Our new study “Can ChatGPT Read Who You Are?” (Computers in Human Behavior)

Read our new study on the psychology of personality and AI in ChatGPT. The study was published in Computers in Human Behavior: Artificial Humans. You can read the study on Elsevier’s website: https://doi.org/10.1016/j.chbah.2024.100088

Generated by DALL-E


Abstract: The interplay between artificial intelligence (AI) and psychology, particularly in the area of personality assessment, is an important emerging area of research. Accurate estimation of personality traits is crucial not only for improving personalization in human-computer interaction, but also for a wide range of applications, from mental health to education.

This paper assesses the ability of a generic chatbot, ChatGPT, to efficiently infer personality traits from short texts. We present the results of a large-scale user study in which texts written in English by a representative sample population of 155 participants were analyzed.We compare ChatGPT’s estimates of personality traits with those of human raters and show that ChatGPT’s performance in inferring personality traits from text is competitive. We also uncover a “positive bias” in ChatGPT ratings across all personality dimensions and investigate the effect of prompt wording on the accuracy of the judgments.

Our work contributes to the understanding of AI’s capabilities in psychological assessment, highlighting both the potential and limitations of using large-scale language models for inferring personality. Our research also highlights the importance of responsible AI development with respect to ethical considerations such as privacy, consent, autonomy, and bias in AI applications.


Derner, E., Kučera, D., Oliver, N., & Zahálka, J. (2024). Can ChatGPT Read Who You Are? Computers in Human Behavior: Artificial Humans. https://doi.org/10.1016/j.chbah.2024.100088