23 July 2012
A multi-disciplinary team of researchers has been studying whether it’s possible to detect psychopathy in people’s tweets. Results suggest that in certain contexts, personality prediction through social media can perform with a “reasonably high degree of accuracy”.
Computer scientists at Florida Atlantic University, the Online Privacy Foundation and Kaggle teamed up to see whether it was possible to identify the “Dark Triad” of psychopathy, narcissism and Machiavellianism through people’s Twitter usage. The team will be presenting the research on 29 July at Defcon in Las Vegas.
The research was triggered by a paper from Cornell University that examined the speech patterns of known psychopaths describing their crimes. That study had found that pyschopaths tend to use more conjunctions like “because”, “since” or “so that”. They also used twice as many words relating to physical needs such as food, sex or money, while non-psychopaths used more words about social needs such as family and spirituality. Psychopaths were also more likely to use the past tense and be less fluent in their speech, using more “ums” and “uhs”.
Kaggle — a company that specialises in predictive modelling competitions — invited data scientists to submit models that could analyse people’s twitterfeeds. Some 113 teams submitted 1071 models to examine psychopathy in Twitter usage.
Each of the models analysed more than three million tweets from a final set of 2,927 tweeters from 80 countries. They looked at profile information, the number of tweets sent, re-tweets and replies, the user’s Klout score and the words in people’s tweets. They created psychopathy scores based on a checklist developed by Professor Del Paulhus at the University of British Columbia. In addition to volunteering to have their tweets analysed, participants were asked to take a personality test, which rated them from 1-5 for eight different traits, one of which was the “dark triad”.
Of the 2,927 people, around 41 users were certifiable psychopaths according to the personality test. Everyone else fell into a spectrum.
The results show that there are “a number of statistically significant correlations between an individual’s darker personality traits and their Twitter activity”, according to Florida Atlantic University’s Randall Wald. They also discovered links between users’ attitudes to privacy, their personality traits and their twitter use.
The team believes that they could potentially use analytical tools to identify potential trouble makers at events such as far-right demonstrations. However, they have ethical, privacy and human rights concerns about using social media analysis to vet or label people according to personality types. The team also recognises that there would be an inevitable issue with false positives on Twitter, so the technique shouldn’t be used to find and target psychopaths.
Chris Sumner from the Online Privacy Foundation told Forbes.com: “Just because someone scores highly doesn’t mean they’re criminally minded. Using this to try to spot someone who’s going to commit a crime is going to result in catching people who aren’t going to. But in general it’s interesting. You can use it to look at large groups of people and ask are we becoming more antisocial?”
A full paper covering Psychopathy, Narcissism and Machiavellianism will be around one month after the conference.
Update 17.40 23/07/2012: Following contact from Sumner at the Online Privacy Foundation (see his comment below), we have changed the headline from “Twitter analysis can be used to detect psychopaths” to “Twitter analysis can be used to detect psychopathy”. This is because the team studies psychopathic traits rather than clinically diagnosed psychopaths. You can read more about the team’s approach in this Online Privacy Foundation blog post.