This is the third post in a series of excerpts from my graduate research at Cornell University; each has been adapted for the purposes of this format. To read the full report, in all its technical glory, please visit my website.
Previous Topic: Principles of Social Networking Theory
Have you ever wondered why that cat meme went viral instead of your agency’s “critical” policy initiative? You aren’t alone. In fact one of the single fastest growing research directives in social networking is predicting what, how, and why something goes viral. Commonly known as cascade modeling, it more-or-less describes the action a post can take as it increases it’s velocity through a network increases. Still unsure? No problem, before we get into the technicalities of a cascade we’ll bring some atoms to the bits.
…harnessing the cascade is akin to becoming God…
Paint a Picture
Imagine a deep ocean, at the bottom of this massive body of water a minor earthquake occurs opening a small fissure in the Earth’s crust. The ocean water fills this new void and the resulting disruption is displayed as a 1cm displacement of water at the surface (not meant to be a precise number). Over time this small disruption ripples out from the center, increasing in size throughout the process; however, at the surface the marginal increase in size isn’t so apparent, its the accumulation of water displacement beneath the surface that is cause for alarm. As shore is approached the displaced water beneath the surface is forced upwards by the increasingly shallow depth. We have described an impossible force of nature — a tsunami — a feat of our planet so strong and overwhelming, but almost undetectable at the origin. Translate this phenomena to a social network and we have our cascade — its fast, strong, and only detectable once we begin to see the displacement taking place.
Let’s get technical for a bit
Consider a social network pressured by the principles described in my last post. Constructural theory suggests that similar nodes (users) within the network will exhibit higher amounts of interaction once self-disclosure of commonalities is provided (i.e. people make it aware they exist and they have stuff in common 👍🏻). To conceptualize how this information is consumed by the network we must consider how this disclosure moves about the information pathways of the network, whether it goes beyond the origin-node’s cluster group, and what pathways provide the shortest distance for the information to reach maximum relevancy.
A final, yet crucial, feature-set of the information traveling the network is the dissemination frequency between nodes: that is, whether it builds enough velocity to produce an information cascade within the network, and — if so — how that cascade might be predicted and leveraged to increase efficiency of information distribution, particularly in terms of influencing non-connected nodes to solicit collaborative action on a network task without social-tie formation. To construct strategies to streamline information delivery amongst the network we must examine the edges linking the network together.
Commonly referred to as information pathways, edges represent information highways between individual nodes and clusters of nodes. Each carries a dynamic range, the shortest path between two nodes if all intermediary relationships are removed, and embeddedness, network density of nodes’ common neighbors. Modeling these characteristics allows us to identify the edges on which packets of information have the potential to travel fastest, which — together — comprise the network’s backbone. By delivering an information packet to any node along the backbone we can expect to minimize our time of delivery, thus achieving substantial low-latency social network distance.
With the shortest path to information ubiquity amongst the network determined, the question becomes which node should be targeted for initial deployment of the packet in order to increase its delivery velocity throughout the network. In this case, we can either focus on nodal relationships with strong ties, load-concentrating, or leverage weak ties within the network which may have lower initial network density, load-leveling. Research shows that by targeting nodes of high acquaintance, load-concentrating, the network will automatically rebalance dissemination latency, increasing velocity throughout the network. We can further degrade latency by deploying information packets to high-density nodes of relative status. Status theory shows the positive social tie of the origin node to its neighbors will elicit packet acceleration equal to tie’s strength.
Traversing a social network is no small feat, but we have managed to construct a conceptual model of doing so. Within each network there exists an immensely powerful force that can flow throughout a network with lossless velocity and holds the ability to warp information between networks. To many studying social networks, harnessing the cascade is akin to becoming God — you wield control over your direct network, and grasp the potential to conduct informational imperialism on entirely separate networks. Today we possess the ability to identify a cascade in formation and predict its size, shape, and temporal curve; however, we haven’t yet been able to identify what information packets possess the capability to originate such an event.
4 Factors of Cascade Growth
Information packet features, origin-node features, structural dynamics of the cascade itself, and temporal features within the affected network.
Holistically, the features of the origin-node and the network’s temporality represent the two most influential components in determining the scope of the resulting cascade. If the origin-node rests within a high-embedded cluster, holds relative status to its neighbors, and has constructed multiple brokerage relationships with non-cluster nodes, then we can forecast the initial velocity of the cascade to exhibit lossless acceleration throughout the backbone. Similar to the depth of the ocean for a tsunami, the deeper the origin density the greater the initial speed of the packet.
As the packet travels between nodal clusters, evidence suggests the temporal nature of the network’s information recognition further degrades clearance of the packet’s features (what the information says); instead, the network simply verifies the packet, disseminating it to the next node at increasing speeds. This speed is critical as the cascade crosses the uncanny valley where engagement stalls as nodes indifferent to the content are responsible for viewing and signing the packet. If the cascade is able to traverse the valley and grow large enough in size, its speed will accelerate rapidly as its content becomes universally signed and shared. If the cascade reaches peak velocity at the shoreline it has the potential to spill over the network’s boundary into a neighboring network (i.e. going viral on Twitter makes it go viral on Facebook).
End of Excerpt
Next Topic: Social Networks, Who Cares?
Wow, we know so much about social networks now. They form, they die, and they can transmit information faster amongst a populace than almost any other mechanism known to man. In the next few posts we’ll begin to form a foundation for which urbanists can begin to construct strategies of digital civic management and communication.
I welcome your feedback; keep in mind this is only a part of a series in which we’ll fully vet the concepts proposed here. Opinions are my own.
Jon Kleinberg et al. 2011. “Maintaining Ties on Social Media Sites: The Competing Effects of Balance, Exchange, and Betweenness.” Association for the Advancement of Artificial Intelligence.
Naaman, More, Ross McLachlan, and Emily Sun. 2017. “MoveMeant: Anonymously Building Community Through Shared Location Histories.” ACM CHI Conference on Human Factors in Computing Systems.
Jon Kleinberg et al. 2008. “The Structure of Information Pathways in a Social Communication Network.” Proceedings of the 14th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining.
Leskovec, Jure, Daniel Huttenlocher, and Jon Kleinberg. 2010. “Signed Networks in Social Media.” Proceedings of the SIGCHI Conference on Human Factors in Computing Systems.
Justin Cheng et al. 2014. “Can Cascades Be Predicted?” Proceedings of the 23rd International Conference on World Wide Web.
Rahmtin Rotabi et al. 2017. “Cascades: A View from Audience.” Proceedings of the 26th International Conference on World Wide Web.