This dissertation explores the use of the efficiency performance measurement paradigm (EM), in terms of its concepts and applications, as an ex-ante mechanism to evaluate enterprise performance and inform enterprise design. The design of an enterprise is driven by decisions that include, but not limit to, which strategies to implement, how to allocate resources, how to shift operating patterns, and how to boost coordination among enterprises, among others. Up to date, EM has been mainly used as a descriptive mechanism, but the fundamental reason for measuring performance in an ex-post fashion, i.e., how well an enterprise does, is also valid in the context of design decisions, i.e., ex-ante evaluation. The contrast between the ex-post and ex-ante use of EM relates to the measurement purpose, i.e., why to measure. Ex-post measurement focuses on evaluating 'what happened' (non-disruptive) while ex-ante measurement emphasizes in informing design decisions exploring changes in current settings (more disruptive). Within this context and to achieve the purpose above, this dissertation is supported by theoretical insights and complemented with three empirical studies. The theoretical insights relate to facts that support, connect to, and challenge (i.e., facilitate or impede) the ex-ante use of EM for enterprise evaluation and informing enterprise design. Those insights are based on the efficiency performance measurement, organizational design and enterprise systems engineering literature. Meanwhile, the three empirical studies situate the application of EM as an ex-ante mechanism to inform evacuation management, bank branch management, and power plants. The theoretical and empirical results indicate that EM is well suited for both evaluating enterprise performance and informing design decisions. The main contribution of this dissertation to enterprise stakeholders is that EM can be not only used to answer how well the enterprise did, but also how well it could do if certain design decisions are taken. / Ph. D.

This dissertation brings together the areas of safety science and operations management through a mixed-methods approach to investigate the complex relationships between two, often conflicting, organizational goals - efficiency and safety, in sociotechnical systems (STSs). This research mainly focuses on production pressure (PrP) which is considered as one of the main negative outcomes of overprioritizing the efficiency aspect of STSs. This work seeks to introduce novel methodologies for assessing PrP in real time for the purpose of mitigating its risks and unwanted consequences, particularly in safety critical environments such as traffic control centers (TCCs). Essay 1 concentrates on the theoretical underpinnings of PrP by systematically reviewing the existing literature to clarify and unify the concept under the context of safety science. It identifies key factors contributing to PrP, its negative effects on safety performance in various industries, and potential mitigation strategies. By doing so, this essay contributes to the field through laying the groundwork for more effective management strategies to improve workplace safety. Essay 2 addresses a significant gap identified in Essay 1 by developing a methodology based on Data Envelopment Analysis (DEA) for the ongoing measurement and monitoring of PrP. This innovative approach introduces a quantitative mechanism that juxtaposes efficiency and safety related outcomes of hourly performance in safety critical environments. This proposed method allows for a detailed analysis of performance dynamics within STSs. The practical application of this model is demonstrated through its implementation in the infrastructure management system of INFRABEL, the Belgian National Railroad Company. Essay 3 advances the conversation by tackling the predictive limitations of the DEA model established in Essay 2. It integrates Machine Learning (ML) techniques with DEA to develop an innovative method for forecasting near-future PrP levels for proactive management of safety risks. The major contribution of Essay 3 is the novel interface between ML and DEA that can improve decision-making capabilities of managers in safety-critical STSs through real-time monitoring and predictive analytics. Together, these studies contribute to the theoretical discussions around PrP and present practical solutions to longstanding challenges in safety science and operational management. / Doctor of Philosophy / In today's increasingly complex world, the systems that run our industries, from traffic control to healthcare, face a dilemmatic balance between pushing for higher productivity and ensuring safety. This dissertation explores the trade-offs between efficiency and safety which has become more pronounced with the advancement of technology. Traditional safety approaches which used to be effective in simpler systems, struggle in modern STSs where causes and effects are not linear but tangled in a web of unpredictable interactions. Production pressure (PrP), at the core of the mentioned balance, is the drive to maximize output and efficiency, often at the expense of safety. This pressure can lead to unintended and sometimes catastrophic outcomes in the long term, especially in environments where safety is critical, such as rail traffic control centers. Despite its vital impact, there has been a noticeable gap in understanding and managing PrP. In fact, existing safety frameworks are struggling to capture the dynamic nature of PrP, consequently, its real-time measurement and control remain difficult to achieve. This work, therefore, tries to broaden our understanding of PrP and to develop methods to monitor, measure, and predict it, to equip managers and policymakers with the tools to navigate the efficiency-safety dichotomy more effectively. Through a series of essays, this dissertation reviews the current state of knowledge on PrP to identify its sources and impacts and also innovates a novel approach to quantify PrP in real-time and predict its future trends.

