‘Lie detector’ to test social media posts

Warwick is collaborating with Sheffield and three other universities to develop a system which will automatically verify online rumours.

The three-year project, called Pheme, will analyse rumours that appear in different forms on social media such as tweets, videos, pictures and blog posts.

Warwick professor of Social Informatics, Rob Procter, who is involved with the research of the project, has previously worked with LSE and the Guardian’s interactive team to manually analyse the spread of rumours on Twitter during the 2011 London riots.

Rumours such as whether the rioters let loose animals from London Zoo and whether the high street shop Miss Selfridge was set on fire (false and true respectively), were manually assessed in terms of their reliability.
This took several months – the new Pheme system aims to investigate speculations on social media in real time as events unfold.

The project, funded by the EU, will classify rumours into four types: speculation – whether interest rates may rise; controversy – such as over the MMR vaccine; misinformation, where untrue information is spread unintentionally; and disinformation, where it’s spread with pernicious intent.

It is hoped that Pheme will be particularly useful for journalists, because it will categorise sources to assess their authority on an emerging story.

Professor Procter told the Boar: “One key role that Warwick has is to produce case studies of rumours in social media. This task involves analysing and annotating the content of social media postings (e.g. tweets) to identify topics and to distinguish the different ways in which other users make sense of and respond to them.

“These case studies will be used to model how rumours spread across social networks and to train algorithms for assessing their reliability automatically.”