Learning how the mind makes meaning
When Jiaxuan Li describes her research, she often begins with a mistake.
Take the sentence “The storyteller could turn any incident into an amusing antidote.” Most would pause briefly before subconsciously correcting “antidote” to “anecdote.” Others might not notice the error at all. A few might register it, and then maybe assume the unusual wording was intentional.
For Li, those split-second responses offer a window into how the human mind works.
“People encounter mistakes or inconsistencies in language all the time,” she says. “Our brains are constantly adapting to make sense of them.”
Li, who earned her Ph.D. in language science from UC Irvine in March, has spent the last five years studying what happens when the brain encounters those moments of confusion. Her dissertation examined what are called “hallucinations” in language comprehension, or moments when people mentally fix errors in a sentence without even noticing.
“Our brains are so good at predicting what is plausible,” Li says. “Sometimes they give a good-enough answer rather than a perfect answer because this is more efficient.”
This, Li argues, is not a flaw but a feature of human cognition. Our brains rely on shortcuts, or what scientists call heuristics, to process language quickly and efficiently. She used EEG recordings, which measure electrical activity in the brain, along with computer modeling to study how the brain responds to language. Rather than simply tracking when people mentally correct errors, Li wanted to better understand the hidden cognitive processes that make those corrections possible.
The same questions that guided Li’s dissertation are now shaping her work in artificial intelligence. After graduating from UC Irvine, Li is bringing those ideas into industry with jobs at Amazon and Apple.
“Human brains process enormous amounts of information with limited resources,” she says. “We can create endless meaning from language, and somehow we do it naturally and almost effortlessly.”
Open questions
Li chose to attend UC Irvine because she saw a rare intellectual culture.
"The department really fosters an atmosphere of intellectual conversation,” she says. “Everyone is so interested in research and so willing to share their expertise and interests."
Her advisor, language science professor Richard Futrell, became central to her development as a researcher. Li describes him as both deeply rigorous and remarkably open-minded, someone who encouraged ambitious questions while helping students connect them to broader scientific debates.
“He gave me the freedom to pursue the topics that interested me most,” she says. “When you’re new to research, it can be hard to identify which questions are worth pursuing. He was very good at helping me shape those ideas and connecting them to broader conversations in the field.”
That flexibility became one of the defining features of Li’s graduate experience. The language science program allows students to shape coursework around their own research interests rather than forcing them into rigid requirements. Li says that freedom made it possible for her to move fluidly across disciplines and methodologies.
Just as important was the department's collaborative culture. Li worked closely with multiple faculty members whose expertise pushed her thinking in new directions. Language science assistant professor Xin Xie, for instance, introduced her to speech research after Li had focused primarily on written language. Li also describes her as a model of what sustained intellectual passion looks like in practice.
“The moment she started talking about research she came alive,” Li says.
Language science associate professor Greg Scontras helped her learn how to share research across different audiences, a skill that proved essential when she moved into industry. That range of mentorship, technical, communicative and personal, reflects something Li sees as distinctive about the program.
“One of the most valuable skills in social science is learning how to tackle complex, abstract problems,” Li says. “You have to connect big ideas about human behavior and thinking with real-world tools and evidence.”
Li’s move into tech did not represent a departure from her research so much as an extension of it. At Amazon, she worked as a language engineer helping develop training and evaluation datasets for Nova, the company's large language model. Her team focused on improving the model's performance in complex domains such as mathematical reasoning, logical reasoning, and medicine. She also felt a certain satisfaction about working at a larger scale. The model she was helping to build would eventually reach millions of users. But over time, she found herself drawn back toward the kind of deep, open-ended inquiry that had defined her years at UC Irvine.
That pull is what led her to Apple. At Apple, she will explore whether AI models make mistakes for the same reason human brains do. Her hypothesis is that they default to what is statistically probable rather than what is precisely correct. If that is true, then the models developed to study the human brain might also help researchers understand, and correct, the behavior of AI systems.
Carrying the research forward
Li’s path to Amazon and Apple began far from Silicon Valley. She grew up in Shanxi, a small industrial city in China known for coal mining and heavy industry. Her parents valued education and encouraged her to pursue opportunities wherever they led. She studied in Hong Kong as an undergraduate before earning a master’s degree at the University of Chicago.
“They were always saying, ‘You can always come back home, and we will support you,’” Li says. “It’s so nice to know that you have a support system while you are working on something difficult far away from home.”
That unconditional support, she says, made it easier to keep saying yes to the unknown, whether that meant a new city, a new discipline, or a new kind of intelligence to examine.
“I benefited enormously from my time at UCI because it gave me the confidence to take on new challenges,” she says. “As long as you’re asking meaningful questions and using the best tools available to answer them, you can feel confident that you’re doing good science.”
For Li, that confidence has a very specific destination. The storyteller is still making mistakes. So are the machines. For Li, that is not a problem to be solved so much as a career's worth of questions to be asked.
-Jill Kato for UC Irvine School of Social Sciences
-pictured: Jiaxuan Li at Big Bear Lake in 2022.