Subjective machines: Probabilistic risk assessment based on deep learning of soft information.

For several years machine learning methods have been proposed for risk classification. While machine learning methods have also been used for failure diagnosis and condition monitoring, to the best of our knowledge, these methods have not been used for probabilistic risk assessment. Probabilistic risk assessment is a subjective process. The problem of how well machine learning methods can emulate expert judgments is challenging. Expert judgments are based on mental shortcuts, heuristics, which are susceptible to biases. This paper presents a process for developing natural language-based probabilistic risk assessment models, applying deep learning algorithms to emulate experts' quantified risk estimates. This allows the risk analyst to obtain an a priori risk assessment when there is limited information in the form of text and numeric data. Universal sentence embedding (USE) with gradient boosting regression (GBR) trees trained over limited structured data presented the most promising results. When we apply these models' outputs to generate survival distributions for autonomous systems' likelihood of loss with distance, we observe that for open water and ice shelf operating environments, the differences between the survival distributions generated by the machine learning algorithm and those generated by the experts are not statistically significant.

Quantifying the impact of environment factors on the risk of medical responders' stress-related absenteeism.

Medical emergency response staff are exposed to incidents which may involve high-acuity patients or some intractable or traumatic situations. Previous studies on emergency response staff stress-related absence have focused on perceived factors and their impacts on absence leave. To date, analytical models on absenteeism risk prediction use past absenteeism to predict risk of future absenteeism. We show that these approaches ignore environment data, such as stress factors. The increased use of digital systems in emergency services allows us to gather data that were not available in the past and to apply a data-driven approach to quantify the effect of environment variables on the risk of stress-related absenteeism. We propose a two-stage data-driven framework to identify the variables of importance and to quantify their impact on medical staff stress-related risk of absenteeism. First, machine learning techniques are applied to identify the importance of different stressors on staff stress-related risk of absenteeism. Second, the Cox proportional-hazards model is applied to estimate the relative risk of each stressor. Four significant stressors are identified, these are the average night shift, past stress leave, the squared term of death confirmed by the Emergency Services and completion of the safeguarding form. We discuss counterintuitive results and implications to policy.

Predicting the Validity of Expert Judgments in Assessing the Impact of Risk Mitigation Through Failure Prevention and Correction.

Operational risk management of autonomous vehicles in extreme environments is heavily dependent on expert judgments and, in particular, judgments of the likelihood that a failure mitigation action, via correction and prevention, will annul the consequences of a specific fault. However, extant research has not examined the reliability of experts in estimating the probability of failure mitigation. For systems operations in extreme environments, the probability of failure mitigation is taken as a proxy of the probability of a fault not reoccurring. Using a priori expert judgments for an autonomous underwater vehicle mission in the Arctic and a posteriori mission field data, we subsequently developed a generalized linear model that enabled us to investigate this relationship. We found that the probability of failure mitigation alone cannot be used as a proxy for the probability of fault not reoccurring. We conclude that it is also essential to include the effort to implement the failure mitigation when estimating the probability of fault not reoccurring. The effort is the time taken by a person (measured in person-months) to execute the task required to implement the fault correction action. We show that once a modicum of operational data is obtained, it is possible to define a generalized linear logistic model to estimate the probability a fault not reoccurring. We discuss how our findings are important to all autonomous vehicle operations and how similar operations can benefit from revising expert judgments of risk mitigation to take account of the effort required to reduce key risks.

Human Error in Autonomous Underwater Vehicle Deployment: A System Dynamics Approach.

The use of autonomous underwater vehicles (AUVs) for various applications have grown with maturing technology and improved accessibility. The deployment of AUVs for under-ice marine science research in the Antarctic is one such example. However, a higher risk of AUV loss is present during such endeavors due to the extremities in the Antarctic. A thorough analysis of risks is therefore crucial for formulating effective risk control policies and achieving a lower risk of loss. Existing risk analysis approaches focused predominantly on the technical aspects, as well as identifying static cause and effect relationships in the chain of events leading to AUV loss. Comparatively, the complex interrelationships between risk variables and other aspects of risk such as human errors have received much lesser attention. In this article, a systems-based risk analysis framework facilitated by system dynamics methodology is proposed to overcome existing shortfalls. To demonstrate usefulness of the framework, it is applied on an actual AUV program to examine the occurrence of human error during Antarctic deployment. Simulation of the resultant risk model showed an overall decline in human error incident rate with the increase in experience of the AUV team. Scenario analysis based on the example provided policy recommendations in areas of training, practice runs, recruitment policy, and setting of risk tolerance level. The proposed risk analysis framework is pragmatically useful for risk analysis of future AUV programs to ensure the sustainability of operations, facilitating both better control and monitoring of risk.

Fuzzy System Dynamics Risk Analysis (FuSDRA) of Autonomous Underwater Vehicle Operations in the Antarctic.

With the maturing of autonomous technology and better accessibility, there has been a growing interest in the use of autonomous underwater vehicles (AUVs). The deployment of AUVs for under-ice marine science research in the Antarctic is one such example. However, a higher risk of AUV loss is present during such endeavors due to the extreme operating environment. To control the risk of loss, existing risk analyses approaches tend to focus more on the AUV's technical aspects and neglect the role of soft factors, such as organizational and human influences. In addition, the dynamic and complex interrelationships of risk variables are also often overlooked due to uncertainties and challenges in quantification. To overcome these shortfalls, a hybrid fuzzy system dynamics risk analysis (FuSDRA) is proposed. In the FuSDRA framework, system dynamics models the interrelationships between risk variables from different dimensions and considers the time-dependent nature of risk while fuzzy logic accounts for uncertainties. To demonstrate its application, an example based on an actual Antarctic AUV program is presented. Focusing on funding and experience of the AUV team, simulation of the FuSDRA risk model shows a declining risk of loss from 0.293 in the early years of the Antarctic AUV program, reaching a minimum of 0.206 before increasing again in later years. Risk control policy recommendations were then derived from the analysis. The example demonstrated how FuSDRA can be applied to inform funding and risk management strategies, or broader application both within the AUV domain and on other complex technological systems.

A Fuzzy-Based Risk Assessment Framework for Autonomous Underwater Vehicle Under-Ice Missions.

The use of autonomous underwater vehicles (AUVs) for various scientific, commercial, and military applications has become more common with maturing technology and improved accessibility. One relatively new development lies in the use of AUVs for under-ice marine science research in the Antarctic. The extreme environment, ice cover, and inaccessibility as compared to open-water missions can result in a higher risk of loss. Therefore, having an effective assessment of risks before undertaking any Antarctic under-ice missions is crucial to ensure an AUV's survival. Existing risk assessment approaches predominantly focused on the use of historical fault log data of an AUV and elicitation of experts' opinions for probabilistic quantification. However, an AUV program in its early phases lacks historical data and any assessment of risk may be vague and ambiguous. In this article, a fuzzy-based risk assessment framework is proposed for quantifying the risk of AUV loss under ice. The framework uses the knowledge, prior experience of available subject matter experts, and the widely used semiquantitative risk assessment matrix, albeit in a new form. A well-developed example based on an upcoming mission by an ISE-explorer class AUV is presented to demonstrate the application and effectiveness of the proposed framework. The example demonstrates that the proposed fuzzy-based risk assessment framework is pragmatically useful for future under-ice AUV deployments. Sensitivity analysis demonstrates the validity of the proposed method.