Beyond the Margin: Targeted Conservation and Household Water Demand

本文基于芬兰775户家庭的预注册随机田野实验,比较淋浴计时器、节水花洒及其配合实时反馈对家庭总用水和淋浴端用水的影响。结果显示,计时器未产生显著节水效果;节水花洒使家庭日用水量下降约5%,加入实时反馈后降幅扩大至约10%,且淋浴端节水大部分传导至家庭总需求,未被其他用水行为明显抵消。成本效益分析进一步表明,技术约束与即时反馈相结合的方案优于单纯提醒策略。

本文基于芬兰775户家庭的预注册随机田野实验,比较淋浴计时器、节水花洒及其配合实时反馈对家庭总用水和淋浴端用水的影响。结果显示,计时器未产生显著节水效果;节水花洒使家庭日用水量下降约5%,加入实时反馈后降幅扩大至约10%,且淋浴端节水大部分传导至家庭总需求,未被其他用水行为明显抵消。成本效益分析进一步表明,技术约束与即时反馈相结合的方案优于单纯提醒策略。
This work proposes a minimal, non-linear exponential random graph model (fit2SM) that can reproduce both the first and second moments of empirical degree distributions within a canonical framework . By replacing linear two-star constraints with fitness-induced components, the model avoids the degeneracy of the mean-field two-star approach and successfully matches the sample variance of degree distributions . Empirical validation on the eMID interbank network across daily to yearly aggregations shows that fit2SM outperforms the traditional UBCM and the density-corrected gravity model (dcGM) in reproducing variance, degree sequence, and spectral properties, particularly in sparser configurations . Moreover, when treated as a generative model, fit2SM supports robust early-warning signal analyses, aligning well with ground-truth indicators even when the true topology is not available .
This study develops a theoretical framework for pharmaceutical licensing that distinguishes incremental from novel innovations, linking information precision to market outcomes. Using a product-level dataset and Double Machine Learning (DML) with a causal DML-IV instrument (Phase III failures), the authors show that licensing raises predicted success probabilities overall, but the impact on returns varies by innovation regime: incremental projects exhibit return equalization (higher licensing success with lower post-licensing returns), while novel frontier projects often lack this offset, implying lemons-type frictions. The empirical results are decomposed across risk profiles, illustrating that competitive equilibrium holds for incremental projects but is weaker at the innovation frontier, where information frictions persist. The paper concludes with policy implications on transparency and contract design to better price frontier innovations
本研究在受控实验中发现,当将明确的非诊断线索嵌入叙事框架时,参与者的 belief 更新会系统性地回撤至最大不确定性0.5,削弱先前累积的诊断信息影响;在等效的抽象情境下及 Story No Info 条件中,该效应显著减弱甚至消失。通过蒙特卡洛模拟,研究进一步显示此信念回撤会延缓信息收敛、增加决策前需处理的信息量,具有经济和组织层面的重要含义

本研究围绕生成式AI在科学研究中的生产力提升与治理风险展开,基于四阶段研究过程的讨论梳理私人与社会回报的潜在冲突及其成因,提出以四项基本原则(披露、差异化、叙事、比例性)为核心的RRAI治理框架,力求在保持AI生产力增益的同时缓解信息不对称、知识共同体负外部性与研究能力枯竭等系统性风险,框架以EU AI Act、UNESCO与OECD等现有制度为锚点,面向不同阶段、不同身份的研究主体实现分层治理与资源配置优化 .