The Impact of Automation Systems on Commercial Aircraft: A Human-Machine Interaction Perspective

The Impact of Automation Systems on Commercial Aircraft: A Human-Machine Interaction Perspective | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article The Impact of Automation Systems on Commercial Aircraft: A Human-Machine Interaction Perspective Lei Du, Ying Li, Xingye Wang, Yangliang Wu, Mingliang Li, Tao Wen, and 5 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7171123/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Since the introduction of automation systems in commercial aircraft, flight safety has improved significantly. However, pilots’ increasing reliance on these systems has led to a decline in their manual flying skills. When automated systems display unexpected behaviors, pilots experience a substantial increase in workload, which may lead to additional unforeseen automation surprises and pose serious risks to flight safety. This study explores the mechanisms underlying automation surprises by mining aviation accident databases and constructing a Bayesian network model. Based on this model, the key factors contributing to automation surprises are further explored, and a human-machine interaction model for commercial aircraft is developed. The findings indicate that building an interaction model between pilots and automated systems helps pilots develop better automation safety awareness. Moreover, “human-centered artificial intelligence” can assist in mitigating safety issues in human-machine interactions from a system design perspective. Physical sciences/Engineering Physical sciences/Mathematics and computing Health sciences/Risk factors Automated systems Commercial aircraft Automation Surprise Aviation accident data Models Human-computer interaction Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Introduction In the field of aviation, safety has always been a major concern. With substantial advancements in automation and aircraft manufacturing technologies, modern aircraft and avionics have become increasingly sophisticated and reliable. As a result, pilots have transitioned from direct operators of aircraft to comprehensive system managers. This shift in pilot role has likely introduced new error patterns 1 . According to the International Air Transport Association's safety report on the causes of aviation accidents from 2017 to 2021, manual operation or flight control accounted for 33% of crew-related errors, ranking first among human factors 2 . The transformation of pilot responsibilities that has brought new error patterns poses additional training requirements 3-5 . In addition, it has increased cognitive and attentional demands, thereby prolonging the time pilots need to detect and correct errors 6 .When an automated system does not perform as expected or fails to operate entirely, an automation surprise may occur in which pilots are surprised by the system's actual behavior 7 . Due to the increasing role of automation management in pilot duties, pilots tend to develop trust in and reliance on automation through prolonged interaction, leading to a degradation of manual flying skills and the emergence of automation surprises 8 . While previous research has identified some factors influencing automation surprise, the specific mechanisms of automation surprise formation in commercial aviation and the associated human-machine interaction (HMI)remain underexplored. With the rapid development of artificial intelligence (AI), AI-enabled automation is expected to evolve from assisting pilots to collaborating with them, and eventually to operating autonomously 9 . In this transition, AI-driven automation systems must be transparent, presenting essential information to help pilots understand system behaviors. Large Language Models (LLMs), trained on aviation-specific knowledge, can serve to communicate and explain system behavior, thereby aligning intentions between pilots and automated systems. This approach helps eliminate the occurrence of automation surprises from a design standpoint and reduces their likelihood through improved human-machine interaction. Introduction Major Automation Equipment: According to a report by the Federal Aviation Administration 10 , Modern commercial aircraft cockpits are commonly equipped with the following automation systems: A powerful Flight Management System (FMS), also known as a Flight Management Computer (FMC), which is used to enhance flight path management; Advanced navigation systems, such as Area Navigation (RNAV), Required Navigation Performance (RNP), and the Global Positioning System (GPS); More sophisticated digital display systems integrated into cockpit components, such as the Primary Flight Display (PFD), Navigation Display (ND), Mode Control Panel (MCP), Multi-Function Display (MFD), and Control Display Unit (CDU). These automation systems are illustrated in Fig. 1 . Main Functions of Automation Implementation: Pilots typically configure combinations of these automation systems to achieve the desired flight performance. The configurations are generally standardized and exhibit inherent logical relationships among the systems (see Fig. 2 ). For example, Moriarty 11 identified several high-level automated flight control modes, including the following: Autopilot (AP) or Autothrottle (AT) engagement, combined with manual selection of Heading (HDG), Very High Frequency Omnidirectional Range (VOR), or Localizer (LOC) navigation modes AP or AT engagement in conjunction with FMS guidance, in which case the aircraft can follow a predefined route with minimal pilot intervention—typically corresponding to the cruise phase of a flight. However,the benefits of these automation systems are not always easily realized. The introduction of complex automated systems requires pilots to spend a significant amount of time monitoring them to ensure timely detection and resolution of anomalies. This shift in pilot responsibilities has introduced new patterns of error and increased cognitive, attentional, and training demands 3–5 . In this study, we manually mined case data from NASA's Aviation Safety Reporting System (ASRS). Elements corresponding to the automation surprise mechanism in the FLight deck Automation Problems (FLAP) model were manually labeled and categorized. The FLAP model examines not only the interaction of humans and automated systems when they are functioning properly, but also includes crew defects, and specifically examines scenarios such as automated system failures and avionics system anomalies.Based on this categorization, a Bayesian Network was constructed to model the formation mechanism of automation surprise. The backpropagation algorithm was then applied to identify the key factors leading to severe consequences and the most influential contributors to automation surprise. A human-machine interaction model for commercial aircraft was ultimately developed. This model helps bridge the gap between the severity of automation incidents and related aspects such as flight management systems, mental models, and decision-making coordination from the perspective of human-machine interaction. Automation Surprise: Surprise is typically a reaction to events that deviate from prior expectations 12 . Automation surprise, in particular, thus occurs when a malfunction in an automated system leads to a mode transition, or when an inappropriate automation mode is activated by the pilot, causing the system to disengage without explicit pilot command. In such scenarios, pilots may struggle to comprehend the automated system's behavior, find it difficult to regain control of the aircraft, and, in extreme cases, lose control in flight (LOC-I) 13 . Numerous studies, particularly those based on questionnaire surveys, have identified various direct and indirect factors influencing automation surprise 14 . Trippe and Mauro 15 categorized the contributing factors to automation surprise into three broad domains: Automation system factors: level of automation, system failure, system design Pilot-related factors: flight experience, risk perception, mindfulness level Interface factors: false or unclear display of information The occurrence of automation surprise is closely related to pilot characteristics such as experience, risk perception, and mindfulness 16,17 . Pilots respond to automation surprise based on their knowledge of aircraft automation and operational experience. Threat to Aviation Safety: De Boer and Hurts 14 classified the outcomes of automation surprise events into different severity levels (as shown in Table 1 ). They found that most events fell into levels 1–3, with only a few rated at levels 4–5. Moreover, most incidents did not result in aircraft damage, and very few involved physical injury or fatalities among crew or passengers 15 . Table 1 Severity classification of automation surprise consequences severity rating Type of consequence 1 No result 2 Reduced automation Reuse automated deviation from FMS program after calibration 3 Manual takeover/correction Prepare to land and resume flight Significant increase in vectoring/air control assistance workloads 4 Speed deviation is significant Serious deviation from course/path Serious deviation from altitude Reporting of security reports/technical logs 5 Unstable flight 6 Organic damage Note: The higher the severity level, the higher the hazard to aviation safety. Nonetheless, from an aviation safety perspective, automation surprise represents at least a potential threat. Dekker 18 emphasized that automation surprise becomes evident only when pilots notice the system behaving in an unexpected manner—but by that point, significant consequences may have already occurred. Automation surprise is considered a specific manifestation of the broader "surprise effect" 19 . Such events can narrow a pilot’s attention, increase their cognitive workload, reduce their situational awareness, and even temporarily compromise their control of the aircraft 20 . Furthermore, handling automation surprises can interrupt a pilot's ongoing tasks, reducing the attention allocated to the primary task. Furthermore, irrelevant information introduced by the interruption may interfere with working memory. Fatigue exacerbates these effects, negatively impacting attention, memory, and task performance 21 . As a result, automation surprise can significantly impair a pilot’s ability to effectively operate the aircraft. In addition, automation surprise may disrupt air traffic operations, as controllers need to redirect other aircraft to accommodate the affected crew's handling of unexpected aircraft behavior 15 . In summary, automation