Explaining Viewing and Sharing Dynamics for YouTube Videos

Finally for this session at Web Science 2016 we move to Sebastian Stommel, who begins by considering what we mean by Web science in the first place. He suggests that 'big data' serve as a macroscope: a new way of looking at things at scale, and an opportunity to create generative models to explain digital traces.

The study applies this philosophy to the analysis of YouTube videos, which have a defined posting date and properties such as the number of views (indicating attention) and shares (indicating word of mouth). A generative model to explain such metrics over time could be the theory of collective coping, which shows that thinking about grief continues for longer than talking about it: the difference between viewing and sharing patterns on YouTube may exhibit similar differences.

Another background theory comes from physics, whose phase transitions (e.g. from fluid to solid) can be compared to the transition from stable to panicked financial markets, for instance, but may also correspond to behavioural patterns on YouTube. Such transitions are driven by a combination of endogenous and exogenous factors that may be identified here, too.

Some videos – especially for anticipated events, such as the transit of Venus or Earth Hour – also build slowly and then fall away again quickly; this can be described as an anticipated shock and needs to be further studied in order to better understand the endogenous and exogenous factors at play here.