Tuan Quang Phan

Assessing the Impact of Granular Privacy Controls on Content Sharing and Disclosure on Facebook

We examine the role of granular privacy controls on dynamic content-sharing activities and disclosure patterns of Facebook users based on the exogenous policy change in December 2009. Using a unique panel data set, we first conduct regression discontinuity analyses to verify a discontinuous jump in context generation activities and disclosure patterns around the time of the policy change. We next estimate unobserved effects models to assess the short-run and long-run effects of the change. Results show that Facebook users, on average, increase use of wall posts and decrease use of private messages after the introduction of granular privacy controls. Also, users’ disclosure patterns change to reflect the increased openness in content sharing. These effects are realized immediately and over time. More importantly, we show that user-specific factors play crucial roles in shaping users’ varying reactions to the policy change. While more privacy sensitive users (those who do not reveal their gender and/or thos...

Uncovering the Importance of Relationship Characteristics in Social Networks: Implications for Seeding Strategies

Seeding influential social network members is crucial for the success of a viral marketing campaign and product diffusion. In line with the assumption that connections between customers in social networks are binary (either present or absent), previous research has generally recommended seeding network members who are well-connected. However, the importance of connections between customers varies substantially depending on the relationships characteristics, such as its type (i.e., friend, colleague, or acquaintance), duration, and interaction intensity. This research introduces a new Bayesian methodology to identify influential network members and takes into account the relative influence of different relationship characteristics on product diffusion. Two applications of the proposed methodology—the launch of a microfinance program across 43 Indian villages and information propagation in a large online social network—demonstrate the importance of weighting connections in social networks. Compared with traditional seeding strategies, the proposed methodology recommends substantially different sets of seeds that increased the reach by up to 10% in the first empirical application and up to 92% in the second.

Hashtag Popularity on Twitter: Analyzing Co-occurrence of Multiple Hashtags

Hashtags increase the reachability of a tweet to manifolds and consequently, has the potential to create a wider market for brands. The frequent use of a hashtag features it in the Twitter trending list. In this study we want to understand what contributes to the popularity of a hashtag. Further, hashtags generally come in groups in a tweet. In fact, an investigation on a real world dataset of Great Eastern Japan Earthquake reveals that 50 % of hashtags appear in a tweet with at least another hashtag. How this co-occurrence of hashtags affects its popularity is also not addressed heretofore, which is the focus herein. Results indicate that if a hashtag appears with one or more other similar hashtags, popularity of the hashtag increases. In contrast, if a hashtag appears with dissimilar hashtags, popularity of the focal hashtag decreases. The results reverse when dissimilar hashtags come along with a URL.

Incremental Algorithms for Sampling Dynamic Graphs

Among the many reasons that justify the need for efficient and effective graph sampling algorithms is the ability to replace a graph too large to be processed by a tractable yet representative subgraph. For instance, some approximation algorithms start by looking for a solution on a sample subgraph and then extrapolate it. The sample graph should be of manageable size. The sample graph should preserve properties of interest. There exist several efficient and effective algorithms for the sampling of graphs. However, the graphs encountered in modern applications are dynamic: edges and vertices are added or removed. Existing graph sampling algorithms are not incremental. They were designed for static graphs. If the original graph changes, the sample must be entirely recomputed. Is it possible to design an algorithm that reuses whole or part of the already computed sample? We present two incremental graph sampling algorithms preserving selected properties. The rationale of the algorithms is to replace a fraction of vertices in the former sample with newly updated vertices. We analytically and empirically evaluate the performance of the proposed algorithms. We compare the performance of the proposed algorithms with that of baseline algorithms. The experimental results on both synthetic and real graphs show that our proposed algorithms realize a compromise between effectiveness and efficiency, and, therefore provide practical solutions to the problem of incrementally sampling the large dynamic graphs.

Investigating the Effects of Self-Presentation at Online Social Network Sites and Brand Pages on Offline Purchase Behavior

The emergence and rapid growth of social media platforms, and particularly social-media based brand communities have spurred significant popular interest in recent times. However, despite the growing economic importance of brand presence on social media, little is understood about whether and how user engagement with these brand communities benefits product sales in brick-and-mortar stores - the online-to-offline (O2O) conversion problem. In this study, we combine two large real-world datasets from a popular social network site (SNS) and the loyalty card dataset from a brick-and-mortar Asian fashion retailer to study the offline purchase behavior of the SNS brand page members. We present evidence that individuals upon joining the brand page reduce their offline purchase expenditure on average. However, using a combination of text mining and statistical approaches, we show that this reduction is significantly attenuated for some individuals who self-present more than others on the SNS. The findings from our study not only illuminate our understanding of the offline economic impacts of online self-presentation, but also present newer ways of performing behavioral targeting of online users.