Diffusion on Social Media Platforms: A Point Process Model for Interaction among Similar Content

Abstract

Social media platforms disseminate a massive volume of user-generated content, some of which convey similar and overlapping information. We study how the diffusion of a given piece of content (called a cascade) is influenced by the diffusion of other cascades carrying similar content (called parallel cascades). We theorize that the diffusion of a cascade can be inhibited or amplified by that of parallel cascades containing similar content. To study this phenomenon, we formulate a generalized version of the self-exciting point process model and showcase a novel approach to evaluating the parallel diffusion of similar social media content. We estimate the model using Twitter data. We observe that, on average, the diffusion of a cascade is inhibited by the concurrent diffusion of parallel cascades with similar content. We further identify an asymmetry among content producers as the diffusion of content contributed by those with larger networks is more likely to be amplified by the diffusion of similar content. Our study underscores the importance of accounting for content similarity as failing to do so may overestimate assessments of a cascade’s diffusion. Our results also suggest that smaller, individual social media content contributors should avoid publishing repetitive content and channel their efforts towards developing novel content, while this is not a concern for larger content contributors.

Key words and phrases:

• cascade
• content diffusion
• competition for attention
• parallel cascade
• point process
• social media
• media platforms
• social media content

Acknowledgements

We would like to thank the guest editors of this Special Section, Rob Kauffman and Thomas Weber, for their invaluable feedback and guidance through the review process. We would also like to acknowledge and thank three anonymous reviewers for their insightful comments. We are grateful to seminar participants at Boston College (2019), Boston University (2019), Hong Kong University of Science and Technology (2019), Pennsylvania State University (2019), Shanghai Jiaotong University (2019), Singapore Management University (2019), University of Texas at Dallas (2019), and Tel Aviv University (2019). We also thank the conference participants at CIST 2018 and the SITES minitrack at HICSS 2019, as well as three anonymous reviewers for HICSS 2019. This work was supported by the Center for Services Leadership, the Office of Knowledge Enterprise Development at Arizona State University, and the AWS Cloud Credits for Research Program.

Notes

1. In reality, multiple producers often publish similar content. In such cases, a piece of content cannot be viewed in isolation as its information overlaps with that presented in other pieces of content.

2. This is an important objective: if diffusion of a piece of content is influenced by the diffusion of similar content, then prior evaluations of content diffusion may be incomplete. Our study recognizes that social media content does not exist in a vacuum, a perspective that better mirrors the reality of social media platforms as an online space teeming with user-generated content.

3. We focus on this compelling context for the following reasons. During these events, information is perishable due to extreme environmental uncertainty and volatility where the events unfold [45]. Therefore, it is critical to understand the factors that facilitate rapid diffusion of information. Second, due to their sudden nature, the amount of relevant social media content surges and competition for users’ attention to event-related content becomes particularly acute.

4. Larger producers are more immune to negative impacts on the diffusion of their content from that of similar content and are more likely to benefit from redundant content diffusion. The advantage of being a large, well-connected producer has been documented [67]. We offer nuanced insights by showing that content diffusion of larger producers may be superior due to these producers’ credibility and reputation and their ability to profit from the propagation of similar content. For smaller producers, however, our results indicate that the diffusion of their content is more susceptible to inhibition by the spread of similar content.

5. Our point process model is not precisely a causal model, but it examines content sharing across time and assesses how a user sharing a piece of social media content impacts other users’ decisions to also share in the future. Thus, our study infers causality in the sense of Granger causality and deepens our understanding of social media content diffusion over time.

6. For cable news shows, greater amounts of social media activity about one show leads to a reduction in viewership for competing shows [61]. Therefore, social media can serve to reinforce and intensify competition for attention among organizations.

7. For example, the self-exciting point process is conducive to analyzing activity by developers on open source projects. This is because developer activity tends to be correlated such that a change to a project’s source code can trigger additional changes [62]. The self-exciting point process is also appropriate to model online auction bidding on sites like eBay since a user’s submission of a bid often motivates other users to submit bids as well [11].

8. We would like to thank Rob Kauffman, co-editor of this Special Section, for the recommendation to refer more appropriately to our arguments as conjectures. Because the theoretical underpinnings are still developing, we argue for both directions of the effect of the diffusion of parallel cascades on the diffusion of a cascade. Therefore, it is premature to label our arguments as hypotheses and is more proper to refer to them as conjectures.

9. Our model can be applied to diffusion of content on other social media platforms that promote content sharing as a feature too.

10. We specified the date ranges that we were interested in as starting from the time the disaster materialized to approximately the end of the initial response period. Humanitarian organizations normally view the first 72 hours after an emergency as the most critical response period. Hospitals are suggested to be prepared for up to 96 hours as the initial response period following a disaster [66]. Therefore, we estimated the end of the initial response period as four days after the date EM-DAT recorded as the end of the disaster. The only exception to this rule was the Black Forest fire event. Because this event was relatively more protracted than the other three events, we ended data collection on the date that the fire was publicly reported to be 100 percent contained.

11. We submitted to WeLink a set of queries specific to each disaster event. These queries contained keywords and phrases that were commonly present in hashtags and content associated with the emergencies as well as combinations of the location of the disaster and event name [52].

Additional information

Notes on contributors

Eunae Yoo

Eunae Yoo ([email protected]; corresponding author) is an Assistant Professor of Supply Chain Management at the Haslam College of Business at the University of Tennessee. She received her Ph.D. from the W. P. Carey School of Business at Arizona State University. Dr. Yoo’s research interests lie at the intersection of information systems and supply chain management, and include social media, online information diffusion, and humanitarian operations management. Her work has previously been published in the Journal of Operations Management.

Bin Gu

Bin Gu ([email protected]) is the Earl and Gladys Davis Distinguished Professor and Associate Dean of China Programs at the W. P. Carey School of Business at Arizona State University. He received his Ph.D. degree from the Wharton School of Business at the University of Pennsylvania. Dr. Gu’s research interests are in online digital platforms, future of work, AI and fintech, digital healthcare, online social media, and social networks, and IT-enabled business models. His work has appeared in Management Science, MIS Quarterly, Information Systems Research, Journal of Management Information Systems, Production and Operations Management, Journal of Retailing, and other leading academic journals.

Elliot Rabinovich

Elliot Rabinovich ([email protected]) is the AVNET Professor of Supply Chain Management at the W. P. Carey School of Business, Arizona State University and the Co-Director of the Internet-edge Supply Chain Management Lab. His Ph.D. is from the University of Maryland. Dr. Rabinovich’s research has focused on the effects that Internet technology applications have on supply chain and operations management. His work has been published in California Management Review, Decision Sciences, Journal of Business Logistics, Journal of Operations Management, Sloan Management Review, and other journals. As part of his research, he has worked with such companies as Cooking.com, eBags.com, Intel, PetSmart, Walmart, and Twitter. His research has been recognized by the University of Maryland with the Nash Outstanding Doctoral Alumni Award, with fellowships from the Institute of Supply Management, and with awards from the Council of Supply Chain Management Professionals. He has recently published a book, Internet Retail Operations: Integrating Theory and Practice for Managers (Taylor & Francis).