Join the Building Healthy Behaviors Across the Life Span Organized Research Center at noon on April 17 for our monthly meeting, focusing on Foundation: AI Capabilities and Limitations.
This session will explore the current strengths and limitations of artificial intelligence (AI) in health care, including pattern recognition, personalization, and real-time adaptation. It will introduce neurosymbolic AI, an approach that combines clinical knowledge with AI to produce safer, more reliable, and clinically meaningful interactions. Examples will highlight applications in behavioral interventions, triage, and decision support, and outline a roadmap for developing more trustworthy health care AI systems.
What AI can do: pattern recognition, personalization at scale, and real-time adaptation
What we want AI to do in health care, but it is not there yet: reliably generate clinically relevant information (e.g., behavioral intervention, triaging, vulnerability assessment, etc.) and ensure safety when interacting with patients and with caregivers
What is the roadmap to develop such an AI:
Focus of neurosymbolic AI: to teach AI how to use clinical knowledge (e.g., CBT Protocols); for instance, in behavioral interventions, how can AI know when to ask Socratic questions and when to do cognitive reframing, based on the user's persona and responses?