surprise remains an issue warranting ongoing attention 22 . Man-Machine Conflicts: Mental models are typically developed through training and operational experience. These models are used to reason about and control the underlying processes of socio-technical systems 23 . In aviation, a cognitive mismatch arises when a pilot’s mental model of the automation does not accurately reflect the system’s actual behavior 19,22,24 . Cognitive mismatches can take various forms, such as mode confusion, in which pilots believe the automation is operating in one mode when it is actually functioning in another 25 . Under such conditions, automation surprises occur 26,27 . Correspondingly, Dekker 18 defines automation surprise as a mismatch between the operator’s mental model (e.g., expectations or intentions) and the actual behavior of the automation (see Fig. 3 ). From this perspective, automation surprise is closely related to human-machine conflict. Automated FLAP Model: Ancel and Shih 28 proposed the FLAP model based on automation-related aviation accidents and the human-machine interaction literature. The FLAP model provides a top-down, systematic framework for simulating automation-related problems in commercial flight operations. For example, Wang et al. 29 employed the FLAP model to successfully trace the root causes of two Boeing 737 MAX crash accidents. The model includes a mechanism of automation surprise (AS), as shown in Fig. 4 . When pilots lack system-related knowledge and a comprehensive mental model of automation, automated behavior may be poorly understood and automation surprises are more likely to occur 4 . Thus, inadequate decision making and degraded monitoring awareness may follow. Reduced situational monitoring can further lead to decision failures. Unexpected automation behaviors may become the source of crew confusion or judgment errors, ultimately resulting in automation surprise 27 . As discussed above, most research on automation surprise has focused on its negative impact on aviation safety. In particular, much attention has been paid to the factors contributing to automation surprise. However, apart from the FLAP model proposed by Ancel and Shih 28 , few studies have systematically examined the formation mechanism of automation surprise and its implications for safety. Furthermore, the aviation accident datasets used to construct and validate the FLAP model remain very limited. Therefore, it is essential to mine additional aviation case data to validate the automation surprise formation mechanism illustrated in Fig. 4 . Automated AI Enabling Psychology: With the rapid development of artificial intelligence (AI), automation is expected to evolve from assisting pilots, to collaborating with them, and, ultimately, to operating autonomously 30 . During this transition, AI-enabled automation must exhibit greater transparency, providing pilots with the essential information needed to understand system behaviors. Research into automation surprise can also help advance intelligent system technologies, particularly in the development of AI systems with high transparency and explainability 30,31 . Large Language Models (LLMs), trained on aviation-specific knowledge, can interpret and explain automation behavior to help align pilot and system intentions 32 . This design approach can significantly reduce the likelihood of automation surprises from the outset. Building on conceptual models of surprise and startle during flight, some researchers have investigated pilots’ emotional responses to unexpected engine failures using physiological indicators such as heart rate, skin conductance, and eye tracking 24,33 . However, few studies have explored the neural activity of pilots during automation surprise events. Liu et al. 34 highlighted the advantages of functional near-infrared spectroscopy (fNIRS) in aviation psychology research. fNIRS allows for continuous, non-invasive, and portable measurement of hemodynamic changes related to brain function 35 , enabling inferences about a pilot’s cognitive state during flight. Thus, fNIRS enables the observation of neural responses to automation surprise and the measurement of mental workload. Additionally, eye-tracking technology has already seen widespread use in cockpit human factors design 36 . Future studies may adopt neuroergonomics approaches by combining near-infrared and eye-tracking technologies in full-flight simulators 37 . The combination will allow researchers to examine the neural impacts of automation surprise on pilots and to assess the effectiveness of enhanced training and AI-empowered system design in mitigating automation surprise. Data and Methods Automation Surprise Formation Mechanism Based on ASRS Data Mining We constructed a Bayesian network based on case reports related to automation surprise from the Aviation Safety Reporting System (ASRS) database. the Bayesian network has an important advantage: when additional information (e.g., data, evidence) on some random variables is available, the Bayesian network enables real-time updating of system state predictions through forward-looking forecasts and backward-tracing analysis. The network allows us to gain deeper insight into the mechanisms underlying automation surprise, identify the key factors contributing to severe consequences, and distinguish