In a study published on Jan. 12, Stanford and University of Cambridge researchers found that computers are better able to judge a human’s personality than other people are.
The researchers determined that a computer that is mining Facebook “likes” will do better in predicting personality than the person’s friends, family or colleagues. However, spouses did nearly as well as the computer in judging personality.
The study, “Computer-based personality judgments are more accurate than those made by humans,” focused on five areas of personality known as the “Big Five”: openness, conscientiousness, extraversion, agreeableness and neuroticism.
To determine relative ability to judge personality, researchers first collected self-assessments about their personality from 82,220 volunteers using a 100-item questionnaire. The human judges used a 10-item questionnaire to judge the subject’s personality while a computer mined each volunteer’s “likes” to produce a digital set of judgments.
The computer performed better than a work colleague (“better” meaning closer to the subject’s self-assessment) after analyzing 10 “likes,” better than a friend/roommate after analyzing 70 “likes,” better than family members after analyzing 150 “likes” and better than a spouse after analyzing 300 “likes.”
People on Facebook had, on average, 277 “likes” for the computer to mine.
The team included Michael Kosinski, a postdoctoral fellow in the computer science department, Wu Youyou, a doctoral student at Cambridge’s Psychometrics Center and David Stillwell, a Cambridge researcher.
The researchers note in their abstract that “computers outpacing humans in personality judgment presents significant opportunities and challenges in the areas of psychological assessment, marketing, and privacy.”
Contact Alice Phillips at alicep1 ‘at’ stanford.edu.