The talk is "Tinkering and Gender in End-User Programmers' Debugging" given by Laura Beckwith.
In the talk, the speaker cited previous evidence that women have a lower sense of self-efficacy when debugging; the low self-efficacy of women leads to low feature acceptance; and that men are more likely to tinker, whereas women are more likely to follow step-by-step instructions. So they added various supportive features to a debugging environment, intending to help women programmers.
But they were surprised by their results. The more women tinkered with a program, the more bugs they fixed. The more men tinkered, the less bugs they fixed. The high-support features helped the women fix bugs, but lowered their self-efficacy.
The men did more repeated tinkering, which harmed their understanding of the task. The women did more exploratory tinkering, which improved their understanding. The explanation they finally discovered was that the women paused to reflect more often than the men, which improved their effectiveness, whereas the men did more counterproductive tinkering.