Sophos Cloud Optix
I’m kind of like Bill Gates.
Just less neurotic, less conscientious and a bit more open to new ideas.
At least, that’s what startup Five Labs figures, having scanned my Facebook wall posts, photo captions and comments for words that predict the Big Five personality traits.
Those are the personality traits – openness, conscientiousness, extraversion, agreeableness, and neuroticism – that psychologists use to describe human personality.
After I clicked a button, Five Labs’ new tool, released last week, went to work crunching my Facebook blitherings, and then it offered to crunch all my friends’ verbiage, as well.
(Warning: NSFW, given its addictiveness. You’ll get sucked in, and we’ll never hear from you again.)
It presents the results in a warped pentagon, with each side assigned a percentage based on how much your personality reflects each of the five traits.
Here’s mine overlaid on the warped pentagon of my almost doppelgänger, Bill Gates:
In fact, with the click of a button, you can scan your friends’ Facebook personas, as well.
You can also choose to compare your results with that of not only Mr. Gates, but Mark Zuckerberg and Barack Obama.
If you’re thinking “targeted marketing on steroids,” join the club.
As the Washington Post noted, this type of tool could mean big bucks to advertisers, who could use it to get inside our heads and craft ads that appeal to us based on our psychological profiles.
The tool was inspired by a study conducted at University of Pennsylvania and published in September 2013 in PLOS One journal.
In that study, researchers led by H. Andrew Schwartz analyzed 700 million words, phrases and topics collected from the Facebook posts of 75,000 volunteers, who also took standard personality tests.
The researchers found linguistic patterns in the volunteers’ Facebook posts, pointing to the use of some words being strongly associated with certain personality traits, as well as gender, age and location.
For example, volunteers who lived in high elevations often talk about the mountains; neurotic people disproportionately use the phrase “sick of” and the word “depressed”, and males more often use the possessive “my” when mentioning their “wife” or “girlfriend” than females use the possessive with “husband” or “boyfriend”.
When it comes to personality traits such as the Big Five, this figure illustrates typical language associated with extraversion (e.g., “party”, “chillin'” or simply a string of exclamation points) and introversion (e.g., “computer”).
The same figure shows different words used with neuroticism (e.g., “hate”, “sick of” and “nightmare”) vs. emotional stability (e.g., “blessed”).
The researchers found that the models they created to analyse language used in Facebook postings are “surprisingly accurate” – for example, they hit the target 92% of the time when predicting users’ gender based only on the language of their status updates.
Those models are the same upon which Five Labs based its Facebook analysis tool, which will be released as a mobile app in August, using data from volunteers, reports the Washington Post.
Nikita Bier, the founder of Five Labs, told the newspaper that after the tool got inside participants’ heads via their Facebook postings, the biggest response was:
This is kind of creepy.
Bier reassured the newspaper that the company doesn’t plan to sell users’ data or use it to create targeted ads.
In fact, the company doesn’t keep user data at all, Bier said.
That’s sort of reassuring, at least in the short term.
Companies need to be upfront with how private data is being used and allow users to then opt out, said Andrew Schwartz, the University of Pennsylvania researcher who advised Five Labs on its project.
The Washington Post quotes him:
It's important that people know how their private data is being used and consent to it. People probably want to know if an advertisement, for example, was specifically selected for them based on their personality ... it's important such applications have policies to inform people how their data is being used and provide means of agreeing to such use.
It’s hard to imagine that marketers and big data brokers won’t want to cash in on the concept of language analysis sooner or later.
And at that point, expect them to be able to get even further inside your skull than they do now.
I did it. I was both reserved and open, yet neurotic. I didnt find ir creepy per se but i was curious about how they weighted words to reach the conclusions.
If I say I ‘Hate’ or I’m ‘Sick Of’ all advertising I wonder if that mean’s they’ll leave me alone finally, or use that to determine my mind set to work out what adverts can be targeted at people who loathe advertising?
Do people even have time to do these kind of stuff? I wonder if there is an appl which tracks the amount of time people waste on FB and says something like “You could have mowed the grass in the same time as you spend on FB last week” or “You could have slept an extra of 26 hours last month if you weren’t on FB”.
It isn’t creepy. The Five Labs tool is, however deeply methodologically suspect.
For example, their “agreeableness” trait should be restated as “Christian-ness”, given that the words they use to score it (according to their own site) are “excited • blessed • a • great • wonderful • amazing • prayers • a • wonderful • an • awesome • the • lord • had • a • great • an • amazing • great • day • excited • for • thank • you • in • christ • thanksgiving • psalm • so • excited • love • you • all • proverbs • merry • christmas • was • amazing • church • praise”
I think what that shows is not suspect methodology, but sample bias. If they ran their personality test in a language other than English, they would probably have significantly different correlations.
It also seems to indicate to some degree what people use their Facebook profiles for — people who belong to non-profit organizations tend to use Facebook a lot for keeping in touch and organizing events, and at least in North America, a large proportion of those non-profit orgs happen to be related to Christian organizations.
So I’d go more with “revealing” than either “creepy” or “suspect”.
Of course, if they don’t continually update their test set as they bring more individuals on board, the data will become less and less relevant.
You have a good point, Andrew.I was thinking about my Facebook persona vs. my non-FB persona in a similar way: many of my posts are made to share Naked Security articles, which are biased toward “eek!” “o no!” “privacy gone to hell!” “beware!” security matters. But I do think that the research is revealing, if only to analyse the personality side that kind of professionally worries about things.
there is quite an unscientific bias in the way they use semantics. You may also take different tests and compare. You will get very different results…
I post stuff without comment all the time. I guess it can poll the 15 times I’ve actually put comments and build my entire personality off of that.
I wonder, if this catches on for targeted advertising, will people start posting and commenting in Facebook in different ways on purpose? For example, phrasing all their comments for 2 weeks as Captain Jack Sparrow, then spend a couple weeks posting with all their wording and phrases sounding like Homer Simpson. You could have some real fun skewing the information that these systems would try to compile on you if you purposely wanted to.