Interesting FB response; it seems that casuistic responses that are hard to algorithmetize are becoming more pertinent across different platforms and services [think of Google’s struggle to determine which search results to take down after individuals appeal to their right to erasure].
31 July 2014 by Hal Hodson
No one really knows exactly how Facebook decides what we see when we log in. Reverse-engineering the algorithm behind it could help us find out
WHO controls your Facebook News Feed? We are fed a specially selected diet of jokes, photos and gossip from our Facebook friends, but not by a person. Instead an algorithm does the work – giving it the power to influence us.
The furore over an experiment in which Facebook researchers attempted to manipulate users’ emotions via their News Feed, albeit only slightly, highlighted the extent of that power.
Facebook’s algorithms are a closely guarded secret. “These are black boxes,” says Christo Wilson of Northeastern University in Boston. “In many cases the algorithms are held up as trade secrets, so there’s a competitive advantage to remaining non-transparent.”
For Karrie Karahalios and Cedric Langbort at the University of Illinois and Christian Sandvig at the University of Michigan, Facebook’s influence is out of balance with our understanding of how its algorithm works. So they are carrying out what they call a collaborative audit, looking at the Facebook experiences of thousands of people to work out the underlying algorithmic rules.
To do this they have created an app called FeedVis, which creates a stream of everything that your friends are posting. When I tried it, I saw an endless stream of comments, likes and posts by friends I’d forgotten I had. To the right I saw my standard News Feed, which was empty by comparison.
In their first, small study using FeedVis, the team found that most people – 62 per cent – didn’t know that the News Feed is automatically curated. People were shocked that they weren’t seeing everything their network posted. In cases where posts of close friends or family were excluded, many became upset.
The team is starting to understand some of the basic rules that govern what people see. “We know that if you comment on someone’s wall, you’re more likely to see a post from them than if you just like something,” says Karahalios. “And if you go to a person’s timeline you’re more likely to see content from them later.” The work was presented at the Berkman Center at Harvard University last week.
But Facebook’s algorithms change constantly. “Even if I figure it out today, that doesn’t necessarily mean it’ll be like that tomorrow,” says Wilson.
To expand the experiment, the team will recreate a person’s profile based on their likes, comments and other Facebook activity and then see if they can detect patterns in what their News Feed shows them.
Already, Facebook appropriates its users’ profiles to create adverts on their friends’ feeds that look like normal content. There are other tricks, too. “I could share a link to the McDonald’s website, commenting that a McLobster sounds disgusting,” says Sandvig. If you like that link, Facebook registers that you like McDonald’s. “It doesn’t appear on your feed, but your friends will get ads that say ‘Hal likes McDonald’s’,” he says.
Understanding these dynamics is crucial, as Facebook is increasingly the tool that people use to communicate and find out about their world. “In the history of mass media, there have been channels with huge reach, but it’s typically a human in the apex of the control loop,” says Wilson. “That’s just not true any more.”
This article appeared in print under the headline “Facebook’s biggest secret”
VELTI has presented USEMP approach to the value of personal data to the 26th International Conference on Advanced Information Systems Engineering (CAiSE 2014), held in Thessaloniki, Greece, June 16-20, 2014! The presentation focused on USEMP use cases, USEMP approach to evaluating personal data value and presented to the community USEMP LIO platform architecture. The slides are available here. There is also multimedia material hosted kindly by FP7 Social Sensor project here.
The third face-to-face meeting took place in Thessaloniki (14-15/5/2014).
Mapping the patterns of sharing data in social networks , so-called data mining, giving businesses and stakeholders huge opportunities to reach individuals and specific target groups with different offers.
Your privacy and personal digital profile is the focus of an interdisciplinary research project , USEMP where Luleå University of Technology and the Centre for Distance -spanning Technology , CDT, collaborates with universities and companies in six European countries.
Apps that give you warnings how the data you provide in social media may be used by others, and tools to help you assess the data you provide into money , will be the result of the project, says Anna Ståhlbröst , senior lecturer of informatics at Luleå University of Technology. Technology can help citizens to take control of the data that leaks out when we are on-line. It also help individuals to do business and sell their data to various companies and stakeholders.
Edward Snowden’s revelations about NSA surveillance of citizens, pin points new issues around privacy online. Results from the USEMP project is also expected to contribute to the debate about mass surveillance . The USEMP project analyzes legislation linked to digital social networks , user’s needs and concerns , marketing and payment models and how new knowledge can be translated to new applications and tools to protect and support the individual user on-line.