2026 Mayo Clinic Research Experiences for Teachers (RET) Poster

2026 Mayo Clinic RET Fellow presents end-of-program poster in collaboration with the Department of Psychology and Psychiatry. Can artificial intelligence help clinicians recognize meaningful changes in speech and emotion? In this presentation, Katie Donlin shares her 2026 Mayo Clinic RET summer research evaluating CASLIM, an AI-based speech emotion recognition system developed to analyze changes in spoken language associated with bipolar disorder. The project compares computer-generated emotion classifications with manual human assessments of clinical speech recordings, examining both model accuracy and the time required for analysis. Results highlight the potential workflow advantages of automated speech analysis while also showing areas where the model needs further refinement, particularly when distinguishing less common emotions. Katie also explores how lessons from her RET experience can translate into the classroom through an AI-powered design project combining artificial intelligence, CAD and 3D printing. Students will use AI as a design coach while learning to critically evaluate its recommendations and strengthen their AI literacy. Topics include: • AI-based speech emotion recognition • Bipolar disorder and clinical speech analysis • Comparing human and machine classifications • Evaluating AI accuracy and workflow efficiency • Responsible AI use and AI literacy • Bringing AI, CAD and 3D printing into education 0:00 Evaluating AI Speech Emotion Recognition 0:19 Speech Changes and Bipolar Disorder 0:43 Introducing the CASLIM AI Model 1:01 Training AI to Recognize Emotions 1:38 Evaluating CASLIM With Clinical Speech 2:29 Measuring Human and Machine Agreement 3:32 Model Accuracy and Performance Results 4:18 Comparing AI and Human Analysis Speed 5:00 Key Findings and Model Limitations 6:02 Bringing AI Learning Into the Classroom 6:45 Building Student AI Literacy Skills 7:07 Closing Reflections and Acknowledgments