During the measles outbreak of 2019 Anti-Vaccine Facebook groups increased their membership more than Pro-Vaccine groups, with some growing by over 300%. Contrastingly there were no Pro-Vaccine groups which grew by more than 100% (and most groups grew by less than 50%). If Anti-Vaccine groups continue to grow at a higher rate than Pro-Vaccine groups, then their membership will outnumber that of Pro-Vaccine groups by 2030.
This was one of the findings of a recent study into “The online competition between pro- and anti-vaccination views” which looked at how members transitioned between different Anti-Vaccine, Pro-Vaccine, and Undecided(1) Facebook groups during the 2019 measles outbreak. It identified several factors which helped Anti-Vaccine groups recruit Undecided candidates, but was pretty hard to read (there was a lot of scientific language, and at least one paragraph made up almost entirely by equations). I wanted to see if I could make that information a little easier to digest:
1. People who are Undecided about Vaccines want to make a decision
People who could be impacted by Vaccines (i.e. everyone, THANKS COVID-19) are put in a difficult position. The consequences of not being vaccinated are dire; the unvaccinated person is susceptible to illness, and can help fuel infection outbreaks. Anti-Vaccine groups propose a wide range of similarly negative consequences if one does get vaccinated.
When told the stakes of your decision are so high, it’s natural to feel pressure to make your mind up. Extra factors contribute to the urgency of this decision; some governments require vaccination at a certain age (imposing a time limit), and some preventable illnesses are pretty undesirable. Furthermore, inaction is not an option (or at least, it’s an option that aligns oneself with the Anti-Vaccine ideology).
Critically, Undecided Facebook users were found to actively want make up their minds about Vaccines; of all viewpoints studied the Undecided had the most out-group movement (i.e. changing from one viewpoint to another). They are not a passive audience who must be sought out and convinced by Pro- or Anti- zealots; they are looking to join one of the two camps.
2. There are more Anti-Vaccine groups than Pro-Vaccine Groups, which enables more interaction with the Undecided
At the time of writing there are more Pro-Vaccine than Anti-Vaccine users, but far more Anti-Vaccine than Pro-Vaccine groups(2). While it may seem that it would be beneficial to have more users than groups, this is not the case.
As an Undecided user browses their Facebook feed, they are presented with a mixture of sponsored adverts, posts made by other users they have connected with (their “friends”) and posts added to groups they have joined. It is much easier for them to join a group (and thus add its posts to their feed) than to connect with another user; it’s unconventional to become friends with a someone you hadn’t met in the real world, and most groups can be joined at a click of a button (rather than waiting for a connection request to be vetted and approved by another user).
Furthermore, posts added to groups tend to relate to the topic of the group, while users tend to post about a variety of topics related to their life. For example, the “Shit London” group I’m part of rarely strays from cataloguing the capital’s oddities (a rule enforced by the page’s admins):
Basically, it is easier to connect to a group than a user, and participation in a group is more likely to expose you posts related to its content than it would if you were only connected to a user of the same group. This means that the majority held by Anti-Vaccine groups is more advantageous than the majority held by Pro-Vaccine users, when looking at converting the Undecided to their cause.
3. Anti-Vaccine groups are better connected with Undecided groups than Pro-Vaccine groups are
If we were to release 100 people into one large room and left them to their own devices, it’s unlikely they’d all end up standing perfectly equidistant from each other; they’d group up and chat with with those who have similar interests, or with people that they know in real life – thus creating separate clusters of individuals. Let’s imagine we left our participants to their own devices for an hour or two; when we return two separate clusters have formed – one of 30 people discussing beekeeping, and one of 70 discussing Genghis Khan.
SURPRISE! Our population also has opinions about Vaccines (did you see that coming?). Using the percentage of Pro-, Anti- and Undecided groups found in the study, we can predict that we’d have 24 Anti-Vaccine, 9 Pro-Vaccine, and 67 people who hadn’t made up their mind about vaccines (but really wanted to). If our 30 person beekeeper cluster contained the 9 Pro-Vaccine proponents and 21 of the Undecided, it’s likely that (once conversation moved on to Vaccines) many of the Undecided would be converted to Pro-Vaccine, and the original 9 would feel pretty good about their efforts. Meanwhile (unbeknownst to them) the 24 Anti-Vaccine enthusiasts are making great headway with the 46 Undecided in the cluster of Khan-ers. Critically, the makeup of these naturally occurring clusters impacted the potential for conversion of the Undecided in the whole 100 person population.
