AI teammates can measure how creative you are through conversation.

Durable skills like collaboration, creativity, and critical thinking matter for work, but schools often ignore them because they are hard to measure. Good tests need to feel like natural real-world human conversations, but they also must stay controlled and reproducible. Researchers used large language models, or LLMs, to solve this problem. In their setup, a person talks with AI teammates. These AI partners mimic real human group work. Behind the scenes, an "Executive LLM" guides the chat to coax out clear proof of the person's skills. A separate AI evaluator then grades the conversation transcripts. The team tested this setup across several durable skills and tested an autorater on complex tasks done by real students for creativity. They found that the Executive LLM significantly increases the evidence drawn out during the talk. Finally, the automated AI scores largely agreed with marks from expert human judges.

Why it matters

Schools and employers know that skills like creativity and collaboration matter, but they rarely measure them because it is expensive and subjective. This work shows a way to assess them through a structured chat with AI, making it possible to track and teach these skills at scale. That could change how education and hiring evaluate what actually drives success.

Questions answered