Email
Password
Remember meForgot password?
    Log in with Twitter

Op-Ed: Using Facebook 'Likes' to predict personality can be dangerous Special

By George McGinn     Aug 9, 2014 in Technology
Speaking the truth in times of universal deceit is a revolutionary act, so said George Orwell, author of "1984."
Computers and the Internet have not only made the most significant advances in our lives possible, it has also been an unwitting betrayer.
It betrays so much that the NSA, Google, Facebook, Microsoft, Apple, Netflix, Hulu, and many other social media companies are finding new ways to use data collected from us for everything from customizing the ads we see, to recommending movies for us to watch, to finding terrorists or profiling criminals.
And while people like Julian Assange and Edward Snowden who have done their revolutionary act by revealing the truths about the NSA's spying on American's program, this leaves it up to the media to pick up where they left off and tell the truth behind the universal deceit going on behind the scenes of websites that may be storing your data for future use.
It is this "universal deceit" that helped create Wikileaks, and caused Snowden to expose the NSA's programs for collecting, storing and analyzing the data of Americans.
And a website that may look or sound benign may really be a wolf dressed in a sheep's clothing.
Take a study by a team at the Psychometrics Centre of the University of Cambridge in England who developed the website "www.youarewhatyoulike.com" that looks at yours and all your friend's Facebook profiles and data and attempts to map out your personality.
Shows one of the methods of taking Facebook likes and predicting who you are.
Shows one of the methods of taking Facebook likes and predicting who you are.
Michal Kosinski, Psychometrics Centre of the University of Cambridge in England
As it says on its website: "It doesn't ask you any questions at all. It looks at the things you like on Facebook, and then tells you who you are. Because you are what you like."
The fact that they do not need any input on you and based their predictions on your past actions is another example of an invasion of your privacy based on your online persona.
The application was based on a study written about in a 2013 paper, "Private traits and attributes are predictable from digital records of human behavior" published in the Proceedings of the National Academy of Sciences (PNAS).
Authors David Stillwell, Michal Kosinski and Thore Graepel claims that since people have migrated to using digital medium to communicate, network with each other and store personal biographical and geographical information, it makes the ability to predict personality traits possible.
The new application uses a database created from a prior study of Facebook profiles and created a formula that cross-references responses between you and your friend's "likes."
The application uses 15 factors from you and all your friend's Facebook activities including: public profiles; likes; friend list; email address; relationships; relationship interests; birthday; work history; education history; hometown; interests (books, music, TV and movies, etc.); current city; religious and political views; and relationship details.
All this data is then put into a formula and uses the "Five-Factor Model" to create a behavior profile score based on how you are rated by you and your friends ratings.
The "Five-Factor Model" is a standard used in psychology to predict:
• Openness (liberal and artistic vs. conservative and traditional);
• Conscientiousness (well organized vs. spontaneous and flexible);
• Extraversion (shy and reserved vs. outgoing and active);
• Agreeableness (warm, trusting and cooperative vs. assertive and competitive); and
• Neurotcism (calm and relaxed vs. emotional and stressed).
And according to the PNAS paper, it is possible to predict additional traits from the digital footprint you and your friends leave on Facebook. Traits such as IQ, sexual orientation, life satisfaction, political and religious views, and more are predictable.
For example, the report claims it can tell the difference between gay and straight men 88 percent of the time, black and white users 95 percent of the time and Democrats and Republicans 85 percent of the time.
Herein lies the problem. The predictions have in many cases a large margin of error. The report's supplemental document explains it clearly:
"An important limitation of our sample is that some of the predicted variables are from Facebook profile information. Individuals who declare their political and religious views, relationship status, and sexual orientation on their profile may be different from non-declaring members of those groups; they may associate with distinct Likes, which may lead to an overestimate of prediction accuracies for these groups. Nevertheless, the model was still able to predict privately reported information, such as personality or intelligence quotient questionnaire results, and survey results on addictive substance use."
"Overestimate of prediction accuracies." Would you base important decisions on predictions of your personality and that of all your friends which are nothing more than a "statistical" estimate?
In many cases the study team's paper states that in many categories they are right only 60 to 75 percent of the time. The only item they get right above 90 percent is gender, and that only because it is a field a user can fill in.
And when the study team compared the results to known psychological questionnaires, the accuracy for many of the "Five-Factor Model" predictions drops even farther away.
