UNH Today

Imagine putting in the work at your job to stand out, only to learn that artificial intelligence algorithms are going to play a large role in your performance evaluation. How would you react?  

If you’re concerned that your supervisor’s biased view of you could negatively impact your review, you might welcome AI’s involvement, according to research from the University of New Hampshire.  

In a paper published in the Journal of Information Systems, Yunshil Cha, assistant professor of accounting and finance at the UNH Peter T. Paul College of Business and Economics, found that when employees anticipated bias or unfair treatment from a human supervisor — such as favoritism or personal disapproval — they considered AI evaluations more trustworthy.  

Cha says early research into the use of AI in performance evaluations revealed some skepticism, with employees feeling that AI evaluations were too focused on numbers and lacking an understanding of teamwork or individual contributions. However, knowing that AI’s objectivity has been valued in other areas, Cha wanted to apply it in the context of performance evaluations.   

"Previous research outside of performance reviews, like in hospital triage decisions or hiring processes, showed that people tend to prefer AI when they believe human decisions could be influenced by bias — such as racial bias in healthcare or gender bias in hiring,” Cha says. “This preference stems from the perceived objectivity of AI’s decisions.”  

Simulating Biased Workplace Situations  

Both hypothetical human and AI evaluations in the study used the same set of objective criteria, such as the number of hours worked and tasks completed, and subjective factors, including teamwork, leadership skills and potential for growth. However, the AI evaluations relied on computational calculations without human involvement, while human evaluations had the discretion to factor in personal observations. 

In her study, Cha ran an experiment by recruiting 120 participants who took on the role of employees in a hypothetical company. Each participant took part in an online survey where they were presented with scenarios where their performance would be evaluated by either an AI or a human supervisor. 

Some scenarios involved positive escalation bias, where participants were told their supervisor had previously recommended them for promotion over their teammates. In the negative escalation bias condition, participants learned their supervisor had recommended promoting their teammates instead, though they were assigned to the team regardless. 

The results showed that when participants anticipated potential bias from a human supervisor, they rated AI evaluations as more trustworthy (4.66 out of 7) than human (2.71 out of 7). However, when employees expected a positive bias, they leaned slightly toward preferring human evaluations (5.08 out of 7) over AI (4.43 out of 7). 

"When people feel there’s potential for bias in their evaluations, they value objectivity. That’s why they tend to trust AI-driven evaluations more,” Cha says. “They see AI as a fairer option because AI decisions are less influenced by personal values and perspectives.”   

The study also found that employees were more likely to consider leaving their jobs after receiving a biased (negative) review from a human rather than AI, suggesting that AI’s perceived fairness could help reduce turnover in certain environments.  

AI’s Role in the Workplace  

While there may always be situations where human judgment is needed, Cha says many companies are increasingly leveraging AI in performance management. For example, Microsoft’s 'Productivity Score' tracks employee behaviors such as document sharing and participation in group chats to gauge productivity. Similarly, PwC uses AI to predict turnover risks and identify pay inequities. 

“AI can analyze vast amounts of data to predict what an employee might need, even suggesting training programs, recommending salary increases, or even identifying employees who might be at risk of leaving,” Cha says. “And as this research shows, it’s viewed as more trustworthy in situations where subjective evaluation bias may be a concern.”  

Cha says potential positives of incorporating AI into performance management include encouraging a fairer workplace culture, supporting diversity and inclusion, improving assessment of the hybrid and remote workforce, and reducing employee turnover. However, she also cautioned that additional research is needed that explores combining AI with human oversight for balanced evaluations, refining AI models to reduce unintended biases and AI’s impact on employee satisfaction and retention.