Digging up facts about fake news: The Computational Propaganda Project


This may come as a surprise to most serious policymakers, but here’s a fact: not all that is “news” is fact-checked information. Worse, non-facts are frequently introduced into stories and passed off as facts. Welcome to the new information world. It is unsettling, and hardly augurs well for robust policymaking. So what can be done about it?


The phenomenon of junk news and its dissemination over social media platforms have transformed (some say destroyed) political debates. The combination of automation and propaganda, also called computational propaganda, can shape public opinion. The trouble is, how can we tell the difference between fake facts and real facts, and indeed, real fakes?


This is the question Samantha Bradshaw and her colleagues from the Computational Propaganda Project at the University of Oxford set out to answer when they analysed the distribution of junk news, including fake news, computational propaganda and ideologically extreme, hyper-partisan, and conspiratorial content, over the social media platform Twitter during the 2016 US presidential campaign in Michigan. As their findings show, junk news was shared to the very same extent as professional fact-checked news.


In this pioneering quantitative research on junk news, Ms Bradshaw and her colleagues studied Twitter conversations happening in Michigan, a swing state in the US presidential elections, between 1-11 November 2016. The research team was interested in finding out what people were sharing as political information and news. They collected tweets with website addresses (URLs), which were classified according to three categories: professional news outlets (both major and minor sites), professional political content (from political parties, experts, think tanks, government), and other political news, which included junk news and further sub-categories such as WikiLeaks and country-related links, notably from Russia. The team found that professional news content and junk news were shared in a one-to-one ratio, meaning that the amount of junk news shared on Twitter was the same as that of professional news.


©Rights reserved

Samantha Bradshaw of Oxford University’s Computational Propaganda Project looks at real trends in fake news


In comparison, the ratios of junk-to-professional news being shared in Germany and France were considerably lower, with a one-to-four ratio during the German federal presidency election and a one-to-two ratio during the French presidential campaign, both in 2017. But does this mean the Europeans are less gullible to fake news? Probably not. Rather, the machine of producing fake news that started in one country will likely spread to others.


Bridging the fact divide


According to Ms Bradshaw, “the proliferation of junk news threatens the core of democracy. We need to have a place where we can collect good, truthful information and come to decisions on politics”. With junk news being spread on social media, it is hard for people to actually develop clear opinions, as they are not receiving the correct information to make informed decisions. This confusion was pointed out by several voters in recent elections. The current situation could create “information inequality” between those accessing fact-checked content and those consuming misinformation about politics and public policies.


What can we do to curb the dissemination of junk news? According to Ms Bradshaw, the rising concerns over misinformation have to be tackled by co-operation between international organisations, research centres and the private sector. She argues that there is a real opportunity for organisations like the OECD and research institutions like hers to collaborate with the private sector and social media companies that possess large quantities of information on what people are reading and sharing.


However, Ms Bradshaw and her colleagues discovered that most companies keep the relevant data firmly closed. In their research, the Oxford Internet Institute’s team could only use the API (application programming interface) of Twitter, which covers 1% of the tweets. It is even more difficult to work with the API of Facebook. While privacy online is a key concern, accessing data is also crucial for understanding and improving the relationship between online information, voting behaviours and political positions. For Ms Bradshaw, more collaboration between international organisations, research institutions and the private sector is urgently needed to push back fake news with facts.


References and further reading

Bradshaw, S. et al. (2017), “Junk News and Bots during the U.S. Election: What Were Michigan Voters Sharing Over Twitter?” Data Memo 2017.1. Oxford, UK: Project on Computational Propaganda

For Data Memos on the German and French elections see

Howard, P. N. et al. (2017), “Junk News and Bots during the German Federal Presidency Election: What Were German Voters Sharing Over Twitter?” Data Memo 2017.2. Oxford, UK: Project on Computational Propaganda

Bradshaw, S. et al. (2017), “Junk News and Bots during the French Presidential Election: What Are French Voters Sharing Over Twitter?” Data Memo 2017.3. Oxford, UK: Project on Computational Propaganda


OECD Forum 2017 issues

OECD work on public governance

‌‌‌‌Rory Clarke

Balazs Gyimesi

OECD Yearbook

© OECD Yearbook


Related Documents