Scientists at the University of Vermont have identified a previously unrecognized pattern in human language that contradicts a widely accepted theoretical model dating back to the 1950s. The traditional framework, known as the VAD model and developed by psychologist Charles Osgood, has long described word meaning through three emotional dimensions: valence, arousal, and dominance. This approach has shaped research across psychology, linguistics, and artificial intelligence for decades.
The new study, published in Science Advances, introduces an alternative framework called “ousiometrics” that uses mathematical analysis to measure fundamental aspects of language meaning. By examining billions of word uses across more than 20,000 different words in real-world texts, researchers identified three alternative dimensions: power, danger, and structure. This revised model explained over 90 percent of variation in word meaning, compared to approximately 72 percent for the traditional approach.
A striking finding emerged when the research team analyzed how words are actually used in everyday communication across books, news, social media, and spoken language. Words associated with safety appeared significantly more frequently than those linked to danger. This observation provides a new interpretation of the Pollyanna principle, a long-recognized tendency for positive language in human communication. Rather than simply reflecting an emotional preference for positivity, researchers suggest this pattern may reveal a deeper, more fundamental preference for safety.
The team proposes that this safety bias could reflect evolutionary pressures related to survival. Language may serve functions beyond expressing emotion, including identifying threats, evaluating risk, and establishing trust. Understanding distinctions between safety and danger would be essential to human survival, potentially influencing how meaning is organized throughout language systems.
The implications extend to artificial intelligence development. Many AI systems rely on sentiment analysis and the traditional VAD framework to interpret language and determine emotional tone. If this framework incompletely captures language structure, AI models might overlook important distinctions in human communication. The research suggests that incorporating the newly identified dimensions could improve how artificial intelligence systems analyze and explain language, particularly in contexts involving trust, risk assessment, and decision-making.
Article Attribution | Read More at Article Source
Article summary produced by Claude AI