The growing reliance on prediction markets as epistemic infrastructures has positioned platforms like Polymarket as providers of objective, real-time probabilistic truth, yet the signals they produce often obscure uncertainty, strategic manipulation, and capital asymmetries, encouraging misplaced epistemic trust. This paper presents a qualitative sociotechnical audit of Polymarket (N = 27), combining digital ethnography, interpretive walkthroughs, and semi-structured interviews to examine how probabilistic authority is produced and contested. We introduce the concept of Prediction Laundering, drawing on MacFarlanes framework of knowledge transmission, to describe how subjective, high-uncertainty bets, strategic hedges, and capital-heavy whale activity are stripped of their original noise through algorithmic aggregation. We trace a four-stage laundering lifecycle: Structural Sanitization, where a centralized ontology scripts the bet-able future; Probabilistic Flattening, which collapses heterogeneous motives into a single signal; Architectural Masking, which conceals capital-driven influence behind apparent consensus; and Epistemic Hardening, which erases governance disputes to produce an objective historical fact. We show that this process induces epistemic vertigo and accountability gaps by offloading truth-resolution to off-platform communities such as Discord. Challenging narratives of frictionless collective intelligence, we demonstrate Epistemic Stratification, in which technical elites audit underlying mechanisms while the broader public consumes a sanitized, capital-weighted signal, and we conclude by advocating Friction-Positive Design that surfaces the social and financial frictions inherent in synthetic truth production.

翻译：随着预测市场作为认知基础设施的日益依赖，Polymarket 等平台被定位为提供客观、实时概率性真相的机构，然而其产生的信号往往掩盖了不确定性、策略性操纵与资本不对称，助长了不当的认知信任。本文对 Polymarket（N = 27）进行了一项定性的社会技术审计，结合数字民族志、诠释性走查与半结构化访谈，以考察概率性权威如何被构建与争夺。我们借鉴 MacFarlane 的知识传递框架，提出“预测市场洗钱”概念，用以描述主观性高、不确定性强的投注、策略性对冲以及资本密集的“鲸鱼”活动如何通过算法聚合剥离其原始噪声。我们追溯了一个四阶段的洗钱生命周期：结构净化——中心化本体论框定可投注的未来；概率扁平化——将异质性动机压缩为单一信号；架构掩蔽——将资本驱动的干预隐藏在表面共识之后；以及认知固化——通过消除治理争议以生成客观的历史事实。研究表明，这一过程通过将真相判定责任转嫁给 Discord 等平台外社群，引发了认知眩晕与问责缺失。通过挑战关于无摩擦集体智能的叙事，我们揭示了认知分层现象：技术精英审计底层机制，而广大公众消费的则是经过净化、资本加权的信号。最后，我们倡导采用摩擦积极设计，以凸显合成真相生产过程中固有的社会与金融摩擦。