Event Selection Procedures

Event selection is a critical aspect of the GCP experiment that merits discussion. Because the project is unique, with no information on relevant parameters available at the outset, procedures for choosing and specifying events had to be developed from scratch, beginning with sheer guesses about what might define a suitable, informative situation or event.

The first step was to define our experimental hypothesis, but this could not be strictly formulaic since our target system, global consciousness, can only be usefully defined in operational terms. Our approach has been to work with a general hypothesis describing a range of conditions rather than a narrow set of parameters:

Periods of collective attention or emotion in widely distributed populations will correlate with deviations from expectation in a global network of physical random number generators.

The hypothesis avoids premature over-specification, but includes the main elements we wish to test for, namely, global correlations between collective conscious activity and the material world as represented by the physical RNG network. Experimentally, this general hypothesis is instantiated in a series of specific, rigorously defined hypothesis tests, each of which is compatible with the general statement. Technically, we propose a composite hypothesis which formulates our broadest guess of how global mind-matter correlations might be defined for the RNG network. We then proceed experimentally with a series of replications using simple hypotheses which are fully specified and can be compared quantitatively against the null hypothesis.

Previous research on group consciousness suggested that synchronization or coherence of thought and emotion would be important, so we focused on major tragedies that make the news, and traditional celebratory events that bring great numbers together in a common focus. In addition to events we expected would produce correlations, we also have chosen to identify some less promising events in order to test the range of criteria and allow assessment of a broad spectrum of potentially relevant parameters. The selection procedure is open to a wide variety of events, allowing exploration to learn what matters while providing for rigorous testing.

To set up a formal test, we first identify an engaging event. The criteria for event selection are that the event provides a focus of collective attention or emotion, and that it engages people across the world. We thus explore events of global character, but allow for variability in their type, duration, intensity and emotional tone. In practice, events are selected that capture news headlines, involve or engage millions of people, or represent emotionally potent categories, e. g., great tragedies and great celebrations.

Once an event is identified, the simple hypothesis test is constructed by fixing the start and end times for the event and specifying a statistical analysis to be performed on the corresponding data. These details are entered into a formal registry before the data are extracted from the archive. We select and analyze an average of 2 or 3 events per month. For a description of the 2010 state of the art, see Event Specification Examples.

While many observers assume we can and should follow a fixed prescription to identify global events this is not straightforward, nor is it actually appropriate for a new research paradigm. We need to learn what matters, and we can do so only if we are respectful of our state of relative ignorance. To give specific examples, we could select a disaster if it results in, say, more than 1000 fatalities. But this would exclude slow moving but powerfully engaging events such as volcanic eruptions or major hurricanes, and it would fail to identify emotionally powerful, extremely important incidents like the politically disruptive attack that destroyed the Golden Dome Mosque in Iraq. What we try to do is to identify, with the help of correspondents around the world, events that can be expected to bring large numbers of people to a shared or coherent emotional state. The following is a partial, illustrative list of criteria and examples:

Suddenness or surprise.

Terror attacks, especially where they are not usual.

Fear and compassion.

Large natural disasters, typhoons, tsunamis, earthquakes.

Love and sharing.

Celebrations and ceremonies like New Years, religious gatherings.

Powerful interest.

Political and social events like elections, protests, demonstrations.

Deliberate focus.

Organized meetings and meditations like Earth Day, World Peace Day.

We have progressively worked toward standardization of the event selection and definition, and for some kinds of events, pre-defined parameters can be applied. For example, events that repeat, such as New Years, Kumbh Mela, or Earth Day, are registered with the same specifications in each instance. For unexpected impulse events (e.g., earthquakes, crashes, bombs) the most used protocol identifies a period beginning half an hour to an hour before the moment of occurrence, followed by several hours (e.g., 4 to 8), depending on the type of event, to allow time for recognition and the spread of news. (Recent analyses of time structure in the database suggest we may not need so much flexibility, and we are testing a simpler event specification: 6 hours beginning just prior to the time of the event.) We specify a relatively brief period to sharpen the analytical focus and minimize overlap and confusion with other potentially effective events.

About half the events in the formal series are identifiable before the fact; the accidents, disasters, and other surprises must, of course, be identified after they occur. To eliminate a frequent misconception, we do not look for spikes in the data and then try to find what caused them. Such a procedure, given the unconstrained degrees of freedom, is obviously inappropriate. There is no data mining, and there is no post hoc inclusion or exclusion of events. All events are entered into the formal experiment registry before the corresponding data are extracted from the archive. analysis for an event then proceeds according to the registry specifications, yielding a test statistic relative to the null hypothesis. These individual results become the series of replications that address the general hypothesis and ultimately are combined to estimate its likelihood.