Transcript Recht

Lessons From The PEAR Lab
Exploring the Possibility of
Anomalous Bias in Scientific
Measurement
The Facts
• PEAR has shown that people can
probably bias random machines
• Through conscious intention (REG)
• Through “subjective resonance”
(FieldREG)
The Big Question
Do scientists influence
measuring devices?
What We Need to Know
•
•
•
•
•
Does anomalous bias exist?
When does it happen?
How big an effect is it?
How can we detect it?
How can we prevent it?
The Sheep-Goat Effect
• Many studies in parapsychology tell us
that people who believe in ESP do better
than chance on tests of ESP
• Nonbelievers do same or worse than
chance
• It is possible (likely?) that this applies to
effect seen by PEAR and thus to
anomalous bias
The Sheep-Goat Effect
•
•
•
•
Some Logic
Say the sheep-goat effect does not apply
Therefore all scientists anomalously bias
their results
But measurement works just fine
So the effect must be small or nonexistent
The Sheep-Goat Effect
•
•
•
•
More Logic
Say the sheep-goat effect applies
We have no idea how big the effect is
As soon as scientists begin believing we’ll
find out (it may be catastrophic)
Therefore PEAR needs to solve this
problem or else stop publishing lest they
convince their colleagues!!!
Scale
• PEAR sees effects of a certain magnitude
(roughly 2 bits off expectation per
thousand bits)
• This is at best an approximation for
anomalous bias
• Variable size?
Scale
Any constant-size bias will eventually
become a problem
BUT!
There is reason to believe bias size will be
smaller for more precise instruments
If this is so, there is no problem
Scale
• Say you measure some quantity to be n
with experimental error σ
• An additional measurement with error ≈ σ
that is within (n - σ , n + σ ) does not tell
you much
• Small information gain → small reduction
in entropy → small energy cost
Scale
• It appears that this effect is low energy
• Thus one shouldn’t be able to bias
measurements much farther than σ away
from n
• By the definition of σ, n is within a few σ’s
of the true value, n0
• So the bias size with respect to n0 should
scale with σ as well
Detection (briefly)
• Volitional (like REG experiments)
– Studies suggest changes in variance
– Catch with statistics
• Situational (like FieldREG experiments)
– Can change direction, stop and start
– Catch with REG array
Conclusion
• Unless anomalous bias has a very specific
set of properties, it will be a serious
problem
• This will be aggravated by the lack of time
or distance dependence
• Also perhaps by high stakes and large
research groups
Conclusion
Even if combining the probability of PEAR’s
effect being real with the chance that
anomalous bias has bad properties yields
a 1% chance of trouble
THAT IS TOO LIKELY TO
IGNORE!
Lessons From The PEAR Lab
Exploring the Possibility of
Anomalous Bias in Scientific
Measurement