Can the structure of speech reveal where a psychotic episode is heading?
In this episode, Bruce talks with computational psychiatrist Natalia Mota, MD, PhD, about her research showing how the structure of speech can help differentiate between emerging psychotic disorders earlier and more precisely than traditional methods alone.
Using graph theory and natural language processing, Natalia and her team transform speech into “word graphs” that mathematically measure thought fragmentation and narrative connectedness. In one striking study with first-episode psychosis patients, for example, Natalia successfully predicted emerging schizophrenia versus bipolar disorder with over 90% accuracy -- simply by analyzing how participants described a recent dream.
Together, Bruce and Natalia discuss:
• The broader clinical implications of computational psychiatry for early detection and intervention
• The classic debate between subjective clinical judgment versus statistical prediction, and why both approaches are needed
• Why Natalia’s methodology is not black-box AI
• How education and socioeconomic factors shape language
• Why technology must "keep the human in the loop”
• What speech fragmentation reveals about dementia
• The adolescent mental health crisis and social contagion
This conversation explores a powerful idea:
There are identifiable structures within natural speech patterns — and these structures can reveal a goldmine of hidden clinical information.
If you care about psychotherapy, early intervention, computational psychiatry, or the future of psychiatric diagnosis, this episode will challenge how you think about listening. -----
🔔 Don’t forget to like, comment, subscribe, and share — your support helps us keep these conversations going.
**** Natalia Mota, MD, PhD is a psychiatrist and neuroscientist with a focus on creating novel quantitative methods to measure the flow of thoughts, and to differentiate between causes of psychosis and dementia. Her current research examines speech in psychosis, wake–sleep cycles, and school-based declarative learning.
**** Bruce Wampold, PhD is a psychologist and leading psychotherapy researcher known for his work on the contextual model of psychotherapy and the science of therapeutic relationships. His research explores how and why psychotherapy works, integrating clinical insight with rigorous empirical methods.
--
The Making Therapy Better podcast is brought to you by CarePaths EHR, an electronic health record system designed to help you improve therapy outcomes.https://www.carepaths.com
Follow the Making Therapy Better project:
🌐 Website: www.makingtherapybetter.com
📺 YouTube: /@makingtherapybetter
📸 Instagram: /making_therapy_better
📘 Facebook: /makingtherapybetter
🐦 Twitter: /therapybetter
💼 LinkedIn: /9223245
Produced and edited by Kevin Riordan and Geissy Araújo. Intro/outro music by Chris Haugen, free for public use.
00:00 Bruce intro
01:35 Bruce: Kahneman's work on confidence vs. accuracy
04:33 Bruce: The practical relevance of computational psychiatry and Natalia’s work
06:39 What Is "Computational Psychiatry?"
12:28 Graph Theory for thought disorders
16:44 Research findings
23:18 Predicting clinical trajectories after first-episode psychosis
30:58 Beyond "black box AI"
35:22 Effects of culture and life experience on language usage
40:25 Applications for dementia and other disorders
45:18 Language, social bonds, and mental health
51:22 Closing remarks
55:11 Outro and farewell