ABSTRACT. Human users’ needs must be considered at the beginning of system design. However, classical sys-tems engineering approaches consider the needs of several stakeholders (clients, authorities, etc.) but those of end users are often less considered. Hence, neglecting or oversighting such needs will lead to unacceptance and non-adoption of systems by end users. This paper introduces an original ap-proach that combines System Engineering (SE) and User-Centered Design (UCD) approaches to ad-dress the needs of end users from the early stages and throughout the design process of a generic system. This approach is applied to the design of a bidirectional AI-based assistant, incorporating principles of human-machine teaming. It aims at assisting operators in real time network supervision and piloting activities.

ABSTRACT. With the democratisation of Large Language Models, academics and professionals are searching for new use cases of conversational Generative Artificial Intelligence (GenAI) for systems engi-neering, including requirements engineering. This paper presents a pilot study to understand the impact of guidelines and templates on the interaction between ChatGPT and a systems engineer for developing system requirements. Results show that when appropriately used, prompting guide-lines and templates improve the quality of requirements. Still, without domain knowledge, the GenAI cannot generate outputs with the quality expected by requirements engineering internation-al standards.

ABSTRACT. The emergence of artificial intelligence (AI) with advanced natural language processing offers promising approaches for enhancing the capacity of textual classification. The aviation industry is increasingly interested in adopting AI to improve efficiency, safety, and cost efficiency. This study explores the potential and challenges of using AI to analyse decision errors in flight operations based on the HFACS framework. In pre-training, the model is trained based on a large amount of data to predict the next word in a sequence which allows the model to learn relationships between the words and their meaning in the accident investigation reports. Initial discoveries demonstrated that the AI model could supply a consistent HFACS framework and populate these dimensions with moderate accuracy. Future research is focused on the development of this HFACS-GPT model through fine-tuning and deep learning, facilitating more reliable and consistent conversations.

ABSTRACT. The automotive industry commonly adopts a component-oriented approach in the development process, where the focus lies on the components. However, due to the current challenges in the industry, this approach is no longer sufficient to fulfill customer requirements. By shifting the focus to functions instead of components, the enterprise can adapt more easily to changing requirements and manage the increasing complexity. Successful implementation and long-term success of this approach require attention not only to technological aspects but also to human and organizational factors affecting the development process. The research aim is to conceptualize a holistic, function-oriented development approach regarding the technological, human, and organizational factors.

ABSTRACT. For nearly four decades, classical cognitive models, such as Model Human Processor, have provided the theoretical basis for human-computer interaction. With the development of technology, human-computer interaction (HCI) is no longer limited to graphical user interfaces, and various ways of creative natural HCI are emerging. As a future trend, natural HCI urgently needs new theoretical models to guide its design and evaluation. In this paper, in combination with existing cognitive models, we proposed a Shared Mental Model based on the characteristics of natural HCI. The model improved the classical three Perception, Cognition, and Action modules. We also incorporated important findings of social cognition, such as the theory of mind, and proposed a Shared Space Module. The Shared Mental Model can describe and predict the natural HCI process and attempts to provide new theoretical guidance and development suggestions for future interaction methods.

ABSTRACT. This contribution focuses on the current findings regarding the exploration of Human-AI teaming (HAT) (EASA, 2024) for development of an assistant to support pilots in adjusting their route due to a meteorological event in cruise or decide of an alternate airport selection. The concept explores the enhancement of bidirectional communication through operational intentions as high-level shared abstractions, thus improving and accelerating the decision-making process (Hourlier, et al., 2022). HAIKU (2022-2025) is a European funded project aimed at developing Human Factors guidelines for the development of human-centric intelligent assistants for aviation (HAIKU, 2022). The project has six different use cases (UC), distributed across cockpit (2 cases), Air Traffic Management ATM (1 case), Urban Air Mobility (UAM, 1 case), Airport Safety Management (1 case) and Airport Passenger (1 case). In the use case, the HAT was defined as System of Interest (SoI). The adopted requirements framework reflects this approach, enabling the study and trade-off of different team architectures. By placing the human within the SoI, the issues of human–machine regarding situational awareness, communication, operational explainability are identified, studied, and detailed in a more flexible and human-centric way then when the human component is considered as a “fixed” interface for the assistant. In a teaming concept, a strong dependency coupling between human, and machine is expected at higher cognitive level. Thus, the realization of team safety, effectiveness, efficiency, and other emergent qualities, are dependent on a proper definition of roles, responsibilities, transparency, task allocation, authority dynamics, among others. The development method is thus iterative and interactive, coupling the team architecture design with subject matter experts’ interviews and trials with low-fidelity prototypes to stress key team anchoring concepts. An operational and system architectural description is developed, coupled with a control architecture from STPA, where focus is placed upon the teaming mechanisms between user and assistant. This will expand upon the approach from (Andrew Kopeikin, 2024). The presentation will thus provide an overview of the use case, method and key insights derived and planned work. The authors aim to invite critical review to strengthen the results.

