After Trump’s surprise election in 2016, Eric Silverman and I wanted to look at Twitter to try and see if it would give us many insights into why this happened. We found a new group of Trump supporters added to, and effectively took over from, an older group of more traditional Republican Party (GOP) supporters during Trump’s campaign. Surprisingly, there was little evidence of Russian infiltration in this new group. We argue that a highly motivated movement of disaffected people played an important role in generating a new wave of support for Trump. The addition of this new support to traditional GOP support, was enough to push him over the bar in several key states in the election.

For more information, we wrote the work up for publication in PLOS One. We also wrote a follow up article in The Conversation.

We found that accounts in the older Republican group (see a in previous figure) switched from following one-another to following accounts in the Trump group (see b in previous figure) .

I have just started as a fellow at Newspeak House. I thought it would be interesting to build a map of the accounts clustered around Newspeak’s Twitter Page. The map I generated shows all the different civic groups which are associated with the Newspeak and how they follow one another. Everything from AI, machine learning, and open data groups, to journalists, social entrepreneurs,  games, arts and the public sector. Political technology…

The grouping is done so that members of each group tend to follow other accounts in the same group. The widths of the arrows and self-loops represent how often accounts from one group follow accounts in the linked group. A second algorithm annotates each group according to which biographical words they use more commonly than the other groups.

I was interested in looking at how cooperative or community organisations use Twitter. Similar organisations will tend to follow each other on Twitter so it might be possible to generate groups of cooperative accounts that share something in common. This can be useful for helping people discover cooperative organisations, or connect into networks of cooperative organisations that are already there.

I sampled over 200K accounts from Twitter which were clustered around the CooperativesUK twitter account. Here’s a map of the accounts I sampled.

The map shows the Twitter accounts collected into groups which share similar interests. As you can see, many of the groups are associated with a particular local area. Other groups are associated with topics like town planning, farming, sustainability, charities, finance, food, social care, etc.

The grouping is done so that members of each group tend to follow other accounts in the same group. The widths of the arrows and self-loops represent how often accounts from one group follow accounts in the linked group. A second algorithm annotates each group according to which biographical words they use more commonly than the other groups.

Executive Director and Senior Research Scientist at the Observatory on Social Media at Indiana University.

I am interested in human social, group and political behaviour. I apply techniques from mathematical modelling and data science to this topic. I have done work identifying clusters of people on Twitter and understanding their language use. I have developed ways of sampling groups on Twitter and characterising them automatically. This can have great value for tracking social/political groups, understanding their dynamics, and targeting members of those groups.

I come from an academic background working in Computer Science and Mathematical Biology fields. I’ve studied social behaviour in a broad range of animals from slime mould, bees and aphids to online human behaviour on Twitter. This included studying how bee colonies are affected by stress from pesticides and other diseases. We showed how colonies subject to stress from pesticides can often be on a knife edge between growth and collapse.