Education

UF engineering professor to present AI risk-attitude research at major conference

University of Florida professor Eric Jing Du is researching whether large language models exhibit consistent, built-in attitudes toward risk — a dimension largely absent from current AI safety testing. His team tested models on repeated medical triage scenarios with erased memory and found the models produced remarkably consistent assessments, suggesting an internal mechanism beyond simple probability. The work will be presented at the NeurIPS conference in December 2026 and aims to make risk-tolerance benchmarking a standard step before AI deployment.

Sources: UF News