For example, using Facebook to determine intelligence, the standard tests get this right by a factor of .78, where the system using Facebook gets it right by a factor of only .39. This means it is twice as likely to get intelligence wrong using Facebook data as it is using standardized tests.
However, the inherent danger is not in the predictions itself, but how they are used. Can an employer use the results to decide whether to hire or fire you? Can law enforcement profile you solely on your online persona's personality? Can you be killed because of a faulty or incorrect prediction?
Yes, yes and and most definitely yes.
According to a Career Builder survey done in 2009, 45 percent of employers used social media sites like Facebook, MySpace, Twitter, and LinkedIn to screen job applicants, even if the job seeker never supplied their online identities. The study reports that it is easy to find you online.
And based on a 2013 article by the Colorado Technical University, law enforcement uses social media sites, like Facebook, in profiling.
The PNAS report also pointed out a deadly consequence of even displaying some predictions. The example they gave is trying to determine if someone is pregnant.
If you are a single woman living in Turkey or some other Middle East country, if you log on you are blasted with ads showing pregnancy clothes, vitamins for prenatal care, and clothing for your newborn, and someone sees this, they may get the idea that you are pregnant, and if tribal law is followed, even if you are not pregnant, the impression that your online screen showed can get you stoned to death.
The paper also states it gets sexual orientation for gay men right 88 percent of the time, and 75 percent for lesbians. That means 25 percent of the time the prediction is wrong or a 1 in 4 chance in those same countries would get you killed if it said you were gay and someone sees it.
Your online data is also used by the federal government to determine if you are a threat to the country, as the media has reported many times throughout the years that the NSA has been profiling people using the same data.
But not all use of online data is harmful or deadly. For example, in the Colorado Technical University also reported that police departments use social media to provide crime information and warnings of wanted criminals to its citizenry.
And a Facebook post in 2009 by Detroit resident Mary Evans on an alumni page led a group of Detroiters into a series of events that eventually freed two brothers wrongfully convicted of murder. Tommy and Ray Highers spent 25 years in prison for a murder they did not commit.
However, any system that uses data posted on social media sites like Facebook, LinkedIn, Google+, and others is only as accurate as the person posting to it. And it's well-known that some Facebook or other social media site's profiles can be fake and hard to tell whether the information provided is accurate.
Online, introverts have been known to become extroverts, men have created women profiles, and child molesters have been known to create profiles to pretend they are teens or children. Because of the impersonal nature posting to online forums allows, hiding behind a computer screen can empower someone to be the person they want to be rather than who they really are.
Why is this important and how does it affect you? Imagine that one or some of your friends are phony and are commenting on your posts and like certain posts based on their false persona. They are used by the application to help predict your personality traits.
Take the case of Jerrod Durr, who for three years used an online social media site to impersonate a war veteran and conned women and businesses out of more than $700,000 writing bad checks.
Durr, of Willard, Ohio, was 18 at the time when he first created a fake online profile of an Iraqi War veteran SSgt. Eli Carter. He used his fake online persona to prey on women online and gain their trust. Out of hundreds of women he communicated with he eventually met four women, and continued the con by dressing the part. He wound up convicted of fraud and theft in four states.
By using a fake social media profile to hunt for his targets, Durr stole the honor and respect we afford those in the military to perpetrate his crimes.
He also posted phony information about himself and on your posts, which the application takes into consideration in determining your personality. And no application can spot a fake profile, as four intelligent women fell for it.
So whether you are denied employment or mistaken by the police for a wanted criminal, the use of your online data isn't going to stop. In fact, its use will continue to grow as more data is stored and people create more ways of using it.
Now, more than ever, you need to be more security conscious of your online activities, deciding how much of you to reveal and to whom you allow access to your data. And you need to start assuming that every time you visit a website, make a post online, upload photos, information about you is being collected and stored.
Even by taking the study team's Facebook personality test they are collecting and storing data about you to analyze later.
I can imagine a scene where George Orwell is sitting on his front porch in his rocking chair, smoking a cigar and now reading the latest news from an iPad. He looks up and says "I warned you all."
This opinion article was written by an independent writer. The opinions and views expressed herein are those of the author and are not necessarily intended to reflect those of DigitalJournal.com
More about online privacy, teens online privacy, facebook privacy, online data, misuse online data
More news from