The famous accuracy figures are real, and they measure something other than prediction. What happens when a divorce model is finally tested on data it has not already seen.
Fifteen leading contemplative scientists published a paper warning about the hype around their own discipline. Almost nobody reported it — including the finding that when you finally ask people properly, most report side effects.
In 1981 the United States recorded 2,422,000 marriages and 1,213,000 divorces. Divide one by the other and you get 50.1%. It was also the peak year. The statistic has not been updated since.
When researchers finally put sensors on employees before and after a redesign, face-to-face interaction fell by about 70%. The study did not measure productivity — and after six decades of fit-outs, almost nothing has.