In Spike Jonze's science-fiction film Her, two star-crossed lovers -- an intelligent computer operating system and a man -- embark on an impossible relationship. Reflecting human-like capabilities, the operating system, personified by female voice Samantha embeds itself deeper into unsuspecting Theodore Twombly's psyche and life.
Switch over to present day reality, where a new study published today in the journal PNAS, reveals that by mining Facebook Likes, computers can suss out your personality traits better than your nearest and dearest.
Not quite an operating systems love story, but surreal enough.
By deploying a new algorithm, researchers from the University of Cambridge and Stanford University have "calculated the average number of Likes artificial intelligence needs to draw personality inferences about you." And they assert that these mean machines can do it as accurately as "your partners and parents."
For this experiment, the researchers took a sample of 86,220 volunteers on Facebook who provided access to their Likes, and who completed a personality questionnaire through the "myPersonality" app.
The results supplied "self-reported personality scores" for the "big five" traits of psychological practice.
Namely; openness, conscientiousness, extraversion, aggreeableness, and neuroticism. This data allows researchers to assess which Likes equate with which traits. For example, liking "meditation" reveals higher levels of openness.
The next stage allowed the friends and family of users of the "myPersonality" app to judge the psychological traits of the user with a shorter version of the questionnaire. In this case, ten items versus the 100-item list completed on the "myPersonality" app.
To ensure fairness and accuracy, the researchers corroborated these online personality judgements with a "meta-analysis of previous psychological studies over the decades, which analysed how people's peers and family judged their personality.
The researchers determine that if there are enough likes, the computers can gauge a participant's self-reported personality better than their family or partner. "Big Data and machine learning provide accuracy that the human mind has a hard time achieving," says Dr Michal Kosinski, co-author and researcher at Stanford. "Humans tend to give too much weight to one or two examples, or lapse into non-rational ways of thinking."
Labelling it as an "important milestone", which might pave the way towards more social human-computer interactions, the researchers acknowledge that the findings reveal how computers can unravel a person's psychological traits through pure data analysis.
So what does this all mean for human-computer interactions? According to lead author Wu Youyou, from Cambridge's Psychometrics Centre, in the future, computers could judge our psychological traits and react accordingly. "In this context, the human-computer interactions depicted in science fiction films such as Her seem to be within our reach," says Youyou.
The researchers assert that automated, accurate and cheap personality assessments could have multiple impacts on societal and personal decision making. "The ability to judge personality is an essential component of social living -- from day-to-day decisions to long-term plans such as whom to marry, trust, hire, or elect as president," noted Dr David Stillwell, a Cambridge co-author.
This research reflects the potential for AI to get to know us intimately by mining through our data, yet researchers also acknowledge that concerns over privacy may be stoked as this technology develops. "We hope that consumers, technology developers, and policy-makers will tackle those challenges by supporting privacy-protecting laws and technologies, and giving users full control over their digital footprints," asserts Kosinski.
This article was originally published by WIRED UK
