit*. Unlike, 
interestingly enough, uploading, cryonics, superhuman AI, time-reversed 
particle theory, and all those topics also derided by most 
feet-on-the-ground skeptics. (This is not an argument for psi, just a 
sociological observation.)

>This type of 'evidence' is worthless. You have to physically separate
>the guesser from the numbers.

It's true that such `evidence' has weaknesses if regarded purely as 
evidence; Ertel is looking to find regularities in a process he already 
regards as established. But his argument, which seems plausible, is that 
clinical lab tests eventually extinguish or suppress the subtle phenomena 
that experimenters are seeking, while putative ethologically grounded `real 
world'-style uses of psi (hand reaching for something unseen, becoming 
aware of a hidden watcher, etc) should be more easily elicited. Obviously 
this makes any experiment more vulnerable. It's the task of the 
experimenter to guard against tampering.

>If these students actually can predict
>numbers in advance. then the lottery is a really excellent
>demonstration of their psi powers.

Not so, because even Ertel's best subjects don't succeed 100% of the time. 
Redundancy can in principle concentrate a small effect size, but it starts 
to get very cumbersome. You can code for six winning Lotto numbers using 23 
binary digits, but how many tedious or artfully diverse repetitions are 
required, given a small effect size, to make identification of each bit 
perfectly accurate? Answer: a hell of a lot.

But this is quibbling, isn't it? If people have `psi powers', they should 
be able to levitate and see the lottery numbers in dreams whenever they 
want and read each others' minds like a conversation, shouldn't they? Well, 
no, actually. Only in the comics.

Damien Broderick