ABSTRACT. At present, a large number of studies have explored the influence of the position of a single piece of information in the head-up display (HUD) system on driving performance, but there are few studies on the range of multiple information visual angle. The study aims to examine the positions and visual angles of driving information presentation that enhance drivers’ driving performance. Considering the direction of horizontal and vertical, we designed two experiments to identify an appropriate area to present driving information. Experiment A was designed as a 4 (horizontal angles: 10°, 20°, 30°, 40°) × 3 (vertical positions: -5°, 10°, 20°) within-subject experiment to discuss the effect of the horizontal angle and vertical position of the information on the reaction time and difficulty rating. The results showed that when the vertical position was − 5 ° or 10 ° and the horizontal angle was 10 °, the driver 's reaction time was faster and felt easier. Experiment B was designed as a 3(vertical angles: 5°, 15°, 25°)×3(horizontal positions: -15°, 10°, 25°) within-subject experiment to explore the effect of the vertical angle and horizontal position of the information on the reaction time and difficulty rating. The results showed that when the horizontal position was 10 ° and the vertical angle was 5°, the driver 's reaction time was faster and felt easier. The two experiment results showed that participants responded faster and rating easier when information was placed near the center visual fields. The results may be useful for designers to present multiple driving information on HUD, providing theoretical guidance for HUD interface design.

ABSTRACT. Improving the safety of the approach control system is important for flight safety. Through the analysis of the risk in the operation process of the approach control system, the risk factor set is constructed by using the fault tree analysis method. Then, on this basis, the evaluation index system of its operation safety is established, and the weight is determined by using the fuzzy order relation analysis method. Finally, the multi-level safety evaluation model is constructed by using the extensibility method, and given the formula for calculating security assessment over multiple time periods of the day. Taking the operation of Qingdao air traffic control station as an example, the operation safety is evaluated and calculated, the results of the assessment are given, and the suggestions for improving the operation safety of the control system in recent years are given.

ABSTRACT. For over 70 years, failure mode and effects analysis (FMEA) has been used in development and assessment across products, processes, and services worldwide. In particular has been its application for useability and use error analysis. FMEA is considered a mainstay of predictive failure analysis and reliability, prescribed by multiple international standards. However, despite this level of adoption, FMEA encounters consistent and regular criticisms, particularly related to its ease of use and effectiveness. Research on improvement often focus on specific elements rather than on overall usability of the tool for practitioners. At the same time, the V-model has become a common approach for product design in systems engineering. However, the integration of these two popular processes together can be cumbersome and incompatible under their current uses. In this paper, we review current methodology for FMEA against similar V-model standards. We identify systemic challenges in using FMEA within the systems engineering V-model and suggest approaches for addressing these challenges to better serve FMEA users.

ABSTRACT. As artificial intelligence (AI) becomes increasingly prevalent in society, ensuring responsible and ethical development and use of AI systems is crucial. This study conducts a risk analysis of human interface with AI systems to identify potential hazards and develop mitigation strategies for promoting responsible AI development. This study is component of a larger research project that is proposing the adaptation of testing strategies for responsible development of generative AI. Risks included in this analysis are bias/discrimination, security/privacy concerns, and lack of transparency/reliability are assessed based on probability, impact of cost, schedule, and performance criteria. Following the implementation of mitigation strategies, a re-evaluation of the risks is conducted to gauge their adjusted system risk. The findings demonstrate that implementing targeted mitigation strategies can effectively reduce the likelihood and severity of risks associated with human interaction with AI systems, thus enabling the development of more responsible and ethically sound AI technology. This study contributes to the ongoing discourse on responsible AI development and provides practical insights for organizations seeking to navigate the complexities of human-AI interaction in a responsible manner.

