Ratings of insight following dream interpretation by ChatGPT

Abstract

There have been many claims that ChatGPT can provide insightful interpretations of dreams. This study tests whether ChatGPT, if given a participant’s report of a recent dream, a report of their recent waking life, and their associations to the dream contents, can achieve  outcome ratings by the dreamer comparable to those obtained after discussions of dreams in a lay group or with a therapist. The Gains from Dream Interpretation (GDI) questionnaire was used to assess participants’ evaluations of the output of ChatGPT regarding their dream. The mean GDI subscale scores were compared to means from previous studies of group and therapist dream interpretations. 30 participants (mean age = 30.72 (SD=13.39)) completed this pre-registered study. Mean insight-exploration subscale score was significantly lower than means for this variable in previous group and therapy studies. However, ChatGPT did not differ from dream discussion groups on GDI items for thoroughness of the dream exploration, the dreamer learning what the dream means, learning how past events influence present behaviour, and the dreamer learning about their waking life issues. Participants rated the ChatGPT dream interpretations as thorough, as very or moderately accurate, and with enabling personal insight. Scores on the insight variables had no significant correlations with word length of reports inputted to ChatGPT, nor with the individual difference measure Attitude Towards Dreams. As ChatGPT, other generative AI chatbots, and dedicated dream interpretation apps develop, the methods used in the current study can be used to assess the value of dream interpretations that they produce.