A new study challenges decades of prior thinking about the biases in human languages and reveals that the use of safe language is more common than previously thought. In fact, safety is the deep code that underlies human language most of the time.
“Language is our great social technology, and it encodes in this hidden way that we are fundamentally social creatures,” Peter Dodds, a complex systems scientist at the University of Vermont, told Discover.
Dodds studies how human language is organized around an axis of meaning. This involves coding the underlying meaning of words and phrases.
This kind of science has been around for a long time. Researchers have long followed the VAD framework created by scientists like Charles Osgood — the term stood for valence, arousal, and dominance. In this concept, the meaning of words could be defined along spectrums of positive versus negative, excited versus calm, and controlling versus submissive. This understanding is important for psychology, linguistics, and even the development of artificial intelligence. These concepts were taken to define the very essence of meaning.
But Dodds and his colleagues, in a study published recently in Science Advances, found that this approach was missing something. This included a fundamental axis of human language that can be boiled down to: safety versus danger.
The team analyzed more than 20,000 words and texts — everything from Wikipedia pages to Google Books, The New York Times, and Arthur Conan Doyle’s Sherlock Holmes stories — using artificial intelligence to test the underlying meaning of human language. The first thing they found was that the traditional VAD framework didn’t seem to capture what was going on.
“The compass was off,” Dodds said.
New Framework for Human Language
From the discovery, the team developed what they call the goodness-power-aggression-danger structure (GPADS).
Rather than the previous spectra, this structure is based on power (weak versus powerful), danger (safe versus dangerous), and structure (ordered versus chaotic). They found that by using these spectra, they could explain more than 90 percent of the variance in meaning, whereas the older VAD framework could explain only about 72 percent.
Among all the books, news, social media, and other communications that Dodds’ team examined, they found that, within this framework, language leaned heavily towards words associated with safety rather than danger.
Read More: Modern Humans and Neanderthals May Share Genetic Regions Linked to Complex Language
Is GPADS Better Than VAD?
Marc Brysbaert, an experimental psychologist at the University of Ghent in Belgium who wasn’t involved in Dodds’ study, told Discover, “It is very well possible that this structure will be better than the valence-arousal-dominance structure,” in part because the dichotomies are easier for participants in studies to understand than concepts like arousal in the traditional VAD framework.
But a large-scale evaluation of these concepts with participants would be needed to confirm their efficacy.
“If not, there is a good chance that researchers will stick with the VAD structure, since a great deal of norms are available for that structure,” Brysbaert said.
He also added that these simplifications may miss some important facets of underlying meaning in language.
“Based on text analysis, other researchers distinguish between 40 or more emotions, which can be useful in other contexts,” Brysbaert said.
Safety Bias and the Pollyanna Principle
Dodds said that the safety bias aligns with the Pollyanna principle, a theory named after the Eleanor H. Porter novel Pollyanna, whose titular character attempts to be happy in every situation. The Pollyanna Principle holds that people tend to remember positive or pleasant things more readily than unpleasant ones. Dodds believes that language reflects this Pollyanna principle, showing a positivity bias.
There are some differences, Dodds noted. The team looked at several languages and found that, while it still skews towards safety, the Ukrainian language has recently leaned more towards danger than other languages.
“It’s a nation or people that has been under siege,” he said. “You can see population-level shifts, and they don’t come back.”
But even analysis of social media platforms like X, formerly known as Twitter, that are often criticized for focusing too much on negativity, skews mostly towards the safety bias.
For him, this makes sense for a social species like humans. Humans rely heavily on each other for survival, putting safety, or safe words at least, at a premium. As a result, overall, language skews positively.
“Terrible things happen at all scales, and this will go on forever,” Dodds said. “But the balance of life is towards positive survival, building, constructing, and helping.”
Read More: What Is the Origin Story of Language? Our Ancient Ancestors May Have Developed Communication Rapidly
Article Sources
Our writers at Discovermagazine.com use peer-reviewed studies and high-quality sources for our articles, and our editors review for scientific accuracy and editorial standards. Review the sources used below for this article:
