International Laboratory Network search

Linguistics lab’s language cognition research could provide insights into AI text generation

Because thought precedes speaking and writing, most of us believe that the mind has full influence on the language we use. But can the language you speak — whether it is English, Mandarin Chinese or another dialect — influence your cognition and change the way you experience the world? If so, how is this evidenced in bilingual brains? How does it impact translations? And what does it mean for chatbots fueled by artificial intelligence that produce language but lack a human mind?


These are the types of questions that Charles Lin, an associate professor of Chinese linguistics in the Indiana University Hamilton Lugar School of Global and International Studies and adjunct associate professor in the Cognitive Science Program, aims to answer in his Language and Cognition Lab.


His lab is equipped to conduct behavioral experiments, measuring response time and eye tracking while reading, and neurological experiments using electroencephalogram. His research focuses on language-processing differences, mostly between English and Mandarin Chinese speakers.


For example, Mandarin Chinese is a tonal language, meaning that small changes in pitch can alter the meaning of a word, while pitch carries less fixed meaning in English. Lin’s group aims to understand how pitch interacts with other types of linguistic information and how language learners acquire tones and tone changes in Mandarin.


“Jacobs School of Music students should be happy to know that musicians are better learners of tones in tone languages,” Lin said.


Lin also conducts experiments specifically related to grammar and how it shapes a language user’s perceptions and decision-making. One such experiment centers on a grammatical device in Mandarin that is used to quantify nouns called “classifiers.” The device, which is similar to how we may say a “piece” of paper or a “strand” of hair in English, is used for every noun in Mandarin.


Lin and his lab wanted to explore how the use of this specific grammatical device influences a person’s perception of what is whole. To test this, the researchers presented items like a cake or watermelon in various stages of wholeness — some partially cut up, some completely cut up and some intact — and had subjects decide whether they were viewing one item. Native English speakers were more inclined to view the cut-up item as one, where native Mandarin speakers saw multiple pieces.


The results of this experiment are a perfect example of why grammar matters, according to Lin. He said grammar can be misinterpreted as being a set of arbitrary rules that don’t play a necessary role in communication and thinking.


“Grammar is fundamental to the nature of our memory systems,” Lin said. “It interacts with our human body to package things and manage our complex thoughts.”


Another focus of Lin’s research is how the fundamental difference between ordering phrases used in English and Mandarin may influence English-to-Chinese translations done by native English speakers. English speakers tend to place complex noun modifiers after the noun. For example, they are more likely to say “the cat that was small, black, scruffy and mean” as opposed to, “the small, black, scruffy, mean cat.” In Chinese, modifiers always come before the noun, which can tend to overload the native English speaker’s processing systems.


artificial intelligences are still mostly a mystery to humans. Using the same methods of working backward to understand the impact of language on the mind, researchers could uncover what imprint language will have on AI. Lin said his research findings may also provide insights to the predictability and reliability of translations done by AI.


Understanding how artificial intelligence is producing translations will be key to facilitating continued cross-cultural communication, Lin said. It will also play a role in preserving humanity in translations. As a leader in global language study, instruction and preservation as well as the creation of human-centered AI, Indiana University is uniquely positioned to do this work.


“The human factor and cross-linguistic differences are still a big challenge for translations,” Lin said. “We are still learning what the AI models are able to do to represent this kind of human variation and convey thoughts that are from a language that are different in grammatical nature. What will be lost in translations done by a machine?”


Charles Lin's lab is equipped to conduct behavioral experiments, measuring response time and eye tracking while reading. Photo by James Brosher, Indiana University