Needless to introduce the background of the 2019 Hong Kong anti-extradition movement which started in March 2019. It has been a year since its inception. Although people are still talking about the protests, the movement gradually steps out the mass media’s front page. An important question to ask is, has this movement ended?
A simple answer is yes.
One of my research projects is studying this movement using data from Telegram, which has a broadcasting function (“channel”) used by the protesters to disseminate information and coordinate action. By monitoring the daily activities of the entire Telegram channel network since early November, I gathered over 1TB accumulated data about this movement. The earliest available data is from mid-2015, and I stopped data collection in mid-January 2020. The project still needs some time to form a research paper, but an initial analysis can provide some evidence for answering the question aforementioned: Has this movement ended?
Information diffusion and coordination are crucial for a decentralized social movement. In order to disseminate information and coordinate, actors must be efficiently connected, forming a “network of protesters.” The network of Telegram channels mirrors the protesters’ information sharing and coordination. Limitations of using this data exist for sure, but a valid study does not have to rely on perfect data as long as we know the constraints. So what I found so far?
A while ago, I posted the following tweet which was a simple visualization of the network. The network is clearly divided into several camps. But how did this network evolve over time?
The graph visualizes 5.5K @telegram channels in HK protests. Yellow=Protestors (mostly trad. Chinese); Blue=Brokers; Red=Pro-govt (mostly simp. Chinese). What puzzles me is the green actors which are about gambling info (football/horse etc). Any ideas? @Comparativist @OSINTHK pic.twitter.com/dYo4piGp5V
— Ji Ma (@we_love_ji) December 19, 2019
The figure below presents a set of important measures of weekly networks from Jan 2016 to Dec 2019. A node in the network is a Telegram channel, two nodes are connected if either one forwards a message from the other. Edge measures the number of connections in a given network. Density measures the “connectedness” of the network. Modularity evaluates how well the entire network can be divided into distinct groups/communities (1 = perfect division, 0 = hard to distinguish). “Birds of a feather flock together” in the same community. The emotional state measures positive or negative emotions.
Figure 1 uses raw units. Blue vertical reference lines indicate every year’s July 4th. The first red line indicates Oct 4th, 2019, when the HK government passed the Anti-Mask Law. Second red line labels Nov 18th, 2019, the day the Hong Kong Polytechnic University was occupied by the protesters and the peak of the entire movement.
The number of edges and nodes climbed to the peak on Nov 18th, suggesting there were more channels and forwarding activities than ever. But they sharply decreased after Nov 18th. What also decreasing was the number of community—more channels and camps were leaving the arena. The Telegram network seems to be a happy community in general, but the standardized scores give us a different story.
Figure 2 shows the standardized values (i.e., z-score). The numbers of edge, node, and density went back to mean, modularity significantly decreased, and there were fewer network communities in the network. What really interesting is the emotion: negative emotion remained higher than the mean, and this condition could date back to as early as mid-2017.
The emotional status probably gives us another answer: the anger never goes away, it is looking for another opportunity.