The Music of Chance
How random is this? I’m sitting in an empty classroom at Princeton, a few feet away from Roger Nelson, the founder and director of the Global Consciousness Project. The setting seems plucked right out of A Beautiful Mind, with a thin band of blackboard framing the room, broken by windows overlooking an immaculate campus. There are chalkboard leavings from the previous occupants: elaborate differential equations from a math class, Arabic conjugations from some linguists. And in the nearby engineering quad sits a server, quietly humming away, polling dozens of computer randomizers around the world for Nelson’s grand experiment. Based on that data, he’s written several carefully worded scientific papers that claim that human minds are somehow influencing the random numbers in subtle ways.
The project surprised me when I first heard about it. Nelson uses well-documented statistical methods to look at 10 million datapoints per day from 65 computers around the world between San Francisco and French Guiana. Nelson examines trends that he claims correspond to global events. Many of his best examples tend to be large catastrophic occurrences — embassy bombings, earthquakes, 9/11 — anything that “rends the social fabric and holds human attention.” There’s more: in a November 2006 paper published by the American Institute of Physics, he explores the idea that trends in random numbers begin hours before an event.
I wasn’t sure I believed it, but Nelson impresses me as a straight-up guy in the first few minutes of our meeting. Herbal tea in hand, he walks around looking for a quiet place for us to talk, finally ducking into the empty classroom. He’s in his mid-60s, with a neatly trimmed beard and striped broadcloth shirt, eminently professorial in appearance.
The Global Consciousness Project, or GCP, is unconnected with Princeton University, except that Nelson’s friends in the engineering department allow him to keep a server, and Nelson himself worked as a researcher there for over 20 years, retiring in 2002. GCP is a bootstraps project with a tiny budget, mostly stemming from PayPal donations. Friends of the project designed one of the matchbox-sized gadgets that perform random-number generation. Nelson’s adult son wrote the software to poll the devices, and 65 volunteers around the globe host the generators in their spare bedrooms, university labs, and home offices.
“They’re like one-way modems, blasting out random bytes at 9600 baud,” Nelson says, holding out one of the randomizing devices used in the project. He passes me the tiny gadget, and I involuntarily wonder if there’s something fishy about the apparatus — the image comes to mind of a magician tapping the sides of a wooden cabinet before the lady climbs in. I know he wouldn’t appreciate the comparison.
Nelson is diligent in his research. Over the last few years, he’s been the focus of respectful, if cautious, articles in The New York Times and Time, the latter going so far as to point out that many of our greatest scientific breakthroughs were once ridiculed. The November paper came out of a symposium on “Retrocausation” — organized by the respected American Association for the Advancement of Science (AAAS) — exploring the theory that time flows symmetrically.
“The theoretical questions about the direction of time go back 150 years,” says Daniel Sheehan, organizer of the AAAS symposium and a physics professor at the University of San Diego. “Nelson and other experimentalists are starting to put some experimental background to the idea that retro-causation may be a real phenomenon. The experimental evidence is not all in, but his results are tantalizing.”
Roger Nelson’s Global Consciousness Project searches for meaning in random numbers.
Photograph by Erika Larsen
Nelson at his office in Princeton, N.J. “Sometimes those meaningful coincidences are so potent that it’s really hard not to believe that there’s something, some kind of connection there that is real and meaningful.”
AT THE FOUNDATION OF GCP’S NETWORK are 65 random-number generators. Truly random numbers are important to hard science, and these common devices harvest the white noise from various physical sources, like the heat coming from a resistor, the time intervals between particles arising from radioactive material, or electrical fluctuations in analog components.
The device used by the GCP is called Orion ($550, randomnumbergenerator.nl) and uses two diodes to generate a chaotic signal. The Orion sends current the wrong way through a diode so that only a few miscellaneous electrons make it across the junction. A very weak, fluctuating current results. The circuit amplifies the signal so that a microcontroller can sample and compare it with a threshold value. Current above the line receives a “1” and current below, a “0.” The circuit compares the bit streams against stored random data using an exclusive-or gate (XOR) to further eliminate mechanical drift. (The GCP network also uses a similar randomizing device with a field effect transistor as the entropy source.) Host computers on the GCP network sample 200 of these bits per second and add them together for sums of between 0 and 200. Since it’s random data, the sums usually come in around 100, with a few outliers in the 80s and 120s.
A few minutes later, the data pours into the Princeton server. The ensuing analysis is a formal, deliberate process. Nelson emphasizes that he conducts research as a scientist, and follows widely accepted research methods. In general, to make a scientific claim, one must set up a formal hypothesis and a null hypothesis before looking at experimental data. Then, he’ll fire up his Linux machine to feed the data through a statistical analysis package called XploRe.
As an example, let’s pick a day in the life of the GCP network — a biggie, September 11, 2001. Nelson, who lives with his wife on a quiet Princeton street and conducts all research at home, was casually watching the network from the five computers he has running all the time. But there didn’t seem to be anything going on in the world worth analyzing. Then an airplane hit.
“By the time the first building collapsed, I realized we had a global event maybe bigger than we’d ever looked at,” he says. He started outlining a formal hypothesis with an expected result and the formal time period for the experiment — a window starting ten minutes before the first plane hit to four hours after. With those parameters fixed, he was then free to look at all the data surrounding the event for a more exploratory look. Nelson has since examined the 24-hour period starting at midnight of Sept. 10. Graphed against a pseudo-random line, the data shows a striking upward trend. The calculated probability of that trend is 1 in 100,000. What’s more, the break from chance begins at 4 a.m., hours before impact.
