Korva is a scroll of the internet’s best ideas.
AI reads personality from diary text
A page of writing
In a recent study, people wrote roughly a page of diary-type text. Large language models like GPT and Claude analyzed it, with no other information about the writer, and produced a personality profile.
Scoring the Big Five
The profile covered the Big Five traits: extraversion, conscientiousness, neuroticism, and the others. The model's scores came close to the ground truth, about as good as a good friend or someone who knows the person well.
Conversations at Caltech
A similar study at Caltech used text conversations, a dynamic back-and-forth with a whole transcript and nothing else, to ask what could be told about a person's stable personality and current mood. It turns out you can tell a lot. What people say and write, and the order in which it unfolds, carries enough information that AI can estimate a personality from text alone.
How did the Big Five personality model arise?
The model came from the lexical hypothesis, the idea that the traits people care about most end up as words in the dictionary. In 1936 psychologists Gordon Allport and Henry Odbert counted nearly eighteen thousand English words describing personal traits. Later researchers, including Raymond Cattell and then Ernest Tupes and Raymond Christal in 1961, reduced those terms by factor analysis and kept finding five clusters. Paul Costa and Robert McCrae turned the five into a widely used questionnaire in the 1980s.
Why might everyday word choice reveal personality?
Psychologist James Pennebaker at the University of Texas built a text-analysis program that counts word categories in writing. He found small but reliable links between the use of small function words, such as pronouns, and traits and emotional states. Function words are produced almost automatically, so they are hard to control and leak information about attention and mood. Some of his early studies found that people who use more first-person singular pronouns tend to score higher on measures of depression.
Why do close friends rate personality accurately?
In a 2010 paper in the Journal of Personality and Social Psychology, psychologist Simine Vazire found that self-ratings and other-ratings each have strengths. People know themselves better on traits tied to internal experience, such as anxiety. Others judge more accurately on traits that show in visible behavior or carry social weight, such as intelligence and creativity. Friends see the person in many settings, which gives them a broad record of how the person acts.
More from Huberman Lab
