She also studied harassment and found that vitriolic threats from a handful of accounts were enough to drive users off the platform, even if their overall experiences were positive—powerful data that she used to convince Twitter executives that good speech wouldn’t naturally drown out the bad.
