The initial vote taken of audience members finds that about one-third of them are undecided. More than half support giving the courts decision-making authority in sexual-assault cases, and just 12 percent support assigning such a role to colleges.
The post-debate vote comes in. Fifty-six percent of the in-person audience supports having the courts decide sexual-assault cases, an unchanged statistic. The share of support for having colleges do so, on the other hand, more than doubles, while the undecided vote declines to 14 percent. Nearly half of the people in the crowd have changed their minds.
56% is unchanged from before and after the poll, so the "nearly half of the people" who have changed their mind are in the other 44%. Presumably the decline in undecideds (33-14=19%) corresponds to the rise in "colleges" (more than doubles, from 12% to >24%). I don't understand where the other 5% went, though there is some wiggle room in "about one third." In any case, 25% is not "nearly half," but more exactly one quarter.
This is poor reporting. The numbers are reported in inconsistent ways and the conclusion does not follow at all.