The Global Consciousness Project

We discussed this on Skeptiko and at the time I had asked one of the researchers if the entire set of numbers was at expectation and IIRC the answer implied that it was, although I can't recall exactly. In any event, that is something to check.

If the RNGs are producing all these non random substrings based on global consciousness then over the vast sample size the 0s and 1s should vary wildly from a random distribution. If the entire set follows a random distribution then supposed substrings that look non random are probably just noise.

At least, that is my understanding.

It's true that if the network is detecting "global consciousness", and if it is producing quite a large deviation from expectation during the chosen events, then one might expect a significant deviation at other times too. That's not the case, which is one of the facts that's consistent with Peter Bancel's eventual conclusion that the results reflected experimenter psi. 

But the cumulative deviation from expectation during the chosen events obviously remains hugely statistically significant.

It's true that if the network is detecting "global consciousness", and if it is producing quite a large deviation from expectation during the chosen events, then one might expect a significant deviation at other times too. That's not the case, which is one of the facts that's consistent with Peter Bancel's eventual conclusion that the results reflected experimenter psi. 

But the cumulative deviation from expectation during the chosen events obviously remains hugely statistically significant.

Another interesting discovery Bancel made concerned the behaviour of the network just outside the periods specified for testing. Sometimes these began and ended at midnight, but in other cases, he found that just before the beginning of the periods and just after the end of the periods, on average the statistic expressing correlations between the RNGs was lower than expectation. He also found that the deficit outside the testing periods balanced the excess within the periods. He considered this to be strong evidence that the experimenter was exercising psi when choosing the start and end points of the test periods.

For the periods beginning and ending at midnight, whose start and end points were fixed, there was also an excess of the correlation statistic within the periods. On Bancel's interpretation, presumably the experimenter was exercising psi when deciding whether or not to include the relevant event in the experiment at all.

Another interesting discovery Bancel made concerned the behaviour of the network just outside the periods specified for testing. Sometimes these began and ended at midnight, but in other cases, he found that just before the beginning of the periods and just after the end of the periods, on average the statistic expressing correlations between the RNGs was lower than expectation. He also found that the deficit outside the testing periods balanced the excess within the periods. He considered this to be strong evidence that the experimenter was exercising psi when choosing the start and end points of the test periods.

For the periods beginning and ending at midnight, whose start and end points were fixed, there was also an excess of the correlation statistic within the periods. On Bancel's interpretation, presumably the experimenter was exercising psi when deciding whether or not to include the relevant event in the experiment at all.

Sounds like he's saying the data was noise but decision augmentation theory used psi to pick a period that expressed a period that gave the correlation.

Sounds like he's saying the data was noise but decision augmentation theory used psi to pick a period that expressed a period that gave the correlation.

Yes, that was his eventual conclusion. But before that he had spent a long time looking for structure in the data, and the earlier paper to which Doug posted a link shows some of those results:
https://www.researchgate.net/publication...JECT_draft

Although some of the structure seemed significant, and wouldn't have been consistent with the decision augmentation interpretation, he seems ultimately to have decided that the structure had just arisen as a result of post hoc testing of multiple hypotheses.

I'm not sure about that. For example, the diurnal variation mentioned by Doug is very strong. If the average value of the correlation statistic is significant at Z=7, is it so easy to dismiss the diurnal variation in the statistic, which is roughly going between 55% and 145% of the average? (Unfortunately the variable Bancel chose to express the diurnal variation makes it a bit awkward to work out its statistical significance.)

[Image: Diurnal.jpg]

Haven't read it, but that's what you would expect to see with power fluctuations... when people are awake there is a lot of fluctuing demand/supply in the power supply grid etc. It's much quieter when people are asleep and not consuming.

I wonder if there have been any tests that correlate randomness with power-line noise? 

I am having a hard time coming up with a reason to think that normal AC line noise would have a measurable impact on software created randomness. We are not talking about an analog circuit that would be directly impacted by things like spikes or harmonics. 

If the power is clean enough to keep the OS is up and running, the software/CPU doesn't give a whit about the power supply sees on it's input. A serious enough power disturbance will cause CPU and other hardware related failures that will be quite obvious (reset, hang, blue screen etc).

Also any spurious error that did occur somehow, would tend to be "random" in nature anyway and so would not really skew the results.

I wonder if there have been any tests that correlate randomness with power-line noise? 

I am having a hard time coming up with a reason to think that normal AC line noise would have a measurable impact on software created randomness. We are not talking about an analog circuit that would be directly impacted by things like spikes or harmonics. 

If the power is clean enough to keep the OS is up and running, the software/CPU doesn't give a whit about the power supply sees on it's input. A serious enough power disturbance will cause CPU and other hardware related failures that will be quite obvious (reset, hang, blue screen etc).

Also any spurious error that did occur somehow, would tend to be "random" in nature anyway and so would not really skew the results.

The random numbers aren't being generated by software. The input comes from devices generating electronic noise. There's a detailed description here:
http://noosphere.princeton.edu/reg.html

I have a very silly question. Maybe too silly... but still...
If RNGs can be affected by non local consciousness (i.e regardless of distance) how the heck do we create a proper control?
Wherever you go you will be affected by the "consciousness field", there's no hope to see how these devices operate when "no consciousness" is interfering.

Sorry, maybe I misunderstand the GCP. After many years I still don't think I get it.  

Cheers

Also... when it's claimed that global consciousness affects RNGs... what propery of consciousness exactly would be at work? Awareness? Emotions (and which one(s))? Will power? Attention? Intention? All of them?

Thanks

The random numbers aren't being generated by software. The input comes from devices generating electronic noise. There's a detailed description here:
http://noosphere.princeton.edu/reg.html

Oh, OK. 

Thought it was via s/w they way we used to do it back in the day. But yes, now that you mention it, those were pseudo-random.

Still wondering however how power noise affects this stuff, and whether tested and quantified.

Oh, OK. 

Thought it was via s/w they way we used to do it back in the day. But yes, now that you mention it, those were pseudo-random.

Still wondering however how power noise affects this stuff, and whether tested and quantified.

One relevant point is that various XOR masks are applied to the raw stream of bits produced by the hardware, swapping certain bits between 0s and 1s in order to eliminate any systematic bias towards one or the other. I was going to post a bit about that when I got a chance, as it's also been argued on that basis that this isn't a kind of field effect owing to consciousness (or at least not the naive concept of a field effect), because the XORing will be different for the different random number generators, and it's the output from that process that is found to be correlated. But equally it makes it even more difficult to see how the output of different random number generators could be correlated because of conventional physical processes.