Do they mean correlation coefficient, the confidence interval, or the p-value? Psychologists often manipulate that aspect. It does say they teamed with applied mathematicians, so maybe the math is valid. It looks like they used a linear combination of facial expressions, voice cues, words and physiological data as the basis for the identifiable space.

There is all this pro-family research out there, presumably due to the difficulty of obtaining funding for anti-family science. It could be that married people are 4 times more likely to suffer high blood pressure, heart attacks, or become victims of domestic violence than single people.