Why Do Human-Machine Teams Fail: Investigating Failure Mechanisms in Human Reliability Analysis


Human Reliability Analysis (HRA), whether qualitative or quantitative, typically focuses on the how of “human failure.” For instance, the possible modes (Crew Failure Modes – CFMs) of failure define how a Major Crew Function (MCF) can be unsuccessfully performed by a crew of operators. The CFMs are typically associated with various Performance Influencing Factors (PIFs) that change the probability of occurrence (and thus, the probability of failing the MCF). However, the cognitive and/or physical pathways that result in a specific CFM (i.e., the question of why a failure occurs) are not often explicitly included in HRA methods. Even in Bayesian network (BN) models, these failure pathways are found implicitly (if at all) in the encoded causal relationships between the lower-level PIFs and higher-level CFMs. Our work posits that failure mechanisms may be revealed by cognitive-physical patterns underlying the presence of identified failure modes, i.e., specific sets of PIFs and/or subtasks (crew activity primitives – CAPs) specify failure mechanisms that can result in the failure mode. Our research reviews previous works that cover failure mechanisms in HRA and identifies possible failure mechanisms in tasks related to information gathering. We then extrapolate this finding to propose a definition for the failure mechanism in HRA and a preliminary procedure for identifying and modeling failure mechanisms in BN models.