The article says “…had professors assess whether responses might mislead or confuse students.”
professors flagged AI responses as pedagogically harmful only 3.5% of the time, compared to 12% for peer-written answers.
Which makes me think that real people came to some conclusion, sometimes biased or wrong, but AI could have produced inconclusive inflated perhaps-maybe-sometimes text (which it would be good at) 96.5% of the time. Response not being harmful doesn’t mean it’s good.
The article says “…had professors assess whether responses might mislead or confuse students.”
Which makes me think that real people came to some conclusion, sometimes biased or wrong, but AI could have produced inconclusive inflated perhaps-maybe-sometimes text (which it would be good at) 96.5% of the time. Response not being harmful doesn’t mean it’s good.