Seems to depend heavily on the hypothesis that a carefully crafted message would sway an uncommitted voter away from Trump.<p>I'm not convinced that works. He has done a fairly good job of showing there's a strong correlation between "any kind of exposure, good or bad", and success.<p>You run the risk of doing the exact opposite of what your goal is.
<i>"The vast majority of people who engaged with our test were willing to add their opinion to the comment thread without feeling any obligation to read the link they were commenting on."</i>