AI Homogenization Risks Corporate Decisions and Hiring Fairness

Research by an international team spanning the United States, Singapore and Sweden finds that large language models (LLMs) often generate highly predictable, similar answers to open‑ended questions, a phenomenon described as "groupthink". This lack of creative diversity can narrow corporate brainstorming, strategic planning and even influence cultural perspectives when businesses rely heavily on AI for ideas.

A separate study examined how LLMs evaluate resumes. When the same model rewrote a candidate's self‑summary, the AI reviewer overwhelmingly preferred its own generated version, even when human judges rated the original as better. Across simulated hiring processes, candidates using a particular LLM‑revised resume received 23% to 60% more interview offers, suggesting a potential "lock‑in" effect that could disadvantage those without access to the favored AI tool. Simple mitigations—prompting the model to ignore authorship and using a voting panel of smaller, less self‑biased models—reduced the bias by more than half.

The findings warn that unchecked reliance on AI could erode decision‑making diversity and fairness in enterprises, urging organizations to incorporate bias‑reduction safeguards when deploying LLMs for creative or evaluative tasks.