Facebook is opening up its data for the first time to more than 60 academics from around the world, who will be able to look, among other things, at how the platform influences elections.
The world’s largest social network, used by 2.4bn people each month, will allow academics to see which websites were linked to by Facebook users from January 2017 to February 2019.
They will also be able to look at the system that advertisers and marketers use to plug into Facebook’s data, and at Crowdtangle, a tool that shows publishers how content is spreading across the site.
The academics, drawn from 30 institutions, were chosen by the Social Science Research Council, a non-profit US organisation. They include researchers who are studying the role that Facebook played in elections in Italy and Germany.
The researchers will have access to one of the biggest troves of information about human behaviour online, and the move may pave the way for other companies with enormous data sets to follow suit.
It comes as Facebook has sought to fend off criticism that it has been too opaque about how its platform has been a host to fake news, following evidence of co-ordinated Russian disinformation campaigns designed to manipulate the 2016 US election.
However, the timeframe selected by Facebook for the links database means the academics will not be able to analyse the most contentious information from the run-up to the 2016 US election or the UK’s Brexit vote.
Facebook has also had to balance concerns over user privacy in the wake of the Cambridge Analytica data scandal.
Facebook’s social science initiative was initially launched a year ago, but faced delays as the company weighed up how to guard against privacy breaches and guard user confidentiality.
Chaya Nayak, Facebook’s strategic initiatives manager, said the company had been “slow out of the gate on the project” in order to spend more time getting “privacy right”.
Facebook’s data sharing partnerships were attacked after it emerged that Cambridge Analytica, a political consultancy that worked for the Trump campaign, had improperly obtained the personal data of some 87m Facebook users via a third party.
On Monday, Facebook said it planned to apply so-called “differential privacy” to the way in which researchers could query its links data, by introducing “noise” to the information that prevents them from being able to personally identify individuals.
Ms Nayak said that the timeframe for the URLs database, which does not include the US 2016 election, was selected in order to “prioritise getting the data out as fast as possible,” adding that the possibility of extending the data set further back in time was “under discussion”.
While academics will not require approval from Facebook before publishing their insights, Ms Nayak said that the company would log how the database is used to ensure it is not abused by academics, who will be required to sign strict legal agreements.
Get alerts on Facebook Inc when a new story is published