Big Data Suggests Big Trouble for Vaccines
An innovative, graph and network theory study, published in Nature, May 13, 2020, analyzed network patterns on Facebook involving “anti-vaccine,” “pro-vaccine,” and “undecided” pages, their membership, and their outside-page linkage (Johnson et al. 2020). Johnson was also lead author on a different study, using what seems to be more rigorous and detailed data categories, of the spread of pro-ISIS online hate groups.
The main conclusions of the anti vs. pro vaccine study:
“Although smaller in overall size, anti-vaccination clusters manage to become highly entangled with undecided clusters in the main online network, whereas pro-vaccination clusters are more peripheral. Our theoretical framework reproduces the recent explosive growth in anti-vaccination views, and predicts that these views will dominate in a decade. Insights provided by this framework can inform new policies and approaches to interrupt this shift to negative views. Our results challenge the conventional thinking about undecided individuals in issues of contention surrounding health, shed light on other issues of contention such as climate change, and highlight the key role of network cluster dynamics in multi-species ecologies.”
I’m not entirely convinced by their a priori node categorization strategy (using Facebook pages, page affiliation, and page recommendations of external links). A study like this has to start somewhere to create a data set to analyze, for sure, yet I wonder to what degree page membership and page node links are a true reflection of the three a priori belief categories they ascribe: “anti-vaccine,” “pro-vaccine,” and “undecided.” Their strongest inferences are in regard to node entanglement with the supposedly “undecided” camp.
I definitely see the value in their breakthrough insight, which is that anti-vaccination nodes, although representing a smaller membership per node and overall, are “more entangled” with undecided people than pro-vaccine nodes are. The study does provide a likely explanation for how anti-vaccination views have managed to increase via social media. The suggested remedy to promulgate more science-based, pro-vaccine perspectives via social media, over the long term, is to reduce direct argument with anti-vaccination people and pages, and get pro-vaccine pages to be more “entangled” with “undecided” pages.
An excellent summary and critical review of the Nature article, by Meredtih Wadman, also was published on May 13th, in Science. A couple of the critical perspectives from that article are offered again here:
“Sinan Aral, an econometrician at the Massachusetts Institute of Technology who has mapped the online spread of misinformation, praises the analysis’s large size. But he advises a “skeptical eye.” He says it’s not clear that a green page’s linking to a red page leads to persuasion or that online interactions trigger actual changes in vaccination. He adds that the predicted online dominance of red groups in 10 years “is extrapolating a lot from … limited data.”
The study only looks at how people’s views circulate, not the content of pages, “as if people don’t have reasons for their views and are only being manipulated,” says another critic, Bernice Hausman, a cultural theorist at Pennsylvania State University College of Medicine. She calls the paper’s battlefield rhetoric “troublesome,” arguing that it betrays the very mindset — casting vaccine resisters as the enemy — that turns the vaccine skeptical away.”
These caveats occurred to me also, in addition to my reservations about the a priori data categories. I think large, network level analyses like this would also benefit from targeted, randomly sampled demographic surveys. This more detailed profile data on values, attitudes, and beliefs could be used to enhance understanding of specific clusters of nodes. I think that the perspectives of people who used to be “undecided” but who shifted to being “anti-vaccine” would be especially instructive, possibly even as a way of weighting graph edges or at least zooming in on the issue of persuasion, which, after all, is central to this study’s inferences.
My 13 years of involvement with Facebook has also led to some insights, admittedly anecdotal, about page membership, page participation, and some of the paradoxical and irrational elements involved with page-related behavior. Facebook pages are incredibly loosely affiliated aggregates, for one thing, much of the time. It may well be that a great many members or followers of pages don’t even adhere to the page ideology at all. There’s also the interesting reality of so-called “Leftbook,” for example, where pages are carefully curated by several moderators, and it seems to me that a great many of the members are just there for entertainment and don’t really care all that much about beliefs, ideology, or promoting a specific agenda. Many of the pages I follow are more like corner bars than they are actual promulgators of ideology or beliefs. These pages are analogous to Chick-fil-A, for example, in that it is probably true that the vast majority of people who purchase food there do not politicize their choice, and are not intending to make an “anti-gay, anti-trans” statement by patronizing the business, in spite of the business’s track record of espousing that ideology.
However, and this is not captured in Johnson et al. 2020, certain pages are much more heavily weighted toward passionate participation of followers, and exclusion of differing points of view. A smaller scale study of the anti-vaccine versus pro-vaccine issue on Facebook that used metrics to weigh nodes would perhaps be informative. It may well be that there are something like only a dozen highly influential and very well connected anti-vaccine pages, and the rest have little persuasive effect. It may be that anti-vaccine views develop in a “soft sell” environment via pages that outwardly are much more generally focused on “wellness,” or mothering/parenting, or “alternative medicine.” I think the landscape could be far more heterogeneous than Johnson et al. 2020 acknowledges when they constructed their categories for analysis. It may be that zooming in to a finer focus and developing more detailed demographic profiles of pro-vaccine, anti-vaccine, and undecided Facebook users would reveal a more detailed dynamic of how these attitudes shift. For example, how passionately committed to their beliefs are the anti-vaccine people? How persuadable are the undecided people? Are pro-vaccine views any more well-informed than anti-vaccine views are, and what are the pathways for pro-vaccine Facebook users to drift toward being undecided or anti-vaccine? What sorts of rhetorical engagement might still be useful in bringing anti-vaccine individuals to either the undecided or pro-vaccine categories? Clearly, there is a tremendous potential in informatic studies of this kind, but more detail would come to light if the network analysis were coupled with survey data.
Nevertheless, Johnson et al. is a highly valuable and powerful analysis, revealing the potential of network analyses of social media to develop public health policy, for example, and help moderate the flow of misinformation. The larger takeaway for me is that there sure is some fairly sophisticated big data available via Facebook, even for non-embedded researchers. This methodology shows great promise in analyzing patterns related to many other hot topics on social media, including climate change, SARS-CoV-2 attitudes, the approaching 2020 election, and, basically, anything that is polarizing in the extreme, yet where there may well be a large, perhaps persuadable group of undecided onlookers. Personally, it’s a great reminder to be conscious and strategic on social media, avoiding being trolled, or drawn into the morass of spin, or otherwise manipulated.
I’ll try to remember that we usually can’t convert the converted, and much of our arduous energy is expended trying to do so, to utterly no avail. But our participation in spreading factual, peer reviewed, reputable information from reliable sources can have a huge impact on the “undecided” bystander. We have no way of knowing who is reading our comments or following our links when we participate in Facebook pages. But we very well could be swaying people in one of those flip flop green nodes to the blue side.
Figure caption: a, Snapshot from 15 October 2019 of the connected component in the complex ecology of undecided (green), anti-vaccination (red) and pro-vaccination (blue) views comprising nearly 100 million individuals in clusters (pages) associated with the vaccine topic on Facebook. The colour segregation is an emergent effect (that is, not imposed). Cluster sizes are determined by the number of members of the Facebook page. Black rings show clusters with more than 50% out-link growth. Each link between nodes has the colour of the source node. b, Global spread of Fig. 1a for a small number of clusters. The ‘global ether’ represents clusters that remain global (grey). c, Anti-vaccination clusters have a stronger growth in cluster size. Each coloured dot is a node; data are from February–October 2019. d, Anti-vaccination individuals are an overall numerical minority compared with pro-vaccination individuals; however, anti-vaccination individuals form more separate clusters. (Johnson et al. 2020)
