Study finds major AI models most often default to Japan in open-ended cultural prompts
Aiden_Tech_Ai · x · 2026-07-24
A Cardiff University and University of the Basque Country study tested 31,680 culturally grounded prompts across 24 languages on frontier models including ChatGPT, Claude, and Gemini.
Key finding
- The usual “Western-centric AI” assumption gets complicated: when asked open-ended cultural questions, Japan was the single most referenced country across most tested models.
- Prompts about festivals, traditional dances, or everyday customs often led models to reach for Japan even when no country was named.
- The paper suggests the bias pattern is not simply “Silicon Valley / America dominates everything,” but something more nuanced.
More from Research
- Hi3D V1.0 open-sourced with 2M training assets and 90% slicing success rate — kimmonismus · 2026-10-26
- Ortet evaluates 34 open-weight models on 50 health tasks, finds general leaderboards insufficient — elmanmansimov · 2026-10-26
- HackBench injects reward hacks into agent trajectories to stress-test monitors — DanielKhashabi · 2026-10-26
- Fern's "context is intelligence" thesis: prediction keeps improving out to 10.2M tokens — bneyshabur · 2026-10-26
- Researchers identify "Forking": train loss falls while validation loss turns up — ZimingLiu11 · 2026-10-26
- Google DeepMind pilots world's first double-blind evaluations for frontier models — scychan_brains · 2026-10-26