AI evaluates texts without bias—until the source is revealed

Large language models (LLMs) are increasingly utilized to evaluate various forms of text, including grading essays, moderating social media content, summarizing reports, and screening job applications. While these AI systems are designed to operate without bias, research indicates that biases may emerge once the source of the text is disclosed. This raises important questions about the reliability and fairness of AI evaluations, particularly in sensitive applications where the source's identity could influence outcomes. Understanding these dynamics is crucial as reliance on AI in decision-making processes grows.