The globalization of digital workspaces has created a demand for real-time translation tools that go beyond linguistic syntax to interpret cultural subtext. While modern LLMs excel at literal translation, they frequently fail at pragmatics—specifically, the ability to discern when to employ high-context indirectness (common in East Asian business cultures) versus low-context directness (prevalent in North American and Northern European cultures).
We recruited 240 participants (120 from the US, 120 from China/South Korea/Japan). Participants were paired cross-culturally and tasked with negotiating a supply chain contract. One participant in each pair used an interface powered by a standard LLM, while the other used the CAF-enhanced model. june liu and zia
This paper presents evidence that dynamic contextual anchoring can mitigate the risks of cultural hallucinations in AI systems. As AI becomes a mediator of global interaction, it must move beyond translating words to translating meaning . The globalization of digital workspaces has created a
June Liu's career, spanning over seven years, has been sustained by these tactical partnerships that introduce her to new sub-communities. As AI becomes a mediator of global interaction,