Factoring out the ad corpus yelping, the interesting counterpoint is that suppressing the risk signal isn't a good thing if it doesn't suppress the risk.