A common problem in performance appraisals is that evaluators avoid using the lowerend of the evaluation scale, failing to differentiate between employees and being too optimistic for underperformers. We argue that algorithm-assisted evaluations can address this problem by attenuating the psychological costs of giving low ratings. Results of an incentivized experiment support this theory. Specifically, evaluators who see algorithmic assessments of employee performance before providing their own rating give low ratings more frequently and provide harsher overall ratings than evaluators without access to an algorithm. These effects occur with both bonus and penalty contracts. More importantly, evaluators are similarly influenced by lower- and higher-quality algorithms, despite evidence showing they recognize quality. These findings suggest that algorithms encourage evaluators to use more ratings across the full evaluation scale, attenuating upward biases for low-performing employees. However, low-quality algorithms can distort evaluations because evaluators uncritically follow inaccurate ratings as a tool to justify low ratings. Thus, organizations should carefully consider algorithm Quality because evaluators may not proactively adjust for inaccuracies