It could be the case that one or two Pro-Vaccine beekeepers who, having said pretty much all there is to say about apiculture, wander over to talk with some of the Khan-ers they recognise from their morning Yoga class. They still wouldn’t be very well positioned within the cluster, and would have to awkwardly hang around at the edge of the group before being able to make any headway with the newly accessible Undecided.
Although Facebook groups don’t occupy any physical space, the same principles apply. People join groups dedicated to their interests (e.g. alternative medicines, or health research), or to people or places relevant to them in real life (e.g. sports teams or schools). When two Facebook groups share many members, it’s likely that ideas will spread from one between the other as its users interact, share posts and add comments. We can picture two well connected Facebook groups as two people with a lot in common standing pretty close to each other and willing to have a chat.
When I said “we can picture this” I meant it literally; the following image visualises how each of the different Pro-Vaccine, Anti-Vaccine and Undecided groups are distributed within Facebook’s ecosystem. Each group is shown as a circle (each green circle represents an Undecided group of Facebook users, each red represents an Anti-Vaccine group, and each blue represents a Pro-Vaccine group). The circles are near each other if there’s lots of connections between them; a red circle near a green circle would represent a pair of Anti-Vaccine and Undecided groups which had many shared users. The lines basically represent connections between the groups (e.g. shared users, or users transitioning between groups)(3).
You can see three main clusters; one small largely Undecided cluster on the left, one medium cluster with many Pro-Vaccine and Undecided groups, and one much larger cluster made up mostly of Anti-Vaccine and Undecided users (and a few Pro-Vaccine groups nodding and smiling at the edge of the cluster).
This means the “people in a room” scenario wasn’t far off what we have in the real world; the largest cluster contains many Undecided and Anti-Vaccine groups tangled up with each other – and the Undecided in this cluster showed the highest group out-growth (i.e. moving from one viewpoint to another) of all groups examined. Meanwhile Pro-Vaccine groups have the misguided view that they are winning over the hearts and minds of Undecided (largely unaware of the uncontested crowd in the other cluster). The makeup of the naturally occurring clusters has already benefited the Anti-Vaccine Facebook groups in recruiting the Undecided.
4. Anti-Vaccine groups offer a variety of potentially attractive narratives for Undecided groups
Pro-Vaccine Facebook groups typically present one theme; vaccines are good. On the other hand, there are lots of different Anti-Vaccine narratives which can entice Undecided users (e.g. conspiracy theories, alternative medicines or safety concerns).
For example if an Undecided person is not very interested in health concerns, then groups which focus on conspiracy theories can provide them an alternate route to making a decision. This is not something that Pro-Vaccine groups can provide; the idea that the widely accepted narrative regarding vaccines is actually true doesn’t make for a great conspiracy theory – although similar strategies have been tried on those uncertain about the moon landing:
So far we’ve covered “4 reasons why Anti-Vaccine Facebook groups are winning over the Undecided masses” – but there are two more small bonus reasons if you have the appetite to keep reading. Based on the average attention span in 2020 I’m humbled and impressed that you’ve made it this far, but the most complicated part is behind us. You’ve got this;
Geographical associations help Anti-Vaccine groups cluster with Undecided
Some Facebook groups are only relevant to you if you live in a particular location (for example I suspect that the average member of “Shit London” lives or has lived in London – although of course it’s possible to celebrate London’s character regardless of geographical background). As users join more groups dedicated to a particular location, highly interconnected “clusters” of groups can form.
Anti-Vaccine groups dedicated to a particular location are not uncommon; the Vaccine Safety Council of Minnesota, Minnesota Natural Health Coalition, National Health Freedom Action and Minnesota Vaccine Freedom Coalition are all examples of Anti-Vaccine groups based in – you guessed it – Minnesota. The fact that Geography contributes to group interconnectivity helps create clusters where Anti-Vaccine users are well connected with the Undecided.
