Citation

Miller, I. D. & Cupchik, G. C. (2014). Meme creation and sharing processes: individuals shaping the masses. In Proceedings of the 2nd Collective Intelligence Conference. Cambridge, MA, USA. arXiv:1406.7579

Links

Preview

paper: Meme creation and sharing processes

Abstract

The propagation of online memes is initially influenced by meme creators and secondarily by meme consumers, whose individual sharing decisions accumulate to determine total meme propagation. We characterize this as a sender/receiver sequence in which the first sender is also the creator. This sequence consists of two distinct processes, the creation process and the sharing process. We investigated these processes separately to determine their individual influence on sharing outcomes. Our study observed participants creating memes in the lab. We then tracked the sharing of those memes, derived a model of sharing behavior, and implemented our sharing model in a contagion simulation.

Although we assume meme consumers typically have little or no information about a meme’s creator when making a decision about whether to share a meme (and vice versa), we nevertheless ask whether consumer re-sharing behavior can be predicted based on features of the creator. Using human participants, web log monitoring, and statistical model fitting, the resulting Creator Model of Re-sharing Behavior predicts 11.5\% of the variance in the behavior of consumers. Even when we know nothing about re-sharers of a meme, we can predict something about their behavior by observing the creation process.

To investigate the individual re-sharing decisions that, together, constitute a meme’s total consumer response, we built a statistical model from human observation. Receivers make their decision to share as a function of the meme’s content and their reaction to it, which we model as a consumer’s decision to share. The resulting Consumer Model of Sharing Decisions describes 37.5\% of the variance in this decision making process.

Finally, we implemented our consumer model of sharing as the infection function in an SIS contagion simulation. When using model parameters similar to our human participants research, we received encouraging results that suggest future refinements will be worthwhile.