between different types of automation systems (e.g., warning and alerting systems, electronic flight instrument systems, and flight management systems). Typical automation surprise scenarios associated with each system type were summarized to inform model development. Case Data Filtering and Marking The data for this study were retrieved from the ASRS database ( https://asrs.arc.nasa.gov/search/database.html ). The criteria for the search, which was conducted on March 1, 2022, are shown in Fig. 5 . The selected time window spanned from January 2010 to December 2021. The operational type was restricted to Part 121 operations, which sets the standard for commercial aircraft opration. The human factor category was limited to human-machine interaction. The text query used root terms related to surprise: “surpri%”, “freez%”, “perplex%”, “confuse%”, “shock%”, and “unexpect%”, in which “%” includes all word derivatives to ensure comprehensive retrieval. The direct search yielded 305 reports, which were then downloaded for further processing. Subsequently, a five-member research team—consisting of three doctoral students in aviation psychology (each of whom is also a captain of a commercial aircraft), one avionics engineer, and one human factors expert—individually read and discussed each incident report. Incidents that mentioned automation, autopilot, autothrottle, flight management system (FMS), flight management computer (FMC), mode control panel (MCP), lateral navigation (LNAV), vertical navigation (VNAV), or their abbreviations were retained. Information related to air traffic controllers (ATCs), passengers, cabin crew, dispatchers, maintenance personnel, ground handlers, or other unrelated individuals was excluded. In total, 281 automation surprise events were identified for further analysis. The five-member team then labeled each case based on factors associated with automation surprise in the FLAP model, such as the following: The specific automation system involved in the incident Whether the pilot's mental model of the system was sufficient or lacking Whether the pilot’s monitoring awareness was high or low prior to the incident Whether the decision-making quality was high or low The severity of the consequences after the surprise occurred This labeling process prepared the dataset for the subsequent Bayesian network modeling. Bayesian Network Model Construction Based on the labeled case data and the automation surprise formation mechanism in the FLAP model, a data-driven Bayesian network was constructed 38 (see Fig. 6 ). The backpropagation algorithm was then applied to identify which factors or combinations of factors contributed to different levels of outcome severity. Results Analysis Distribution of Surprise Events The distribution of pilot-reported surprise events by manufacturer was as follows: Airbus (29%), Boeing (30%), Bombardier (20%), Embraer (10%), and McDonnell Douglas (5%) (see Table 2 ). A chi-square test indicated a significant difference in the distribution of automation surprise across aircraft types (χ²(4) = 78.398, p < .001). However, current data suggest that all types of commercial aircraft are susceptible to automation surprise and that such events may negatively impact safe operations. Understandably, as Boeing and Airbus dominate modern commercial fleets, these two manufacturers collectively accounted for 59% of all surprise events. Table 2 Incidence of automated pilot surprise effects by aircraft type Aircraft Manufacturer Aircraft type N N % Cum.% Airbus A300 7 82 30 30 A320 Family 35 A330 14 A340 15 A350 11 Boeing B727 3 86 31 61 B737 Classic 24 B737 NG 19 B757 5 B767 8 B777 16 B787 11 Bombardier CRJ Series 33 57 20 81 Dash 8 Q400 24 Embraer EMB 135/145 17 28 10 91 EMB 170/175/190/195 11 McDonnel Douglass MD-11 13 13 5 96 Others 15 15 5 100 Total 281 281 100 Note: Total percentage may not equal 100% due to rounding of values. Others indicates aircraft of unknown type. Bayesian Network Model Based on the analysis of the Bayesian network model, high-severity outcomes were found to be primarily associated with the flight management system, pilot mental models, and decision-making quality 38 . For example, failures in the FMS may lead to hazardous or high-severity aircraft behavior, such as steep pitch or roll angles or altitude loss. In contrast, low-severity outcomes were often linked to warning systems and monitoring awareness. Specifically, many alerts generated by warning systems turn out to be false alarms, posing minimal risk after verification 39 . These observations are consistent with previous exploratory analyses of ASRS data. Discussion Interaction Failures With ongoing technological advancements, various new automation technologies have been integrated into commercial aircraft. Does higher automation necessarily lead to safer flights? The answer is not always in the affirmative. According to an FAA 40 investigation, approximately one-fourth of aviation accidents or serious incidents were associated with flight crew performance during abnormal situations. In such cases, pilots may be diverted when attempting to understand the abnormal aircraft state and respond to automation surprise, thereby degrading their situational awareness. Pilots make aviation decisions by interacting with the automation interface when current flight