Best viewed in full screen.
Twitter’s popularity as a forum for discussing and sharing media events as they unfold offers a valuable source of data for measuring engagement. But much of the marketing, advertising, and social science research in the field is bound by siloed methodologies.
So what does audience engagement mean in the age of social media? HI collaborated with neuroscientists at Columbia University and City College of New York to learn more. Using the pilot episode of AMC’s drama The Walking Dead as our stimulus, we set out to find the drivers of neural and social response.
To do this, we worked with online analytics company Crimson Hexagon and Twitter to obtain and query all the Twitter activity referencing the show during its original airing in 2010—what turned out to be more than 19,000 tweets over 90 minutes. We parsed and labeled this content according to a customized coding scheme that took into account indications of sentiment, humor, and “immersion,” or engagement with the subject matter. We then pegged these tweets to 194 scenes, or narrative units, within the episode.
Using demographic information mined from this data, we recruited 20 subjects to watch the show while hooked to electroencephalogram (EEG) monitors to measure neural activity. Because this was calculated for every second of the episode, we were able to identify specific moments in the show that were associated with high levels of activity for three corresponding brain waves, which have previously been linked to attention, affect, and memory encoding.
We then took an innovative, design-centered approach to analyzing and combining the data from each of these sources. HI’s Graham Technology Fellow Clint Beharry developed custom software to compare brain and social media data with the content of the episode. The program also displays content categories from HI’s coding scheme, all in real time as the show unfolds. For example, we see brain activity spike when a little zombie girl is shot in the head. Following the real-time progression of tweets connected to this scene, we observe that tweets immediately following the event demonstrate high levels of immersion and positive sentiment (excitement, celebration); those occurring after a delay are more often humorous (jokes to relieve tension).
In general, we find that moments that produced high levels of brain activity also produced high volumes of social response, suggesting a link between compelling content and social media engagement. Yet there were clear points of divergence, suggesting that there may types of content that people are engaged with, yet refrain from sharing or discussing online. Below is a video exploring the app and communicating some of our most interesting findings.