Back in June we posted about the bullshit metre. MIT measured what happens when you outsource your judgement to an AI. People got 21% better at spotting fake news with the help. Take the help away and they scored 15 points worse than where they started. A quarter of them felt they were improving while their scores fell.

The warning was that the skill is not transferable. The dependency is.

A new Penn State study just landed the second half. 62% of US adults report knowing little or nothing about basic statistical concepts. Not a lab sample. A nationally representative survey.

So here is the full picture. Half the population never had the detector. The other half is training itself to stop using it.

And the environment both groups live in is flooded with confident numbers. Cheating rates and market shares and temperature records, all served with the same authority, aimed at people who cannot check any of them.

You do not need to be lied to. You just need to be unable to check.

The bullshit metre was a skill built by growing up next to the medium that produces the fakes. The literacy underneath it, the ability to read a number and feel its weight, was always rarer than we pretended.

Now both are going at once. That is not a character flaw in the audience. That is a design feature of the machine.