Is AI Above the Curve in Evaluating Employee Performance?
Traditional performance management tools “need improvement.” That’s the message from managers asked to rate how their companies conduct performance reviews. Companies like Accenture have publicly tossed aside the annual employee performance review in favor of a system of “ongoing feedback.” Yet, no matter how frequent the review or feedback, human error and bias can compromise the integrity of the employee review process. When these errors are present, a performance review may hurt employee morale and job performance, rather than function as a helpful evaluation tool.
Can software and tech tools that use artificial intelligence (AI) improve the performance management process? The research to date suggests that AI may automate some aspects of the performance review cycle, but its ability to provide useful feedback is still limited.
AI Can Streamline Performance Evaluations
Quick and Continuous Feedback
Traditional performance reviews usually occur once a year and can become a low priority in a high-volume workplace. As a result, traditional performance reviews may provide limited opportunity to assess a worker's performance and organizational contributions.
Using AI Can Counteract Certain Reviewer Biases
Performance evaluations are adversely impacted when managers employ logical fallacies based on emotional reasoning that ignores objective facts. A similar result flows from the influence of “recency bias,” in which managers weigh what the employee appears to have done in the last weeks or months, rather than assessing the entire evaluation period. “Contrast bias,” when a manager incorrectly compares an employee to his or her peers, can also adversely impact the performance review process. For example, a manager may overlook employee potential by giving undue weight to a recent mishap in a performance review.
AI-driven technology that leverages data can help reduce certain biases that can impact the efficacy of performance reviews. For example, giving managers tools to immediately identify changes in performance, in real time, can eliminate biases that may otherwise plague the performance review process. In addition, real-time feedback supports managers by giving them the tools to immediately identify, evaluate, and correct, operational inefficiencies.
Potential AI Pitfalls
AI is often praised for eliminating conscious and unconscious human bias. Yet, research shows that AI engenders cognitive bias, because the technology largely relies on parameters that reflect the conscious and unconscious biases of the humans who program it. While AI technologies are vigorously tested for biased results prior to implementation, the beauty of AI technology is that it learns and grows based on new data received as time goes on. Since new data is not tested for bias, it can yield biased results. If AI is programmed to evaluate biased metrics, then performance reviews based on that criteria will lead to biased results, which exacerbate issues already inherent in performance reviews.
While AI can help to alleviate workplace bias by providing data that allows management to identify and place talented workers in ways that promote operational efficiency and immediately recognize improvement in employee performance, employers should also be wary of how management reads data that AI collects. The risks of “contrast bias” are still relevant when reviewing multiple employees who perform similar tasks, according to the same criteria. Organizations need to ensure that harmful generalizations do not flow from reported indicators of employee performance.
Impact on Employee Morale
Negative performance evaluations can cause employees to become less invested in project outcomes and organizational objectives. AI provides objective indicators to evaluate employee performance but can fail to provide a holistic analysis of the employee by removing empathy from the equation. Also, the platform fails to consider that real-time and automated criticism could impact various workforce contingencies differently. To preserve employee morale despite these shortcomings, organizations should not blindly rely on AI technologies when conducting performance reviews. Instead, employers must recognize that people are needed to establish the evaluation parameters and cadence, evaluate the data, and communicate results.
Spotting Liability Early
In the #MeToo Era, allegations of sexual harassment and discrimination in the workplace are currently at the forefront when assessing legal risks. As we previously reported, some third-party AIs meet these concerns by providing anonymous reporting platforms where employees can report instances of sexual misconduct or harassment without fear of retaliation. Yet, overall, AI’s automated data-focused approach fails to meet employer needs in this context. AI does not always provide information that allows organizations to consider necessary exceptions or explanations for poor performance that are personal and cannot be objectively measured, nor do all platforms provide a reporting mechanism necessary to resolve workplace problems. As a result, it is important that organizations do not rely exclusively on AI-provided information when assessing performance. Human interaction is still necessary to ascertain the morale and tone of a team, which allows employers to accurately assess legal risks.
The benefits of AI are clear - by providing timely feedback, employers can recognize talent and growth opportunities early, thereby allowing for efficient management, overall lower costs, and dual-success for employees and the organization. Yet, HR-savvy humans are still necessary to ensure that performance evaluations facilitated by AI produce accurate data, identify legal risk, and that performance results are communicated in a way that does not impact employee morale.
EDITOR’S NOTE: An earlier version of this post stated that 15Five uses data-collection platforms powered by AI to help employees conduct weekly self-evaluations; 15Five has not implemented AI in its products.