CNetS researchers reach finals of the Prosocial Ranking Challenge

We are developing a “bridging” algorithm to rerank social media feeds, aiming to foster civil discussions and appeal to a diverse audience, thus creating a more positive and inclusive online environment. We accomplish this by promoting content that elicits diverse and positive responses while demoting toxic content and content eliciting toxic responses. Our algorithm achieves this by evaluating posts based on two factors: the emotional response they evoke and the diversity of engagement they attract. The emotional response includes the polarity of the emotions and the presence of harmful reactions while the audience diversity is estimated by using the range of ideological perspectives among engaged users. We believe that our comprehensive evaluation of posts will help mitigate the negative effects of recommender algorithms, particularly in terms of affective polarization.