Quantifying with words: An investigation of the validity of narrative‐derived performance scores
研究利用文本挖掘从绩效评估的叙事评语中提取情感分数，发现这些分数能可靠预测未来绩效、离职、晋升和加薪，比传统评分提供更多信息。
Abstract Performance appraisal research has focused almost entirely on traditional numerical ratings despite narrative text comments regularly being collected within appraisals. This study investigated the theory and utility of leveraging narrative comments to better understand employee performance. Narrative sentiment scores were derived using text mining on a large sample of narrative comments. These scores were then applied to an independent set of 2 years of performance data. It was assumed that narrative comments would reflect true performance variance that overlaps with traditional ratings, but also that they would capture incremental variance due to increases in total information and a reduction in rater‐motivated biases in contexts in which narrative data were not explicitly linked to administrative outcomes. The derived narrative scores were reliable across years, converged with traditional numerical ratings and explained incremental variance in future performance outcomes (performance ratings, involuntary turnover, promotions, and pay increases). Collectively, this study highlights how narratives can enhance performance measurement and demonstrates how these data can be economically scored in applied settings.