文中指出，2010年代后的大规模再现检验推翻了许多高影响力但低可靠性的社会心理学效应，包括“自我损耗”“权力姿势”“面部反馈”与多种行为启动研究。关键问题集中在抽屉文件偏差、研究者偏差与部分数据造假，使早期显著效应在跨实验室复现实验中趋近于零或大幅缩小。相对之下，语义与知觉启动等较低层次的认知效应、以及结构性制度变革（如预注册、开放数据、多实验室合作）显示再现率显著高于早期个别实验阶段。

经元分析与多实验室项目筛选后，若干经典理论仍具稳定效应，包括认知失调、社会认同理论、刻板印象威胁、旁观者效应（但不包括其著名案例叙事）与心理安全感。这些理论的共同特征是效应量中等、机制明确、跨情境可重复，且在真实场景中具有可观察的行为后果。然而，部分效应量在田野环境中比实验室更弱，例如刻板印象威胁的平均效应在多次元分析中呈现高度异质性，显示其依赖情境、研究设计与抽屉文件偏差。

行为经济学的“助推”研究也经历了类似分化：自动加入式养老储蓄等结构性助推具有大效应且可复制，但依赖情绪或道德暗示的助推在后续高功率研究中多呈小效应或零效应。文中强调最佳证据来自大型再现、随机对照试验、跨情境验证与统计合成，并建议对单一研究的显著结论保持怀疑，将已发表效应视为真实效应量的上限。总体趋势显示社会心理学正借由危机走向更稳健的累积性科学。

Large-scale replication efforts in the 2010s invalidated many high-profile but low-reliability social psychology effects, including ego depletion, power posing, facial feedback, and various behavioral priming claims. Core problems involved file-drawer bias, researcher bias, and occasional data fraud, causing early “significant” findings to shrink toward zero in multi-lab replications. In contrast, lower-level cognitive effects such as semantic and perceptual priming, along with structural reforms like preregistration, open data, and multi-lab collaborations, show substantially higher reproducibility than early single-lab studies.

After filtering through meta-analyses and multi-lab projects, several classic theories retain stable empirical support, including cognitive dissonance, social identity theory, stereotype threat, the bystander effect (excluding its original narrative), and psychological safety. These theories share moderate effect sizes, identifiable mechanisms, cross-context replicability, and observable real-world behavioral consequences. However, some effect sizes weaken in field settings; for example, stereotype threat exhibits high heterogeneity across meta-analyses, reflecting sensitivity to context, study design, and file-drawer bias.

Behavioral economics–driven nudging shows similar divergence: structural nudges such as automatic enrollment in retirement savings display large, reliable effects, whereas emotion-based or moral-framing nudges often collapse to small or null results in high-powered follow-ups. The text emphasizes that the strongest evidence arises from large replications, randomized controlled trials, cross-context validation, and statistical synthesis, advising skepticism toward single-study claims and treating published effects as upper-bound estimates. Overall trends indicate that social psychology is evolving through its crisis into a more robust cumulative science.