Performance Evaluation Systems of the Future: Evaluating and Predicting Employee Performance using Algorithms

Abstract

Algorithms are being increasingly used by organisations to support managers’ performance evaluation judgments and improve the quality of feedback provided to employees. However, there is limited understanding of how managers and employees respond to the use of algorithms for performance evaluation and feedback purposes. This dissertation comprises two studies, which examine how managers use algorithms to evaluate employees’ performance (Study One) and how employees respond to performance information generated by algorithms (Study Two). Study One investigates the extent to which managers’ willingness to use an algorithm-advised rating to evaluate subordinate performance is influenced by the valence of the advised rating and managers’ decision rights to adjust the algorithm. The study finds that managers are less willing to use the algorithm-advised rating to evaluate subordinate performance when the rating indicates below-average rather than above-average performance of the subordinate. When the rating indicates subordinate performance is below-average, allowing managers to adjust how the algorithm computes the advised rating increases their willingness to use it, compared with allowing them to adjust the final advised rating. Study Two examines whether and how providing employees with the forecasts of their individual performance (i.e., forecast performance) improves their acquisition of skills that are critical for future work roles (referred to as “distal skills”). The study also investigates the extent to which this effect depends on the forecast source and its past accuracy. The study finds that the provision of employees’ forecast performance increases their willingness to acquire distal skills. This benefit, however, decreases if the forecast source is an artificial intelligence (AI)-based performance evaluation system (PES), rather than a human advisor. Further, the negative effect of having an AI-based PES as the forecast source is amplified when the forecast source displayed a higher accuracy in predicting employees’ performance in the past. This dissertation extends the extant accounting literature investigating the use of algorithms for performance evaluation. Additionally, it offers organisational advice on how to increase managers’ willingness to use algorithmic advice when evaluating employees’ performance. Further, it provides practical suggestions on how to implement algorithms to provide performance feedback to facilitate employees’ skill-acquisition decisions.