The increasing importance of consumer ratings raises the question of whether people adjust for potentially fake or biased extreme opinions when judging products. Two studies tested treatments that trimmed the extremes of rating distributions. Neither removing extreme ratings while preserving the mean, nor flagging suspicious extreme ratings, nor priming individuals about review manipulation significantly affect judged product quality on average. However, judgments for specific distributions may be made less extreme by flagging or trimming. On average, it is difficult to override usage of the mean rating as the strongest proxy for product quality. When a weighted-mean model is fitted, the estimated weighting profile is hump-shaped and asymmetric. Consumers appear to discount 5-star ratings but are particularly susceptible to being misled by disingenuous 1-star ratings. The weights suggest that there is a binary bias with an inflection point at 2-stars for product ratings, meaning that any rating above this broadly sends an equally strong positive signal of quality. Further theoretical work is required to understand how people form weights for ratings, and applied work should continue to search for decision aids that could help consumers to better adjust for review bias. (PsycInfo Database Record (c) 2023 APA, all rights reserved).

中文翻译：

---

加权评级：人们是否会针对极端评论中的偏见进行调整？

消费者评级的重要性日益增加，引发了一个问题：人们在判断产品时是否会根据潜在的虚假或有偏见的极端观点进行调整。两项研究测试了缩小评级分布极端值的治疗方法。无论是在保留平均值的同时删除极端评级，还是标记可疑的极端评级，也不是引导个人进行评论操纵，都会显着影响平均判断的产品质量。然而，通过标记或修剪，可以使对特定分布的判断不那么极端。平均而言，很难超越平均评级作为产品质量最强指标的使用。当拟合加权平均模型时，估计的权重分布是驼峰形且不对称的。消费者似乎对 5 星级评级打折扣，但特别容易被不诚实的 1 星级评级误导。权重表明，产品评级存在二元偏差，拐点为 2 星，这意味着任何高于此的评级都会广泛发出同样强烈的积极质量信号。需要进一步的理论工作来了解人们如何形成评级权重，应用工作应该继续寻找可以帮助消费者更好地调整评论偏见的决策辅助工具。（PsycInfo 数据库记录 (c) 2023 APA，保留所有权利）。