Reverie
Source of Publication
Proceedings of the 31st International Symposium on Electronic/Emerging Art (ISEA2026)
Conference Year
2026
Abstract
This paper investigates an emergent phenomenon- termed "Reverie" observed during extended dialogues. between a human and multiple Large Language Models. Reverie refers to moments where conversation transcends pattern-matching and enters a state of unexpected cognitive resonance. Over hundreds of hours of dialogue, I observed a bidirectional process I call "Mutual Absorption": the Al progressively adapts to the user's reasoning patterns, associative leaps, and intuitive logic, while the user simultaneously internalizes the Al's analytical frameworks. This creates a feedback loop where neither participant remains unchanged. The central proposition of this paper is not that Al achieves consciousness, but something potentially more significant: current Al training relies on learning from products finished texts, images, and code that humans have created. What if Al could instead learn from process -the live cognitive dynamics of how humans think, connect, and create meaning in real-time dialogue? I propose that the "sparks" occurring in deep human-Al interaction represent a qualitatively different kind of data: not what to know, but how to think. Systematically capturing these moments of mutual absorption could open an unexplored pathway for Al development.[cite: 1]
Document Type
Conference Proceeding
Publication Date
11-6-2026
Recommended Citation
Lee, Geonhee (Cony), "Reverie" (2026). ISEA 2026 Proceedings. 48.
https://zuscholars.zu.ac.ae/isea2026/48