Facebook2Cprivatetraitsandattributes_Predictionsfromdigitalrecordsofhumanbehavior-Journalist27sResourceJournalist27sResource.pdf

10/26/15, 6:11 PMFacebook, private traits and attributes: Predictions from digital records of human behavior – Journalist’s Resource Journalist’s Resource

Page 1 of 3http://journalistsresource.org/studies/society/internet/private-traits-attributes-predictable-digital-records-behavior

INTERNET, SOCIAL MEDIA

Facebook, private traits and
attributes: Predictions from
digital records of human
behavior
Tags: facebook, privacy, technology | Last updated: March 25, 2013

(iStock)

Governments, businesses and policy makers are increasingly
able to capture and organize vast quantities of personal
information — a trend known as “Big Data” — to help them
make better-informed decisions. However, this data can also
pose a serious threat to individual privacy. In a widely
publicized 2012 episode, the retailer Target was able to
guess from a young woman’s recent purchases that she was
pregnant, before she’d even told others she was expecting.
Even criminal enterprises have gotten in on the trend.

Data gathering extends far beyond simply tracking credit card purchases and includes the
information people share online about themselves and others. A 2007 report by the Pew
Internet and American Life Project found that 60% those surveyed were not concerned about
what was available about them online and 61% do not try to control it; however, by 2012, Pew
found that many people were employing privacy settings to reduce the information available
about themselves. Other research has indicated that those who use social media sites are
typically more relaxed about privacy settings and often feel a false sense of control.

A 2013 study from the University of Cambridge [U.K.] and Microsoft Research in the
Proceedings of the National Academy of Sciences (PNAS), “Private Traits and Attributes Are
Predictable from Digital Records of Human Behavior,” examines the degree to which information
available online can successfully predict an individual’s personal — and private — attributes. The
researchers correlated public records of Facebook “Likes” from more than 58,000 users with
results from personality and intelligence tests and information from public profiles. The study

http://journalistsresource.org/studies/society/internet

http://journalistsresource.org/studies/society/social-media

http://journalistsresource.org/tag/facebook

http://journalistsresource.org/tag/privacy

http://journalistsresource.org/tag/technology

What is Big Data? Research roundup, reading list

Emerging cyber threats report, 2013

http://pewinternet.org/Reports/2007/Digital-Footprints.aspx

http://www.pewinternet.org/Reports/2012/Privacy-management-on-social-media/Main-findings/Main-findings.aspx

Can social navigation inform online privacy preferences?

Misplaced confidences: Privacy and the control paradox

http://www.pnas.org/content/early/2013/03/06/1218772110

Home

10/26/15, 6:11 PMFacebook, private traits and attributes: Predictions from digital records of human behavior – Journalist’s Resource Journalist’s Resource

Page 2 of 3http://journalistsresource.org/studies/society/internet/private-traits-attributes-predictable-digital-records-behavior

focused on whether a user’s history of Likes could accurately predict sexual orientation, ethnic
origin, political views, religion, personality, intelligence, satisfaction with life, substance use,
age, gender, relationship status, and “whether an individual’s parents stayed together until the
individual was 12 years old.”

Key study findings include:

The researchers were able to accurately predict a user’s sexual orientation 88% of the
time for men and 75% for women. While less than 5% of user profiles were explicitly
linked to gay policy or advocacy groups, “predictions rely on less informative but more
popular Likes, such as ‘Britney Spears’ or ‘Desperate Housewives’ (both moderately
indicative of being gay).”

The model was able to predict a user’s ethnic origin (95%) and gender (93%) with a high
degree of accuracy. “Patterns of online behavior as expressed by Likes significantly differ
between those groups, allowing for nearly perfect classification.”

The model predicted whether a user was Christian or Muslim (82%), a Democrat or
Republican (85%), and used alcohol, drugs or cigarettes (between 65% and 75%), and
was in a relationship (67%) with a high degree of accuracy.

The model was less accurate when attempting to predict the length of the parents’
marriage (60%). “Individuals with parents who separated have a higher probability of
liking statements preoccupied with relationships, such as ‘If I’m with you then I’m with
you. I don’t want anyone else.’”

The low prediction accuracy for satisfaction with life (only 17%) may be linked to an
inability to separate short-term affect (such as bad mood or mood swings) with longer-
term happiness.

“Although liking ‘Barack Obama’ is clearly related to being a Democrat, it is also relatively
popular among Christians, African-Americans and Homosexual individuals.”

The researchers caution against the potential negative outcomes that ready access to this type
of personal data might have: “Commercial companies, governmental institutions, or even one’s
Facebook friends could use software to infer attributes such as intelligence, sexual orientation or
political views [that] could pose a threat to an individual’s well-being, freedom or even life.”

In a related study, “Silent Listeners: The Evolution of Privacy and Disclosure on Facebook,”
researchers found that although Facebook users exhibited more privacy-seeking behaviors, this
was offset by changes to Facebook’s privacy policy. “The amount and scope of personal
information that Facebook users revealed privately to other connected profiles actually
increased over time and because of that, so did disclosures to ‘silent listeners’ on the network:
Facebook itself, third-party apps, and (indirectly) advertisers.”

Tags: technology, Facebook, privacy

http://repository.cmu.edu/cgi/viewcontent.cgi?article=1098&context=jpc

10/26/15, 6:11 PMFacebook, private traits and attributes: Predictions from digital records of human behavior – Journalist’s Resource Journalist’s Resource

Page 3 of 3http://journalistsresource.org/studies/society/internet/private-traits-attributes-predictable-digital-records-behavior

Writer: Margaret Weigel | March 25, 2013

Citation: Kosinskia, Michal; Stillwell, David; Graepelb, Thore. “Private Traits and Attributes
Are Predictable from Digital Records of Human Behavior.” PNAS, March 2013. doi:
10.1073/pnas.1218772110.

http://journalistsresource.org/author/margaret-weigel