BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Phoenix
X-LIC-LOCATION:America/Phoenix
BEGIN:STANDARD
TZOFFSETFROM:-0700
TZOFFSETTO:-0700
TZNAME:MST
DTSTART:19700101T000000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20241014T203101Z
LOCATION:Grand Ballroom
DTSTART;TZID=America/Phoenix:20240910T152000
DTEND;TZID=America/Phoenix:20240910T154000
UID:HFESAM_ASPIRE - Presented by HFES_sess157_LECT432@linklings.com
SUMMARY:Assessing GenAI's Intra-Agent Reliability in Coding Unstructured H
 uman Factors Data within Patient Safety Reports: An Empirical Investigatio
 n
DESCRIPTION:Lecture\n\nDouglas Wiegmann (University of Wisconsin - Madison
 ), Arsalan Ahmad and Felix Nguyen (UW-Madison), and Scott Shappell (ERAU)\
 n\nThe rise of large language models (LLMs) and Generative Pre-trained Tra
 nsformers (GPTs) has sparked interest in leveraging specialized, generativ
 e AI (GenAI) agents for analyzing and coding human factors data in patient
  safety reports. This has led to questions about these agents' capabilitie
 s to detect human factor (HF) issues within unstructured, narrative data a
 nd then reliably code them using various human factors frameworks. To expl
 ore this potential, our lab is conducting systematic experiments with vari
 ous specialized GPT agent configurations, focusing on enhancing their abil
 ity to analyze patient harm reports to identify HF issues.  Our preliminar
 y results are encouraging, demonstrating generative AI's transformative po
 wer in processing unstructured data. These findings suggest a future where
  AI not only supports but significantly advances analytical methods. Howev
 er, we proceed cautiously, mindful of the challenges and limitations at th
 is early adoption stage, including issues with GenAI’s output reliability 
 and interpretability.\n\nTrack: Health Care\n\nSession Chairs: Maryam Tabi
 bzadeh (California State University, Northridge) and Shababa Matin (Rice U
 niversity)
END:VEVENT
END:VCALENDAR