Large language models (LLMs) have the potential to transform standardized psychological tests and assessments. Artificial intelligence algorithms are capable of analyzing natural human communication and thus complementing information obtained through targeted patient questioning. However, their use also entails specific risks and a number of challenges.
AI could benefit psychologists by bringing a person’s personality into focus through language. According to the authors of a comprehensive review published in Advances in Methods and Practices in Psychological Science, language as an assessment tool has several advantages. It is behavioral, provides a more objective approach to evaluation, and can be natural. Another advantage is scalability. It is also rich, allowing individuals to express themselves in ways that go beyond traditional rating scales.
As the researchers noted in their work, language will likely contribute rapidly to expanding knowledge about psychological characteristics thanks to extraordinary advances in technology. Validated LLM-based tools could also be more easily integrated into routine research and clinical activities involving speech. Their use could save time and resources for both participants/patients and researchers/clinicians.
Josh Oltmanns, one of the co-authors of the comprehensive review of large language models for psychological assessment and an assistant professor of psychological sciences at Washington University in St. Louis, pointed out that words are windows into the brain, and how we choose them and how we say them reveals a great deal about our personality and even our mental health.
Thoughts, feelings, and behavior are reflected in speech. It is not only word choice that matters, but also the manner in which words are spoken. For example, slowed speech may be a sign of depression, whereas overly rapid speech is often associated with anxiety. Other indicators include loudness, tone, or pitch. Rather than subjecting patients to lengthy testing, psychologists could gain valuable insights from speech analysis.
“Artificial intelligence tools trained to detect telling cues in speech could revolutionize psychological assessment,” said Josh Oltmanns. According to him, clinicians may not always catch important verbal signals, but a properly trained computational model will not overlook these often subtle cues.
In theory, a psychologist could ask a client to describe their life and problems, which is a standard part of an initial assessment. In addition to using their own clinical expertise, the professional could then feed this conversation into a program designed to detect personality traits and signs of mental health problems. A computer program analyzing the patient’s speech could subsequently help verify the clinician’s observations or draw attention to something that may have been overlooked.
The advent of language models opens up a new world of possibilities. AI programs can be much faster, more thorough, and more accurate than previous computational models. However, Josh Oltmanns also warned of potential risks. AI tools are often trained on information from the internet, which can introduce a certain degree of bias. If such biases are not addressed in a timely manner, certain cultural differences in speech patterns could be inaccurately labeled by AI as signs of mental health problems.
To prevent such distortions, artificial intelligence models should be trained on diverse patient populations. To this end, the authors are studying hundreds of hours of interviews with more than 1,600 adult residents of St. Louis, collected in the SPAN study. These respondents represent the city’s linguistic diversity, and the use of their speech patterns could help train AI models to approach every patient more fairly.
Editorial Team, Medscope.pro
Sources: 1. Brickman J., Gupta M., Oltmanns J. R. Large language models for psychological assessment: A comprehensive overview. Adv Methods Pract Psychol Sci 2025; 8 (3): 1–26, doi: 10.1177/25152459251343582. 2. Kjell O. N. E., Kjell K., Schwartz H. A. Beyond rating scales: With targeted evaluation, large language models are poised for psychological assessment. Psychiatry Res 2024; 333 : 115667, doi: 10.1016/j.psychres.2023.115667. 3. Woolston C. What do our words say about our minds? WashU Ampersand 2025 Sep 1. Available at: https://artsci.washu.edu/ampersand/what-do-our-words-say-about-our-minds
Did you like this article? Would you like to comment on it? Write to us. We are interested in your opinion. We will not publish it, but we will gladly answer you.
Don‘t have an account? Create new account
Enter the email address that you registered with. We will send you instructions on how to set a new password.
