Events Calendar

Kristen Altenburger is a staff research scientist at Meta.

Title: Social Networks for Product Innovation

Abstract: Technological innovations have fundamentally transformed information flow and social behavior. As society’s digital platforms continue to promote instantaneous connectivity, network science can help advance machine learning and causal inference methods aimed at driving product innovation. This talk will first provide an overview of network science in machine learning applications. Node attribute prediction tasks arise in a wide range of classification tasks on social networks. Examples include detecting spam accounts, identifying compromised accounts, and inferring user demographics for targeted marketing. Organizations have numerous digital social networks available for graph learning problems with little guidance on how to select the right graph or how to combine multiple edge types. For example, while user-to-user interactions are directed in nature, many graph learning approaches use the undirected version of the network. In this paper, we introduce how to incorporate edge direction, edge weight, and multi-relational data for node prediction tasks. Next the talk will introduce "causal network motifs" for addressing interference in A/B tests on networks. In experimental settings such as social networks, users are interacting and influencing one another, which may violate conventional assumptions of no interference for credible causal inference. Existing solutions account for the fraction or count of treated neighbors in a user's network, but typically ignore the network structure beyond immediate neighbors. Our study provides an approach that accounts for both the local structure in a user's social network via motifs as well as the treatment assignment conditions of neighbors. Finally, the talk will conclude with an overview on future research directions. In sum, these projects highlight the importance of network science in advancing machine learning and causal inference approaches and demonstrate network science's unique role in product innovation.