ABSTRACT. Complex sociotechnical systems are likely to display emergent properties that are difficult to an-ticipate at design time. One of the causes for emergence is the system's sensibility to its own op-erational context. Since context heavily drives the behavior of such systems, a proper HSI endeavor should strive to define the relevant contextual elements early, even before the system has been integrated and project resources committed. This paper presents an early context elicitation methodology along with a companion software tool in development that combines the HSI litera-ture, scenario-based design principles and previous context studies. The use of our tool is illustrated through the case study of the design of a remote air traffic control center.

ABSTRACT. Increasing levels of automation is a solution for more efficient and better capacity for railway transportation. For the French railway company SNCF, the first step to enable this transition from manual driving to automated trains is to introduce automated train operation (ATO) to the existing train system. ATO provides a more precise train operation and speed control during the journey. By controlling the train at the operational speed calculated, ATO contributes to minimizing the energy consumption for train driving. In our work, we intend to integrate humans into system design at this early design phase to gain more flexibility and security of the system. This paper presents the ATO functional architecture from its specifications and functional analysis to clarify the task distributions between ATO and train drivers. This analysis identifies the safety-critical functions and tasks in the semi-automated train system. We emphasized these safety-critical functions and tasks while comparing human-in-the-loop simulation (HITLS) activities and the prescribed tasks. These comparisons enable the identification of the safety-related design gaps in the ATO system.

ABSTRACT. In this early stage research paper, we outline our approach for a three-year study that aims to examine the potential influence of work design on control room operators' performance in the context of advanced nuclear reactor operations. The study is inspired by ongoing developments in the nuclear industry towards advanced reactors that are expected to partly transform the work of control room operators. The research intends to answer three key questions concerning anticipated work characteristics in advanced reactors, their differences compared to conventional plants, and how these characteristics can affect control room operators' performance. The proposed approach, along with the study's strengths and potential challenges, are outlined.

ABSTRACT. This early-stage research addresses an essential need to deepen our understanding of enhancing manufacturing resilience by focusing on internal factors such as workforce, manufacturing pro-cesses, and physical assets. Employing Human-System Integration (HSI) principles, the study focuses on the assembly operator within the assembly cell on a real manufacturing process envi-ronment. Recognizing insufficiencies in standards concerning human strain and its connection to performance, the study will eventually propose a simulation model for decision-making. The aim is to deepen our understanding and enhance manufacturing resilience through risk management, considering human factors, and continuously improving the design interface of the manufacturing process. Through simulation, the study will experiment with changes in different parameters re-lated to human factors and assess their effects on process performance. The focus is on the as-sembly operator's physical performance and force generation, considering variables like gender, age, individual differences, and injury recovery timelines. By analyzing the force generation of human operators with various variables, we aim to address the effects of changes on performance. The simulation aims to build decision-making scenarios and assess the impact of changes on per-formance. In the context of HSI in manufacturing, the study promotes for system design that in-corporates technical and human aspects of in manufacturing processes. This integration aims to proactively contribute to the development of a resilient manufacturing system, fostering adapta-bility and robustness to address diverse challenges.

ABSTRACT. Establishing a structural model of stressors for pilot cadets in flight training, this study developed a scale through interviews and literature research. Using the SHEL model for dimensional analysis, 18 stressors across five dimensions were identified. These stressors were evaluated through dimensional analysis and expert scoring, and modeled using DEMATEL and ISM. The model calculated the influencing, influenced, center, and cause degrees of each stressor and established a hierarchical interpretative structure. Key underlying stressors identified include weather conditions and unsafe incidents. Stressors like psychological pressure, difficulty, and progress in flight training were sig-nificantly impacted by other factors. Notably, psychological pressure, training difficulty and progress, and instructor relationships were found to be most influential on cadets’ stress. The DEMATEL-ISM method effectively established a structured hierarchical model of stressors in cadet flight training. The findings suggest that flight training schools should enhance safety management and instructors should positively influence cadets to manage stress effectively.

