Collective belief revision as a saturating function of accumulated knowledge: cross-platform evidence from social media
Year: 2026
Type: preprint
Abstract: Probability is usually treated as static given the knowledge that conditions it. This paper proposes a dynamical interpretation of collective belief — expressed belief as a saturating function of accumulated knowledge V with a revision ceiling Kₘ (working sigmoid form \(\:p={K}_{m}·\sigma\:\left(\alpha\:V+\beta\:\right)\)) — tested on 297 Twitter rumour threads (PHEME/RumourEval-2017) and 3,387 Sina Weibo repost cascades (CED). Pooled revision curves are saturating and monotone; they survive within-thread, cross-thread and block-permutation placebos (p ≤ 0.02), de-cumulation, partial-pooling and leave-one-event-out refits. The ceiling separates false from true rumours by ≈ 2-fold (1.7–2.4 under sensitivity analyses; paired event-cluster bootstrap P(reversal) = 0.012; with 7/5 usable events per side, event-level inference remains resolution-limited by design); non-rumour cascades are flat. Because comment streams under true claims are not truth-conducive, the contrast is read as a correctability measure for false claims only; on Weibo the ceiling is measurement-conditioned (Kₘ ≈ 0.50–0.92 across denominator corrections). The logistic family is not identified, so only family-independent findings are claimed: a ceiling, monotone saturation, and sequence-dependence.
Source: Research Square
Authors: Yuexin Lan, Bo Mao, Sijia Li, Qifeng Li, Shuangshi Zhang
Institutions: China People's Police University, Guangdong Police College
Language: English
Subfield: Sociology and Political Science
Field: Social Sciences
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7235434343