As Nelson himself points out, the possibility of network precognition is clearly exploratory. “There’s no way you can use that as evidence in the scientific sense,” he says. Nelson frequently underscores that he wants to be critiqued principally on his formal hypotheses. But the results of the formal hypothesis are actually pretty boring. In more than 200 time parameters studied under hypotheses, only 21 show a significant deviation against a distribution of purely random events. For example, the formal September 11th events showed 2.8 percent odds of occurring against chance. Some larger odds come from other formal studies such as the May 2006 Indonesian earthquake (.2 percent); the September 2004 Chechen hostage standoff in Russia (1.2 percent); and the March 2000 Papal visit to Israel (.8 percent). Given the number of tests, however, it’s possible that the GCP researchers are getting lucky by picking small, anomalous windows in random data.
HOW THE GCP WORKS
1. Random-number-generating devices (right) are placed in spots around the planet. They connect via serial cables to PCs in home offices, workplaces, and the basements of academic institutions.
2. The RNG gadget is the size of a matchbox. It works by sampling a fluctuating signal inside the device, generating a random stream of 1’s and 0’s. Each host computer grabs 200 samples per second, adds them together, and sends the sums off in batches every 10 minutes to a central computer in Princeton.
3. At the GCP headquarters, a Linux PC collects the data and analyzes it for deviation from normal statistical probability.
“Nelson and other experimentalists are starting to put some experimental background to the idea that retro-causation may be a real phenomenon. The experimental evidence is not all in, but his results are tantalizing.”
“Statisticians will say, ‘Well, there are so many opportunities to find anomalies,’” says Nelson. “But sometimes those meaningful coincidences are so potent that it’s really hard not to believe that there’s something, some kind of connection there that is real and meaningful.”
MANY UNIVERSITY ACADEMICS HAVE dismissed GCP on its face, but few have been willing to crunch the numbers themselves (the raw output is available on noosphere.princeton.edu). A common critique involves electromagnetic waves — doesn’t the use of cellphones, television transmission trucks, and two-way radios increase during world tragedies? Wouldn’t that addle the randomizing hardware? Nelson counters that the devices are carefully shielded. Besides, each device does generate random data during the experiments. It’s the statistical trends among the 65 devices that are the subject of study.
There have only been two serious looks by scientists critical of the work. S. James P. Spottiswoode, chief statistician for entertainment-monitoring company Neilsen, reproduced the computer analyses of the formal 9/11 findings, and confirms Nelson’s result. He notes, however, that any event with a 2.7 percent chance of happening shouldn’t surprise anyone.
Nelson’s other auditor, Jeffrey Scargle, a staff astronomer at the NASA Ames Research Center in California, contends that the math Nelson uses to plot graphs makes minor trends look more meaningful than they are. Scargle also questions the XOR operations inside the random-number generators. “They’re scrambling the bit stream in order to get rid of drifts in the instrument, but by doing that they’re getting rid of any signal that would possibly have a direct effect on the random-number generator,” he told me. Nelson himself clearly seems interested in mind-quanta interaction. Why not let the bits fall as they may?
Nelson responds by saying that the XOR doesn’t entirely remove the influence of the original quantum events; it merely smoothes out the data. “He thinks I’m throwing the baby out with the bathwater,” he says. “But from laboratory tests, we know we still have some baby.”
Nelson’s critics also dismiss the work as carefully selecting data to guarantee a positive result. The exploratory tests boast larger departures from chance, but these sessions are often conducted by laying out all the statistical deviations on a timeline and zeroing in on the surprising ones. Nelson deflects this by saying the exploratory claims are clearly labeled as such. “It’s unacceptable that they criticize the GCP’s hypothesis testing on the basis of postfacto analyses,” he says calmly. “These are independent projects. It would be stupid not to see what the data looks like.”
While we talk, I try to drill down on the exact mechanism of the phenomenon. “During these global events, do you think the devices are not performing in a scientific manner?” I ask.
“That’s a pretty logical kind of thought,” he says affably. “It sounds as if you’re saying human consciousness is changing the way these devices operate. But probably not.” Nelson suggests the possibility that the numbers aren’t changed at all while they are being generated, but don’t become fixed on computer hard drives until they are analyzed. Then he stops talking and sort of shifts in his chair. “This is a lot of hand waving, for sure,” he says.
Clearly, there are plenty of other people curious about global consciousness, too. Traffic to the GCP website nearly doubled to 60,000 last May during a busy month that included the Indonesian earthquake and Earth Day. And on Dec. 22, 2006, the pro-sexual liberation group Baring Witness held a “Synchronized Global Orgasm” and used the network data to explore the impact of mass nookie on GCP’s spreadsheets. The GCP reported a “positive outcome, but it is a mild trend“ from the experiment. “The picture has a feeling of promise, but it is not possible to know whether the trend is evidence of a signal or not.”
People are encouraged to try harder next year.
Bob Parks is the the author of Makers: All Kinds of People Making Amazing Things In Their Backyard, Basement or Garage, published by O’Reilly Media.