ABSTRACT. Abstract. To ascertain the psychological factors needed for the pilots in SPO (single pilot operations) crew configuration, a study investigated the effects of professional ability on pilots’ workload in a simulated DPO(dual pilot operations) and SPO task. 46 pilots performed approaches with low visibility using a B737 full flight simulator in DPO and SPO crew configuration respectively, and their workload measured by NASA-TLX. A pilot’s psychological competency measurement tool was used to collect pilots’ professional ability data. The results showed that there were significant differences detected in crew configuration regarding workload and relative indexes. Mostly, the workload in the SPO crew configuration was higher than it was in the DPO crew configuration. Meanwhile, in DPO crew configuration, as the Pilot Flying (PF), better teamwork ability was significantly correlated with a worse performance index in NASA-TLX. In SPO crew configuration, spatial orientation ability was negatively correlated with the mental demand index and physical demand index but positively correlated with the performance index(all ps<0.05). These findings contribute to the selection of pilots working in future SPO aircraft while demonstrating the practical application value of the pilots’ psychological competency measurement tool in safeguarding SPO flight safety.

ABSTRACT. The perceived mental workload of Single Pilot Operations (SPO), as an emerging trend in commercial aviation, has received significant attention. The objective of this study is to investigate the differences in pilots’ perceived mental workload between different role assignments and crew configurations. A total of 57 pilots with commercial pilot licenses participated in this study, undertaking three low-visibility approaches as pilot flying (PF) within a crew setting, pilot monitoring (PM) within a crew setting, and PM within a single-pilot setting, respectively. Their perceived mental workloads were evaluated using the National Aeronautics and Space Administration-Task Load Index (NASA-TLX). The results indicated that pilots experienced higher mental workload when performing as PF within the crew setting compared to their role as PM in both the crew and single-pilot settings. As PMs, compared to the crew setting, pilots reported lower levels of effort in the single-pilot setting while perceiving a higher level of physical demand. The significance of this study lies in providing empirical evidence from the perspective of perceived mental workload regarding the feasibility of normal SPO scenarios.

ABSTRACT. Verification is an integral stage of the system engineering process, partially to capture human errors during the process. There is, however, seemingly less attention given to the potential human errors caused by verification engineers themselves, that is, errors that result from ineffective verification planning and/or execution. We focus on two sets of possible cognitive overloads during the verification process: short term overloads during each verification event and long-term overloads over the system lifecycle. A graph-based mathematical approach is proposed to lower such cognitive overloads, utilizing orthogonality and graphical representation. The research is in progress, with theoretical and empirical validations remaining. Practical engineering considerations will be added to finalize the proposed approach, which will then go through an empirical study for validation.

ABSTRACT. While verification is an integral process of systems engineering, there is no consensus on measures for a full-scale verification complexity and what it represents. Verification engineers can rely on their implicit expertise to determine relative complexity differences, but this is resource intensive and scales badly with large systems. This research aims to define the verification complexity in an explicit and mathematical manner, suggest relevant measures, and propose indicators for verification complexity. Data gathering and experiment design have been finished with varying sizes and interconnections. Background research is being conducted on the verification complexity definition and relevant measures. Once finished, machine learning models will be trained on the measures with the proposed definition as a dependent variable. The trained model will then be analyzed to deter-mine accurate, explicit indicators of verification complexity. These are expected to aid more accurate information propagation between system stakeholders, especially engineers, reducing system development costs.

ABSTRACT. While system complexity is considered an integral piece of information throughout the system development life cycle, the complexity of the verification is given comparatively lower focus in the field of systems engineering. There is no domain-wide consensus on the definition of verification complexity, resulting in disputed complexity measures or lack thereof. Verification is a pervasive task throughout the system development; its insufficient measurement is detrimental to both the system engineers and users. We propose the Verification Complexity Framework as a formal definition of verification complexity. A cube-shaped framework is proposed to cover both static and dynamic complexity through the time axis and the hierarchical complexity layers, covering from external effects to the verification structures. Its modular design allows the framework to be nested to mimic information flow between verification at multiple integration levels. This framework provides a common vocabulary for verification complexity, where both its definition and measurements can be discussed.