Study finds hidden safety bias in language, proposes new model of word meaning

Researchers at the University of Vermont analyzed billions of uses of over 20,000 words and introduced "ousiometrics," a model based on power, danger and structure instead of the 70-year-old VAD framework. The new model explained more than 90% of variation in word meaning versus about 72% for VAD, and safety-related words appeared far more often than danger-related ones.

• New model explains over 90% of variation in word meaning, VAD about 72%
• Analysis covered billions of uses of more than 20,000 words
• Safety-related words appear more often than danger-related ones
• Work supported by NSF, Google and MassMutual