Words are windows into the brain. The words that we choose - and how we say them - speak volumes about our personalities and even our mental health, said WashU psychologist Josh Oltmanns. "Our thoughts, feelings and behaviors are reflected in language," he said.
Instead of subjecting people to endless batteries of tests, psychologists could gain valuable insights from samples of language. But they might need high-tech help to find the right signals in all of the chatter.
Artificial intelligence (AI) tools trained to detect tell-tale signs in speech could revolutionize psychological assessment, he said.
"Psychologists are people, and people are fallible, so even a good clinician might not always pick up on important verbal cues," said Oltmanns, an assistant professor of psychological and brain sciences in Arts & Sciences at Washington University in St. Louis. "But a properly trained computer model will catch those cues."
In theory, a psychologist could ask a client to describe their life and concerns, a standard part of an initial assessment. In addition to using their own clinical expertise, the psychologist could feed that conversation into a program designed to detect personality traits and signs of mental health concerns.
"The computer program could help validate their observations or warn them about something they might have missed," Oltmanns said.
Oltmanns is working with his collaborators, who are also WashU PhD students, Tu Do, Tong Li, and Tongyao Ran, to develop AI tools to help psychologists uncover these hidden cues in language. He recently described the potential of such tools in the journal Advances in Methods and Practices in Psychological Science. Mehak Gupta, of Southern Methodist University, and Jocelyn Brickman, of Xavier University, were co-authors.
Language can convey psychology in many ways. Word choice matters, whether in a deep conversation or a casual post on X or Facebook. Early in his career, Oltmanns studied how word choices in social media posts broadly reflected a person's Big Five personality traits: openness to experience, neuroticism, agreeableness, conscientiousness and extraversion.
But the way a person says words matters, too. "You can tell a lot by how a person speaks," Oltmanns said. "Slowed speech can be a symptom of depression, while overly rapid speech is associated with anxiety."
Speed is just one measure. Spoken words also vary in loudness, tone and pitch. "Speech samples have hundreds of different acoustic parameters that could be meaningful," he said.
With so much potential information buried in each conversation, psychologists have long wanted the help of computers to analyze speech. More than 20 years ago, researchers developed Linguistic Inquiry and Word Count, a software program that could score people on various psychological aspects based on written text. Those tools have been refined over time, but the advent of AI opens a new world of possibilities, Oltmanns said.
"AI programs could be far faster, more thorough and more accurate than previous computer models," he said.
Still, Oltmanns cautioned that AI also has risks. "It's often trained on information on the internet, which means it can be biased," he said. If those biases aren't addressed, it's possible that certain cultural differences in speech patterns could be inaccurately labeled as signs of mental health problems.
To avoid such problems with bias, AI models should be trained on diverse patient populations. To that end, Oltmanns is studying the hundreds of hours of interviews collected over the years through the SPAN Study, an ongoing investigation of more than 1,600 St. Louis adults who represent the city's diversity.
"We're particularly interested in looking at speech patterns in white and Black participants to ensure that the AI models treat each group fairly," he said.
Oltmanns sees several other important questions moving forward. It's not clear how word choice in written language differs from word choice in speech, or how many words it takes to truly gain insights into a person's psychology. "We have a lot of ideas and a lot of work to do," he said.
Given the speed of innovation in the AI field, he hopes to find answers sooner rather than later.
"Companies are already selling AI psychological assessment tools to hospitals and clinicians, but it's not clear to me how well they work or how thoroughly they've been evaluated," Oltmanns said. "This sort of technology could be a huge advance for the field of psychology, but it has to be done carefully. We have to be smart."
Originally published on the Ampersand website.
