Reducing Unconscious Bias When Conducting Performance Reviews Checklist Use the following checklist to identify and reduce bias when conducting written and verbal performance reviews. These items are based on the best evidence from research on patterns in performance evaluation. Does the evaluation stay focused on the requirements for and performance of the job? Does the evaluation include specific evidence or examples that support all evaluative claims? Are numerical ratings explained with written narratives that include specific examples? Does the evaluation include specific feedback about how the employee can improve? Does the evaluation overtly or subtly include stereotypical or gendered adjectives or descriptions (e.g., good communicator, works well with others, compassionate, collaborative, supportive), reflecting assumptions or unconscious biases about who best performs the job? If these items are important to the job, are they balanced with other strengths, reflecting a range of diverse capabilities beyond stereotypes? Does the evaluation avoid vague “grindstone” adjectives (e.g., hardworking, dutiful) and/or make sure these are balanced with adjectives that accurately reflect an employee’s ability or talent? Does the evaluation include “doubt raisers” or inadvertently “damn with faint praise” (e.g., “so and so does a really nice job, but...” or “dutifully and adequately completes tasks” or “it appears her home life is stable”)? Does the evaluation use title and surnames in ways that are consistent with other evaluations (e.g., the evaluation uses first name more often than in evaluations of other employees)? Does the evaluation give unnecessary attention to the personality of the applicant? Is the evaluation of similar length to evaluations of employees in similar positions (e.g., is it inadvertently shorter)? Does the evaluation reflect as much time and attention as do evaluations of other employees? (E.g., typos, grammar mistakes, and the appearance of being hastily written can subtly communicate negative messages to the reader.)