Presentation
Investigating a Real-Time Adaptive Procedure System: A Machine Learning Proof-of-Concept for Bridging Work-As-Imagined and Work-As-Done
SessionSF1: Student Research
DescriptionProcedural issues contribute significantly to incidents in high-risk industries, with analyses showing operating procedure problems account for approximately 17% of major chemical process incidents (Baybutt, 2016). Current procedural systems fail to address the persistent gap between Work-As-Imagined (WAI), how work is prescribed, and Work-As-Done (WAD), how workers actually perform tasks in practice (Hollnagel, 2017). This gap reflects workers' necessary adaptations to changing conditions that the "one-size-fits-all" approach of static procedures cannot address. While research demonstrates that effective procedures vary based on worker experience, task frequency, and procedural features (Hendricks & Peres, 2021; Peres et al., 2020), no evidence-based guidance exists for real-time procedural adaptation. This study investigates the feasibility of developing a Real-Time Adaptive Procedure System (R-TAPS) that uses machine learning to understand procedural deviations and adapt procedures dynamically, representing a paradigm shift from compliance monitoring toward learning-oriented systems that support worker expertise in high-risk industrial settings.
Event Type
Lecture
TimeWednesday, October 21st2:10pm - 2:30pm PDT
Location