Thursday 3rd May 2012 | 10:58
LSE warns Twitter not representative of public opinion
LSE press release:
As London elects a new mayor today, new research from LSE’s Media Policy Project raises concerns about the way data mined from Twitter and other social networking sites is used by the media during elections.
During recent Mayoral debates Twitter buzzed with #MayorDebate, #BBCLondon, and #SkyLondonDebate related comments. But research by LSE’s Nick Anstead and Royal Holloway’s Ben O’Loughlin points to problems with the way these comments are quantified, interpreted and repackaged for the public by companies they call “Semantic Pollsters” -
“ ‘Semantic polling’ is the employment of natural language processing technology to ‘read’ and codify vast datasets gathered online, and then the use of this data to illustrate and understand public opinion.”
The researchers find little consistency in the way this data is used, and serious concerns about the way it is communicated to the public. They maintain it should not be presented as representative of overall public opinion. This is particularly important when one considers that reports on opinion polls by the media can drive, as well as reflect, public opinion itself.
Although Anstead and O’Loughlin see great potential in the future development of semantic polling, they warn that:
“Both those carrying out semantic analysis and those in the media reporting it have a responsibility to offer appropriate explanations about the meaning and limitations of the conclusions, and the methods used in data analysis.”
Among other recommendations, the new report calls for an increase in media literacy among citizens; increasing data literacy among journalists and editors; and a greater level of methodological transparency for pollsters.