Medium sized Anti-Vaccine groups showed the largest growth during the 2019 Measles outbreak
Typical modelling of group dynamics predicts that the larger the group the more users it will attract – however it was found that medium sized Anti-Vaccine groups saw the most growth. While large Anti-Vaccine groups tend to take up the most attention of Pro-Vaccine users, these smaller groups can expand largely unchallenged.
It was found that Anti-Vaccine Facebook groups increased their membership more than Pro-Vaccine groups during the measles outbreak of 2019. On examination the following factors were identified as contributing to that outcome:
- People who are Undecided about Vaccines want to make a decision
- i.e. Undecided users don’t need to be found and approached, they want to decide
- There are more Anti-Vaccine groups than Pro-Vaccine Groups
- Anti-Vaccine groups are better connected with Undecided groups than Pro-Vaccine groups are
- i.e. The number of and Anti-Vaccine groups and the way they are connected with Undecided groups means they are better positioned to recruited Undecided than Pro-Vaccine groups are
- Anti-Vaccine groups offer a variety of potentially attractive narratives for Undecided groups
- i.e. If someone isn’t interested in the health benefits of Vaccines, there’s not much left Pro-Vaccine users can use to entice them, whereas Anti-Vaccine users have other topics they can draw on
Factoring in these elements (and a couple of others) it was predicted that Anti-Vaccine users would outnumber Pro-Vaccine users by 2030.
Conclusion (/So what?)
Traditional forms of modelling group dynamics which don’t consider these key factors won’t be effective for informing policy or strategy to counter the spread of the Anti-Vaccine ideology. Bad things happen when communities reject vaccinations, so we should probably do something about this.
WHO’s guidance on “How to respond to vocal vaccine deniers in public” highlights the importance of making the public audience “more resilient against anti-vaccine statements and stories” and to “support the vaccine hesitants in their vaccine acceptance decision”. With that in mind, one example of how Facebook could use cluster data would be to predict which Undecided groups are at the highest risk of Anti-Vaccine conversion, targeting them with specific interventions designed to increase resilience to Anti-Vaccine narratives.
There are many different ways in which Facebook could work with the data provided to mitigate Anti-Vaccine groups’ advantages, but how can the average member of society (who doesn’t have much sway on Facebook’s development) use this information? What I’ve taken from this is that there is a higher probability than I realised that Undecided people who I know will become Pro- or Anti-Vaccine at some point in the near future. I think it’s a good idea to be Pro-Vaccine, so I’m going to try to be properly prepared to inoculate those close to me to the tactics, statements and stories used by Anti-Vaccine proponents.
If you want to do this too, the “How to respond to vocal vaccine deniers in public” guide I mentioned earlier isn’t a bad place to start. If you want to learn more about real-world Anti-Vaccine disinformation, check out this pre-filtered list of articles related to Vaccines in the Disinformation Database.
(1) In this case “Undecided” groups refers to Facebook pages which could become associated with the vaccine debate, but with no overt stance – for example a School-Parent Association page, which could be brought into discussion about Vaccines due to mandatory vaccination of children.
(2) The following table shows the number of users and groups associated with each viewpoint:
|Group Type||Total Users||Amount of Facebook Groups|
|Anti-Vaccine||4.2 million users||317 groups (24% of all groups studied)|
|Pro-Vaccine||6.9 million users||124 groups (9% of all groups studied)|
|Undecided||74.1 million users||885 groups (67% of all groups studied)|
(3) This analysis was done using Gephi software with the ForceAtlas2 layout. My understanding is that initial random positioning of the circles (Facebook groups) was transformed into this image by setting the circles to push away from each other, while having the connections between them act as springs; meaning that highly connected groups were pulled back together in spite of the “push”.
It’s not very clear how long you leave the circles pushing and pulling against each other before you consider the visualisation to be “done”, but I think that it doesn’t matter too much (as long as enough time has elapsed for circles to properly break free from each other). If you are more familiar with the software than I am, please reach out and let me know!