[extropy-chat] Scientific standards of evidence

[Eliezer S. Yudkowsky]

Mitchell Porter wrote:

> 
> Eliezer said
> 
>> It doesn't matter whether the explanation I gave is correct.  There's
>>  no ESP in the universe, no paranormal phenomena, no gods, no demons.
>>  It's just us within the laws of physics.
> 
> There is no rational necessity for such dogmatism, especially when (i)
> today's standard-issue laws of physics already contain a mechanism for
> nonlocal correlation,

(Disagree; Everett.)

> and (ii) anomalous coincidences are a very common experience.

*Coincidences* are a very common experience.  The Bright Hypothesis is 
that there has not been one single *anomalous* coincidence in the entire 
history of the universe to date.

> Considerations of human irrationality cut both ways here.
> 
> I recommend the following essay by C.D. Broad: 
> http://www.ditext.com/broad/rprp.html

Note, incidentally, the credence in Rhine's data, which was later 
debunked, and other data which repeated experimentation failed to 
replicate.  Yet another success for the Bright Hypothesis.  It takes great 
daring to stand by your principles in the teeth of early data, but to do 
so and succeed is a tremendously impressive accomplishment for a theory.

The Bright Hypothesis has triumphed over anecdotes, public wisdom, 
overeager researchers, wishful thinking, and outright experimental fraud. 
  It is one damn strong hypothesis.

> None of Broad's principles are logical necessities. They are all 
> hypotheses. Furthermore, we know how to build physical models which
> violate the principles in the first category (quantum nonlocality,
> closed timelike curves), which means that we can describe, in the
> abstract, material cognitive systems which violate principles from the
> second and the fourth categories. Even if a Bayesian reasoner only
> considered mathematical models already devised by human beings, it 
> would have to assign a nonzero probability to the reality of paranormal
> phenomena (and not just in the form of a world-as-simulation
> hypothesis).

Yes, that we live in a magic-free universe is a hypothesis.

It happens to be an overwhelmingly supported hypothesis, and I am willing
to put my weight down on it, and to wield it as my understanding.

The weight of the alternatives is not zero.  It is small enough that, 
confronted with supposed anomalous coincidences, I am willing to deny the 
data, as history suggests will turn out to be the correct course of 
action.  I deny the data openly rather than covertly, so that I can keep 
track of what it is I have denied.  For example, I earlier denied the data 
on an experiment purportedly showing 50% recovery rates for prayed-for 
patients versus 25% recovery rates for un-prayed-for patients: as I 
commented at the time, a nice non-marginal result, blatant effect size, 
and clear and straightforward claim, far healthier for debate than all the 
marginal claims of conventional psychic science.  Of course, as the Bright 
Hypothesis predicts, the study failed to replicate; and so that is one 
more denial off the stack and one more triumph of the Bright Hypothesis.

History shows that the Bright Hypothesis is stronger than anecdotes and 
early experiments.  Needless to say, not even the Bright Hypothesis would 
be stronger than an experiment with a non-marginal effect size and 
non-marginal statistical significance that replicated several times in the 
teeth of surveillance by professional magicians.  Needless to say, I 
predict, wielding the Bright Hypothesis as my understanding, that no such 
experiment will ever occur.  I can take my weight back off the Bright 
Hypothesis, given a damn good reason to do so; no such reason has been 
forthcoming, and on the Bright Hypothesis none ever will.

-- 
Eliezer S. Yudkowsky                          http://singinst.org/
Research Fellow, Singularity Institute for Artificial Intelligence