NHS Debate on Twitter — One Month Data
In a previous article I examined the Twitter conversation in a one week period 7th-13th March.
In total we collected a data set for a period of 25 days 22nd Feb — 17th March. In total 670k Tweets collected on the topic of NHS.
In the first week analysis I identified that there is a very vociferous community of Twitter Users who are tweeting about Save the NHS issues which includes privatisation and the junior doctors contract dispute. Here they are shown in a Twitter map shown in blue.
Over any one week period there are factors and events which have a significant effect on the nature of the conversation overall. In this case the week of 7th-13th March there was a junior doctors strike. We can see the evidence of this in the conversation in several ways.
In this enlarged section of the map it can be seen that there is a very significant volume of connections with Jeremy Hunt who is the focus of attention as the Secretary of State for Health.
Secondly the topics identified (by automated machine learning) clearly indicate that junior doctors were some of the main topics of conversation during the period.
By analysing a longer period of time it would be expected that the week to week variations will be smoothed. Depending on the specific area of interest we may wish to focus on these variations i.e. understand what are the topical issues of the day. However, I will focus on the smoothed out conversation over the whole 25 days to get a better understanding of the zeitgeist and the most important influencers overall.
Initial Impression
The first impression is that there is a similar structure to the conversation as in the single week’s analysis. With two big communities as highlighted here along with some of the smaller ones together with some of the topics of discussion.
In the first analysis I concentrated on the the four main communities which are visible at this level. In the months data there are a few more identifiable at this level but as stated above the structure is broadly the same.
The communities indicated by the different colours are detected by machine learning community detection algorithms. Whilst only the main ones are shown at this level there are many thousands of communities of various sizes identified and consolidated into these main groups.
By looking at a more detailed breakdown of the communities it is possible to reveal a bit more structure about smaller communities and the topics of interest within them.
Zooming in to the NHS Operations section of the map some of the sub communities are visible.
Highlighted here are some of these and the influential Users and the topics of interest within these groups.
The top community is clearly highlighted in green. The most important User in this community is Geraldine Strathdee.
As well as the visualisation of the community we can also analyse the members of that community in more detail in our dashboard.
Here is a list of the top members of the community which further illustrates that there is a coherent topic of interest around mental health.
Top Influencers
Here are the tables showing the top influencers in the overall network.
One issue identified in the first post on this debate was the lack of engagement from those who could be broadly categorised as pro privatisation or supportive of the Government position on the junior doctors dispute.
In short the green community in this map consists of Users Tweeting in support of a ‘Save the NHS’ and ‘Support Junior Doctors’ agenda. What is completely absent is any form of alternative view.
It could be that there is no other view and that everyone in the UK supports the Save the NHS campaign. Given the fact that the current government is only recently elected and the Save the NHS is broadly anti-government I think that is probably unlikely. I think it more likely that there is a strong Confirmation Bias within the community and little effort to engage with those who have differing opinions.
The Inverse Michael Buble Effect
This is a theory to which I have given the name The Inverse Michael Buble Effect — here he is:
The Michael Buble Effect is a very positive one for recording artists. It means that an artist has a fixed number of fans (in this case 2.3m Twitter Followers). The objective is to sell his records. An echo chamber is exactly what is required to achieve this. 2.3m people all talking about how much they love Michael will sell a lot of records. The point is that it doesn’t matter how much the remaining 7 billion people hate his music — in fact the more they hate it the better as it will only make his fans more devoted.
The Inverse Michael Buble Effect is the opposite. In a political situation you have to convince the people outside the echo chamber who have a vote. It is pointless just strengthening the cohesion within the group as it will never increase the number of supportive voters.
Additionally the increased devotion to the cause has a negative effect of groupthink, intolerance of others and confirmation bias which means that communicating to the people that matter is ineffective.
The analysis of the longer period does not show any evidence to disprove the concern of an echo chamber.
Of course it is not possible to prove a negative. So, we cannot prove that there is no engagement that we have missed. What we can show is that there is no evidence to support a wider engagement so far.
What we can say is that a large data set of over 600k Tweets the lack of any evidence to show a wider engagement strongly suggests that there is very little taking place. We can also say that there is a lot of noise within this community which, at best, is wasted effort at worst a serious reputational risk.
In addition I have been asked by a few people involved in the campaign how to go about reaching other groups of people to spread the message.
