Abstract: Curiosity drives us to explore vast amounts of information, while compression distills this abundance into memory structures that guide action. We examined three styles of curiosity—the wandering “busybody,” the focused “hunter,” and the creative “dancer”—by studying behavior in both lab settings and a large-scale naturalistic dataset of 482,760 Wikipedia readers across 14 languages and 50 countries. By modeling readers as biased random walkers atop Wikipedia’s knowledge network, we show how curiosity unfolds as an extended and open-ended search guided by simple rules on novelty-seeking and information foraging. At the same time, the brain simplifies information through lossy compression, discarding redundancies to capture key patterns. This balance between preserving detail and efficiently condensing knowledge follows a fundamental principle of efficient coding: maximizing useful information while minimizing resource costs. Analyzing brain networks in 1,041 individuals, we find that different balances of information transmission and compression relate to cognitive performance, including memory. Compression underlies memory’s reconstructive nature—rather than storing exact records, the brain distills the present with prior knowledge. Efficient coding explains why some memories blur together, making it harder to detect novelty. Together, curiosity and compression shape what we seek and retain, transforming vast complexity into structured knowledge through simple, local principles.
Speaker
Dale ZhouHewitt Postdoctoral Research Fellow & Cunningham Scholar University of California, Irvine