Research

People come to social media to satisfy a variety of needs, such as being informed, entertained and inspired, or connected to their friends and community. Hence, to design a ranking function that gives useful and personalized post recommendations, it would be helpful to be able to predict the affective response a user may have to a post (e.g., entertained, informed, angered). This paper describes the challenges and solutions we developed to apply Affective Computing to social media recommendation systems.

We address several types of challenges. First, we devise a taxonomy of affects that was small (for practical purposes) yet covers the important nuances needed for the application. Second, to collect training data for our models, we balance between signals that are already available to us (namely, different types of user engagement) and data we collected through a carefully crafted human annotation effort on 800k posts. We demonstrate that affective response information learned from this dataset improves a module in the recommendation system by more than 8%. Online experimentation also demonstrates statistically significant decreases in surfaced violating content and increases in surfaced content that users find valuable.