Sociotechnical systems (STSs) rely on the collaboration between humans and autonomous decision-making units to fulfill their objectives. Highly intertwined social and technical contextual factors influence the collaboration between these human and engineered elements, and consequently the performance characteristics of the STS. In the next two decades, the role allocated to STSs in our society will drastically increase. Thus, the effective design of STSs requires an improved understanding of the human-autonomy interdependency. This dissertation brings together management science along with systems thinking and uses a mixed-methods approach to investigate the interdependencies between people and the autonomous systems they collaborate within complex socio-technical enterprises. The dissertation is organized in three mutually exclusive essays, each investigating a distinct facet of STSs: safe management, collaboration, and efficiency measurement. The first essay investigates the amount of work allocated to safety-critical decision makers and quantifies Rasmussen's workload boundary that represents the limit of attainable workload. The major contribution of this study is to quantify the qualitative theoretical construct of the workload boundary through a Pareto-Koopmans frontier. This frontier allows one to capture the aggregate impact of the social and technical factors that originate from operational conditions on workload. The second essay studies how teams of humans and their autonomous partners share work, given their subjective preferences and contextual operational conditions. This study presents a novel integration of machine learning algorithms in an efficiency measurement framework to understand the influence of contextual factors. The results demonstrate that autonomous units successfully handle relatively simple operational conditions, while complex operational conditions require both workers and their autonomous counterparts to collaborate towards common objectives. The third essay explores the complementary and contrasting roles of efficiency measurement approaches that deal with the influence of contextual factors and their sensitivity to sample size. The results are organized in a structured taxonomy of their fundamental assumptions, limitations, mathematical structure, sensitivity to sample size, and their practical usefulness. To summarize, this dissertation provides an interdisciplinary and pragmatic research approach that benefits from the strengths of both theoretical and data-driven empirical approaches. Broader impacts of this dissertation are disseminated among the literatures of systems engineering, operations research, management science, and mechanical design. / Doctor of Philosophy / A system is an integrated set of elements that achieve a purpose or goal. An autonomous system (ADS) is an engineered element that often substitutes for a human decision-maker, such as in the case of an autonomous vehicle. Sociotechnical systems (STSs) are systems that involve the collaboration of a human decision-maker with an ADS to fulfill their objectives. Historically, STSs have been used primarily for handling safety critical tasks, such as management of nuclear power plants. By design, STSs rely heavily on a collaboration between humans and ADS decision-makers. Therefore, the overall characteristics of a STS, such as system safety, performance, or reliability; is fully dependent on human decisions. The problem with that is that people are independent entities, who can be influenced by operational conditions. Unlike their engineered counterparts, people can be cognitively challenged, tired, or distracted, and consequently make mistakes. The current dependency on human decisions, incentivize business owners and engineers alike to increase the level of automation in engineered systems. This allows them to reduce operational costs, increase performance, and minimize human errors. However, the recent commercial aircraft accidents (e.g., Boeing 737-MAX) have indicated that increasing the level of automation is not always the best strategy. Given that increasing technological capabilities will spread the adoption of STSs, vast majority of existing jobs will either be fully replaced by an ADS or will change from a manual set-up into a STS. Therefore, we need a better understanding of the relationships between social (human) and engineered elements. This dissertation, brings together management science with systems thinking to investigate the dependencies between people and the autonomous systems they collaborate within complex socio-technical enterprises. The dissertation is organized in three mutually exclusive essays, each investigating a distinct facet of STSs: safe management, collaboration, and efficiency measurement. The first essay investigates the amount of work handled by safety-critical decision makers in STSs. Primary contribution of this study is to use an analytic method to quantify the amount of work a person could safely handle within a STSs. This method also allows to capture the aggregate impact of the social and technical factors that originate from operational conditions on workload. The second essay studies how teams of humans and their autonomous partners share work, given their preferences and operational conditions. This study presents a novel integration of machine learning algorithms to understand operational influences that propel a human-decision maker to handle the work manually or delegate it to ADSs. The results demonstrate that autonomous units successfully handle simple operational conditions. More complex conditions require both workers and their autonomous counterparts to collaborate towards common objectives. The third essay explores the complementary and contrasting roles of data-driven analytical management approaches that deal with the operational factors and investigates their sensitivity to sample size. The results are organized based on their fundamental assumptions, limitations, mathematical structure, sensitivity to sample size, and their practical usefulness. To summarize, this dissertation provides an interdisciplinary and pragmatic research approach that benefits from the strengths of both theoretical and data-driven empirical approaches. Broader impacts of this dissertation are disseminated among the literatures of systems engineering, operations research, management science, and mechanical design.