To understand which users and communities were more prominent, we built a network of quote tweets and retweets out of the almost 50K tweets that were classified as potentially driving hesitancy in phase 1 of the vaccine rollout. Then, we ran clustering and community-detection algorithms on the network, identifying two major clusters.

We can define the two communities as being composed of ‘hard’ and ‘soft’ detractors, with quite significant differences between the two. While hard detractors typically appear to oppose vaccination altogether, the soft detractors do not; rather, they are critical of the way the vaccine rollout has been conducted by the Government. It is likely to be difficult to convince hard detractors to change their minds about their stance on vaccinations, whereas we believe that improvement in the Government’s communication policy could prevent, or at least minimise, the presence of this second community in conversation driving hesitancy.

The clearest example of this dynamic is the controversy that emerged after the Government’s decision to extend the wait period for the second vaccination dose to 12 weeks, against the recommendation of both the World Health Organization and manufacturers. On January 23, critiques of the policy by soft detractors resulted in a major peak in hesitant conversations and kept on driving posts expressing hesitancy over the following days and weeks. Notably, it was doctors speaking against the extension, claiming it could result in reduced efficacy, further infections and deaths. Moreover, users felt that the Government was being unclear on the reasons why it was taking such a decision, as demonstrated by this highly retweeted tweet by broadcaster Piers Morgan.

The difference in attitudes towards vaccines was also reflected in the network structure and the tweeting volumes of the two communities. The hard detractors cluster is slightly larger than the soft detractors one, with the first comprising 10,702 users, and the second 8,828. The former is also somewhat tighter knit, with members interacting with each other very often. Hard detractors tended to post more tweets expressing hesitancy: each member posted on average 2.4 tweets compared to the 1.4 of soft detractors. These discrepancies in both the density of the network and tweeting volume can be attributed to the fact that soft detractors do not appear to be opposed to vaccination as a whole, and hence will post tweets expressing hesitancy only on the occasions in which they disagree with aspects of the rollout. For instance on the 23rd of January, the community posted 11% of its total tweets and generated 13% of its impressions accusing the Government of handling the timing of the second dose of the vaccine in an irresponsible way. Conversely, hard detractors are a tighter community of vaccine sceptics, regularly expressing their concern about, and outrage at, an alleged vaccine conspiracy.

This can also be seen from the way the conversation network unfolded over time as it can be seen in the ‘Dynamic Network’ video. Notice how the red cluster representing soft detractors grows by big leaps, whereas the growth of the hard detractors cluster is much more gradual and constant.

Beside differences in network structure and behaviours, the two communities clearly diverge in terms of values and identities. By analysing the keywords present in the members’ Twitter bios, a political divide between the two communities can be easily recognised: while hard detractors assert their national identity with mentions of Englishness, Britishness, conservatism and Brexit, soft detractors claim a European and progressive identity, with bios commonly featuring keywords relating to Europe, Labour and socialism. It is also interesting to observe how the two clusters contrast in terms of values, with the former making appeals to ‘freedom’ and ‘truth’, telling of a conspiratorial worldview confirmed by the presence of the names of fringe right-wing platforms like Parler and Gab, and the latter to ‘justice’. Lastly, it is interesting to note that the keyword ‘NHS’ was a frequent among soft detractors’ bios: indeed often it was healthcare professionals who criticised the Government on some aspects of the vaccine rollout.

The heterogeneity of the two communities was also clearly reflected in our SANTEE model, which analyses concerns of those expressing hesitancy about vaccines concerns in six categories: Safety, Agency, Need, Trust, Efficacy and Ethics. In applying the SANTEE model, we found that the two communities express very different types of concerns. While it is important to keep in mind that hard detractors authored abound double the number of tweets posted by soft detractors, the former generated a majority of tweets in all of the areas but one; Efficacy, where soft detractors expressed the majority (51%) of concerns. The main driver behind this was controversy around the government’s choice to delay second doses of the vaccine. Around the issues of Agency (the freedom to choose to be vaccinated) and Ethics, hard detractors almost monopolised the conversation, accounting for over 70% of the tweets.

The finding that hard detractors generated the most tweets in most areas, might lead one to assume that they had a bigger impact across hesitant conversations. But it would be wrong to do so, as soft detractors tweets had a significantly higher reach than hard detractors (almost 30 million impressions more). Given the tight knit nature of the community of the hard detractors conversation network, one can assume that a big chunk of impressions for this community were from individuals who are already part of vaccine sceptic communities, though further analysis on followers relationship would be required to confirm this.