References to ancient Britain linked to hostility online

Political posts on social media that most frequently referenced ancient history tended to be more extreme, hostile and overwhelmingly negative in tone than average, finds a new study by researchers from UCL and the University of Edinburgh.

Their study, published in PLOS ONE, scrutinized nearly 1.5 million posts using a combination of AI, computational and manual techniques and used a range of sentiment analysis tools to evaluate the attitudes behind those posts that reference Britain's distant past.

Senior author Professor Mark Altaweel (UCL Institute of Archaeology) said, "Misrepresentations of history are widespread online. This research helps us identify when the past is used to advocate for hostile and extreme positions in political discussions.

"Ultimately, we hope that this work can help understand interpretations leading to dangerous discussion threads online."

Researchers examined often-heated debates around Brexit on Facebook as a test case to see how references to the Iron Age, Roman and medieval periods in Britain are inserted into online political debate.

Facebook users arguing for any specific side of a complex political issue are more likely to justify their positions through "historical thinking" if they hold extreme views.

Lead author Dr. Chiara Bonacchi of the University of Edinburgh, said, "The ancient past is often used to express political identities. Our research shows that various periods of pre-modern British history are most commonly invoked to advocate and justify negative and extreme positions.

"When people reference the ancient past in political debates on social media, it's usually connected with hostile and polarized views rather than with moderate ones."

The researchers collated almost 1.5 million posts published on 364 individual Brexit-themed Facebook pages between 2015 and 2017. These posts reflected both positive and negative attitudes towards Brexit.

From these, they identified more than 2,500 posts that contained at least one reference to Britain's Iron Age, Roman or early medieval periods.

These key terms included the names of Iron Age tribes and chiefs, Roman emperors and Anglo-Saxon kings, Roman towns, Iron Age and Anglo-Saxon settlements in Britain, as well as general terms describing the period from 800 BCE to AD 800.

The researchers then used a range of language processing algorithms to gauge whether the emotion of these posts was positive, negative or neutral, as well as the extremity of the sentiment.

They also spot-checked random posts manually to ensure that the algorithms were faithfully capturing the emotions of users.

Co-author Dr. Jessica Witte, a Post-doctoral Research Fellow at University of Edinburgh, said, "This study was challenging and required a combination of AI and manual analysis of text.

"There are a range of sentiment analysis techniques available, but these need testing to make sure that they are suitable for use in a specific communication context."