New AI Benchmark Tests Social Norm Compliance in Real-World Tasks
Researchers created a benchmark to test if AI planners can follow hidden social norms, not just explicit goals. This helps AI act more appropriately in everyday environments, like knowing not to interrupt people.
Researchers introduced NormAct, a new benchmark to test AI planners' ability to follow hidden social norms. NormAct evaluates how well multimodal large language models (MLLMs) — AI systems that process text, images, and other data — can perform tasks while respecting implicit social rules. For example, an AI might know how to set a table, but NormAct checks if it also knows not to knock over a glass while doing so.
This matters because AI is increasingly used in real-world settings, like robots assisting in homes or offices. If an AI follows instructions but ignores social norms, it could cause awkward or even harmful situations. NormAct helps ensure AI behaves appropriately, making it more useful and safer for everyday interactions.
To see how AI planners perform, you can explore the NormAct benchmark on arXiv. While you can't directly test it yourself, you can read about the research and its implications for future AI development. Check out the full paper for more details.