Kresge Center Lecture: AI's promise and pressure in nursing and healthcare
On Feb. 6, 2026, more than 450 attendees gathered at Binghamton University for the Kresge Center Lecture hosted by Decker College of Nursing and Health Sciences. Dean Mario R. Ortiz opened the event and introduced keynote speaker Maxim Topaz, PhD, RN, MA, FAAN, FACMI, FIAHSI. Topaz is the Elizabeth Standish Gill Associate Professor of Nursing and Data Science at Columbia University, a senior research scientist at VNS Health, and a visiting scholar at the University of Maribor in Slovenia.
His message was clear: "We need nurses involved in AI technology right now." The discussion focused on practical AI use in clinical settings-what works, what breaks, and how to build workflows that actually help patients and clinicians.
Why nurses must be in the AI loop
Topaz is a pioneer in applying natural language processing to nursing documentation. His research explores AI, text mining, and automated speech and video processing to improve care, from predicting patient deterioration to reducing documentation burden.
He emphasized that if AI is implemented well, clinicians can make informed decisions faster. "If AI tools are integrated appropriately into our practice, we can make informed decisions based on the best available literature and the best available data analysis." The heavy lift-synthesizing evidence and data-can be offloaded to machines, while humans keep final say.
Lots of questions, lots of bad answers
Topaz noted that an OpenAI report estimated more than 40 million people use ChatGPT daily for health information. Yet studies show hallucination rates between 10-30%. That means a sizable share of answers can look credible but be wrong-or irrelevant.
This is where nursing judgment matters. AI can summarize, suggest, and surface patterns. It cannot replace clinical sense-making, context, or patient advocacy. Continuous involvement of nurses from development through refinement is non-negotiable if these tools are going to work for patients and staff.
Panel: ethics, accountability, and patient choice
Following the keynote, Topaz moderated a panel with three leaders in AI and nursing: Ann Fronczek, PhD, RN (Binghamton University), Laura-Maria Peltonen, PhD, RN (University of Eastern Finland / University of Turku / Warwick Medical School), and Meghan Reading Turchioe, PhD, MPH, RN, FAHA (Columbia University). The group tackled real decisions clinicians face as AI shows up in everyday care.
- Should patients allow tools like ChatGPT to access their medical records?
- Should nurses be held legally responsible for decisions made with AI assistance?
- Should patients be able to refuse AI involvement even if outcomes could be worse?
The panel also highlighted a new legal boundary: Oregon's 2025 law barring AI or other nonhuman entities from using the title "nurse." The intent is to protect patients from being misled and to preserve the human skills-empathy, critical thinking, nuanced judgment-that define nursing.
From left, speakers at the 2026 Kresge Center Lecture were: Maxim Topaz, Dean Mario R. Ortiz, Laura-Maria Peltonen, Ann Fronczek and Meghan Reading Turchioe.
What healthcare leaders can do now
- Stand up a multidisciplinary AI governance group that includes bedside nurses, informatics, legal, risk, and patient reps.
- Pick focused use cases (e.g., documentation support, triage prioritization) and run small pilots with clear success criteria.
- Require rigorous validation, bias testing, and drift monitoring. Set thresholds and fallback plans when models underperform.
- Make AI use visible to patients. Offer a path to request human-only workflows where feasible.
- Define documentation standards: when AI informed a decision, what was accepted/overridden, and why.
- Limit data exposure. If AI touches medical records, enforce PHI safeguards, access logs, and vendor BAAs.
- Train staff on prompts, verification, and escalation. Teach when not to use AI.
- Target documentation burden reduction (voice-to-text, NLP summaries) and measure time saved vs. error risk.
- Set vendor requirements: audit trails, EHR integration, downtime plans, human override, explainability, and model update cadence.
- Build continuous nurse feedback loops from deployment through maintenance. Ship updates based on real workflow pain.
Research and community
The lecture included a poster session with projects from across Decker College, the University, and healthcare partners. Sponsors included Decker College's Office of Research and Scholarship, the Center for AI in Society, and the Roger L. and Mary K. Kresge Center. The theme across the day: keep nurses at the center of AI decisions so tools serve patients-not the other way around.
Resources
Your membership also unlocks:
