This paper investigates the evolution of user interactions in the Facebook social network, focusing on the activity network, which is based on actual user interactions rather than mere friendship. The authors collect data on friendship links and wall posts from a large subset of the Facebook New Orleans network, analyzing over 60,000 users and 800,000 interactions over two years. Key findings include: 1. **Microscopic Level Analysis**: User pairs with infrequent interactions are often triggered by site mechanisms, such as birthday reminders. Even frequent interactors show a general decline in activity over time, with only 30% of pairs consistently interacting from one month to the next. 2. **Macroscale Level Analysis**: Despite the rapid change in individual user pairs, many graph-theoretic properties of the activity network remain stable over time, such as average node degree, clustering coefficient, and average path length. 3. **Discussion**: The study highlights the impact of site mechanisms on user interactions and suggests that examining the content and cause of interactions can provide insights into link strength. It also discusses the implications for systems that leverage social network properties, such as SybilGuard and SybilLimit, which aim to prevent Sybil attacks and unwanted communication. The paper concludes that while the activity network exhibits high churn in individual links, its global structural properties remain relatively stable, providing a nuanced understanding of user interactions in online social networks.This paper investigates the evolution of user interactions in the Facebook social network, focusing on the activity network, which is based on actual user interactions rather than mere friendship. The authors collect data on friendship links and wall posts from a large subset of the Facebook New Orleans network, analyzing over 60,000 users and 800,000 interactions over two years. Key findings include: 1. **Microscopic Level Analysis**: User pairs with infrequent interactions are often triggered by site mechanisms, such as birthday reminders. Even frequent interactors show a general decline in activity over time, with only 30% of pairs consistently interacting from one month to the next. 2. **Macroscale Level Analysis**: Despite the rapid change in individual user pairs, many graph-theoretic properties of the activity network remain stable over time, such as average node degree, clustering coefficient, and average path length. 3. **Discussion**: The study highlights the impact of site mechanisms on user interactions and suggests that examining the content and cause of interactions can provide insights into link strength. It also discusses the implications for systems that leverage social network properties, such as SybilGuard and SybilLimit, which aim to prevent Sybil attacks and unwanted communication. The paper concludes that while the activity network exhibits high churn in individual links, its global structural properties remain relatively stable, providing a nuanced understanding of user interactions in online social networks.