How to spread your message
In (very) simple terms there are nodes within any network that connect different groups — connectors.
Here is one in this network.
The map shows that Roy looks like he connects two groups — the algorithm that detects this also shows him as the top individual connector in this network.
Here is a view of the network created just from Users connected to Roy.
Roy is a key influencer in both the main communities in this network. In order to reach into other communities you need to find Users who are influencers in the communities you are trying to reach. Ideally people who are already involved in this network.
Reaching people in communities who may have strongly differing opinions is tough. These connectors are the bridgehead, but identifying them is not straightforward.
Women in Healthcare
My analysis of the World Economic Forum in Davos focused on the prominence of Women in business promoting gender equality issues.
Here is one of the network maps in that analysis highlighting some of the prominent women in that network.
I also noticed that there are some prominent women in the network for the NHS conversation.
The women in the WEF network are primarily business focused, the women in the NHS primarily healthcare. There is also a clear overlap e.g. Melinda Gates.
It is a purely personal observation that business issues appear to be poorly understood by the Save the NHS campaign.
The primary concern of the Save the NHS campaign is privatisation of the NHS. As a result there is a great deal of content tweeted warning of this danger. However, the content almost universally demonises private business making no distinction between good and bad private companies.
Over 80% of people in the UK work for private businesses of which 99.9% are small or medium sized businesses. These businesses and their employees bear no relation to the corporate giants which (rightly or wrongly) are cause for concern.
The maths is pretty simple — most people who have a vote work in businesses which are demonised daily by the Save the NHS campaign.
This does not seem to make much sense if those people are the ones who need to vote to support the policies being promoted by the campaign.
It would also suggest that there are many people in business who share the values and objectives of the Save the NHS campaign and would be very useful in spreading the message into new communities of people currently not engaged.
So to put these ideas together, there are a group of women in the healthcare and business communities who will share many common interests and objectives. Also, critically the women in business will have a bridge to help influence others in business and spread the message.
Finding the Influencers in Other Topics of Interest
We use a (free) service called Right Relevance. To find people who are influential in specific topics of interest.
Here is screenshot of the list of influencers for Women in Business. You can find the list here.
Once you have an objective of reaching out to an identifiable group and you can discover who the influencers are in that group the rest is down to each individual to make relationships and spread the word.
Conclusion
Firstly, in response to feedback from my previous article I should point out that this post and the previous one are intended as discussion pieces to illustrate some of the techniques and power of network analysis and graph theory.
The list of things I have omitted is very long indeed. I am very interested in feedback and criticism of my analysis or methods. However, I am not interested in joining the debate as that would conflict with my objective of observing the structure and nature of the network
So, here is a quick summary of the issues I have covered.
I have demonstrated how our analysis tools can identify communities and the topics of interest; and influential people within them.
This is an automated machine learning process which provides information and is guided by a human analyst. This means it starts with no preconceptions but identifies the information by analysis of the information contained within the network.
In this case I showed how this process illustrated some major topics within the conversation about the NHS in the UK.
One of the main topics of interest is the Save the NHS campaign. I showed that there is some evidence of an echo chamber effect in this conversation. In order to combat this problem I showed how to implement a process of identifying influencers in other topics of interest who may be useful in spreading the message beyond the group of people who are already convinced.
It is important to note that the women in business is just a single example of how this could work. There are as many different approaches to this as there are people in the Save the NHS campaign to make connections.
Imagine how powerful the campaign would be if everyone involved contributed to spreading the word to the people who count — those whose minds you need to change.
Where is Wally?
So one mystery remains. Where is the alternative view? Twitter is an imperfect reflection of life and you would expect a major public debate to have both sides reflected — e.g. Brexit.
There are many plausible theories as to why they are not present in this network worthy of further investigation
There is one very dangerous one, however. This one:
History teaches us that there is a great danger in underestimating the strength of the opposition we face.
The great danger of living in the filter bubble of an echo chamber is starting to believe your own stories.
I am sure George Foreman believed he was landing a lot of punches and scoring well — don’t be George.
Footnote
I really couldn’t finish this blog without mentioning the achievements of the NHS Choir.
The more observant reader may have spotted this.
Here’s why:
So well done to NHS Choir.
In one picture this is the way to make a difference. You might need a bit of luck to get Justin to come to your meeting, but everyone can do something genuinely unique and engaging with a bit of thought and a lot of hard work.
