Daily experiment
Every day at noon UTC, for thirty minutes, anyone who wants to can take part. Each person chooses a number, a shape, a colour and a word.
Afterwards we measure how much those choices overlap, and compare it against how this group normally chooses. Every session is published, whatever it shows. We have no prediction and nothing to prove. The point is to run it long enough to find out.
Today's session
Not enough participants
0 people took part. A score needs at least 8. Below that the result is mostly noise, and with a handful of people it could reveal what an individual chose.
The session runs again tomorrow at 12:00 UTC.
0 scored sessions · 1 too small to score
The shaded band is where a day lands when nothing unusual is happening. Its width is set by how many people took part, so it narrows on busy days: the same score counts for more when more people produced it. Breaks in the line are days that did not reach eight participants, left blank rather than drawn as zero, because no data is not the same as no synchronicity.
The session runs from 12:00 to 12:30 UTC, every day, without exception. One fixed half hour is what makes it a shared moment rather than a form people fill in whenever they get round to it. Noon UTC is the compromise that puts the most people awake at once: evening across Asia and Australia, the middle of the day in Europe and Africa, early morning on the American east coast.
Does a group of strangers, choosing at the same moment, agree more often than chance allows? Every session is scored and published, whatever it says.
If collective attention is a skill rather than a fixed property, it should improve with repetition. A daily session gives it somewhere to improve, and the running trend is where that would become visible.
Those two aims pull in opposite directions, and the design has to protect the first from the second. Practice is why the session is daily. Measurement is why nobody is ever told what the group chose, and why the options are reordered for every participant.
Nothing here is required and there is no way to do it wrong. It is only worth saying because how you go about choosing changes what is being recorded.
Give yourself a moment before choosing. The window is half an hour, so there is no advantage in being quick.
Whatever comes to mind before you start weighing the options.
If you catch yourself swapping an answer for a more sensible one, keep the first.
Working out what other people are likely to choose measures something else entirely.
During the half hour you pick one number, one shape, one colour and one word. There is no right answer. You can change your mind until the session closes.
After the window closes we compare every participant with every other one, and count what share of all those pairs made the same choice. With 60 people that is 1,770 comparisons per dimension. Nobody is sampled and nothing is estimated: it is the exact figure for that session.
People are not random. Ask a crowd to choose freely and some options come up far more often than others, and it is the same ones every time. Scored against a coin flip, an entirely unsynchronised crowd would appear to synchronise every single day. So the comparison is not against chance in the abstract. It is against how this community has actually chosen across every past session, and the score is what is left over.
Zero means the day looked like people always look. Above zero means the group converged more than usual. It can also go below zero, meaning the day was more scattered than normal. That is a real result too, and it gets published.
Nothing is hidden here. For one dimension, with N participants and ni of them choosing option i:
The threshold drawn on the chart is 52 ÷ N, the level below which 95% of results fall when nothing unusual is happening. It comes from simulating this exact set of dimensions twenty thousand times at each participant count, and it tightens as more people take part.
If we published which option the crowd favoured, far more people would choose it tomorrow, and the experiment would be measuring memory instead of synchronicity. Individual choices are never published, not during the session and not after. Only the score leaves the room.
Every participant sees the options in a different order, fixed for the session but different for each person. Layout is persuasive. If everyone saw the same first option, the crowd would drift toward it together, and that would look like synchronicity. Randomising per person turns that pull into noise.
Below 8 participants the number is mostly noise. With two people, a single coincidence reads as total agreement. It also works backwards: in a very small group, a high score tells you what the others picked. Those days stay blank rather than showing a figure we do not trust.
A single day above the threshold is not evidence of anything. Run enough sessions and some will land high by chance alone, which is exactly what the threshold is there to tell you. What would be meaningful is a persistent shift across many sessions. That is why the running trend matters more than any one result, and why every score is published, including the unremarkable and the negative ones.