Play It Again, For the Next Time: Iteratively Reinterpreting Dreams with Generative AI
Yifan Zhao, Bowen Liu, Zhi-Peng Jia, Yuxuan Bai, Ray Lc
Adjunct Proceedings of the 2025 International Conference on Human-Engaged Computing November 21, 2025 DOI: 10.1145/3775021.3787083 (opens in new tab)
Study at a glance
AI-extracted from the abstract| Characteristics | Qualitative study Peer reviewed |
|---|---|
| Sample size | 4 |
| Population | One dream recorder and three dream interpreters |
| Duration | Six rounds and two months of iterative feedback |
| Key findings | The authors demonstrate a workflow in which one person's eight recorded dreams were visually reinterpreted by three interpreters using generative AI across six iterative feedback rounds over two months. They present the reinterpretations as storyboards of AI-created images and as a video game scene from the "bottled memories" dream, arguing that narrative and interactive components together create an immersive backdrop for revisiting dream interpretation. |
Abstract
Our dreams are so fleeting that every time we try to recall them, we reinterpret them anew. To explore how human psyche can iterative reconstruct dreams to re-imagine dream content in novel contexts, we created a workflow whereby one dream recorder wrote down eight distinct dreams, and three dream interpreters use these narrative fragments to visually recreate these subconscious experiences using Generative AI (GenAI) in six rounds and two months of iterative feedback. The themes of the narratives include spying in a library, escaping from bedrooms, conquering the castle, underwater bubbles, cruising on a cargo ship, and bottled memories. We show the reinterpretations in two ways: (1) Storyboards of GenAI-created images; (2) A video game showing a scene from the "bottled memories" dream where interactions represent what the dreamer was doing in the dream. The narrative and interactive components together create an immersive backdrop to revisit the dream interpretation process.