Twitter provides a space for both social and broadcast communication; a tweet by a celebrity can reach millions of users, some of whom may reply and engage in conversation with the celebrity or any other user. This bidirectional communication channel means that influencers can both speak and listen to their audiences on issues they care about. But how will an audience respond to a particular message by an influencer? In this work we are combining deep learning models of text with social network analytics to predict audience response. We are focused on influencers using their platforms for social good, with the goal of developing models of influencers, language and audiences to support positive social impact.