conditions deviate from their expectations 41 . For instance, the fatal crashes of Lion Air Flight JT610 (October 2018) and Ethiopian Airlines Flight ET302 (March 2019)—both involving the Boeing 737 MAX—were attributed to unexpected behavior from the Maneuvering Characteristics Augmentation System (MCAS). Investigations revealed that the MCAS's operational logic exceeded pilots' understanding of the 737 series’ flight control system. This mismatch triggered automation surprise, resulted in human-automation interaction failure, and ultimately led to the loss of control 20,29 . These cases highlight the potential safety risks inherent in human-automation interaction. Automatic Surprise Human-Computer Interaction Model Based on abnormal reports and subsequent data analysis, the research group developed a human-machine interaction (HMI) model for automation surprise (see Fig. 7 ). Specifically, pilots with stronger risk perception are more likely to detect abnormalities. A “+” indicates a reinforcing relationship, while a “−” indicates a weakening one. Experienced pilots typically show higher tolerance toward abnormal conditions caused by automation due to their past exposure. Psychological competence refers to the congruence between a pilot’s mental health and adaptability and their role's psychological and occupational requirements 42 . Pilots with good psychological competence are more likely to remain calm and focused under stress, better equipping them to handle automation surprise. In contrast, other pilots may be more prone to emotional reactivity, such as anxiety and agitation. Meanwhile, mindfulness can help pilots regulate their basic psychological needs 17 , reducing their susceptibility to the psychological impact of surprise. When experiencing surprise during flight, if pilots judge the event—based on experience and standard operating procedures—to be a false alarm, no action is required other than monitoring. However, if the event impacts flight safety, partial disengagement of automation and manual control may be necessary. Thus, to maintain safety, it is essential to repeatedly train for surprise scenarios in simulators. Additionally, Crew Resource Management (CRM) training can integrate surprise scenarios to help reduce psychological responses to automation surprise through improved communication and coordination 43 . Importantly, the factors explored in this study do not act in isolation; rather, they interact dynamically. To ensure safe flight operations, pilots must continuously monitor their internal and external environments, collect relevant information, rapidly assess risks, and make timely decisions 8 . Commercial Aircraft Human-Computer Interaction Model In summary, although the original intent of automation design is to enhance efficiency, reduce workload, and improve safety, it has introduced a range of unforeseen operational risks. These risks include the degraded monitoring and situational awareness of pilots as well as their declining manual flying skills. Post-“black swan” phenomena such as the “lumberjack effect,” a dramatic increase in workload, and automation surprise are additional risks. These phenomena are often referred to as the "paradox of automation" 44 , the root cause of which lies largely in the failure of human-machine interaction. Historically, research on pilot error has often been fragmented and reductionist, attributing errors solely to either the technology (automation) or the pilot (human). These attributions do not consider the interactive nature of errors arising from the interplay between pilots and automated systems 45 . This perspective is gradually evolving. In recent years, researchers have adopted top-down, systems-thinking approaches to model factors influencing human-automation interaction and to understand their underlying mechanisms 46,47 . Inspired by such work, and drawing upon the model developed in this study, we propose the human-machine interaction model for commercial aircraft shown in Fig. 8 . This study makes several key contributions. First, the study conducts a multidimensional exploration of automation surprise formation. By combining ASRS data mining with Bayesian network modeling, the study reveals the mechanism behind automation surprise, offering a new perspective on human-machine interaction challenges in aviation automation systems. Second, the study provides a systematic assessment of training impacts. Specifically, this research evaluates how enhanced training methods affect automation surprise, providing scientific guidance for flight training innovations and improving pilots’ capacity to handle unexpected automation behaviors. Third, the study conducts an exploration of explainability in aviation AI. By investigating explainable aviation-focused Large Language Models (LLMs) and their role in mitigating automation surprise, the study contributes to expanding the application of AI in aviation, offering new ideas to improve automation transparency and interpretability. Finally, the study explores how system transparency influences automation surprise, providing important design guidance for user-friendly automation systems and fostering innovation in human-centered interface design. Conclusion By constructing a human-machine interaction model for commercial aircraft, this study identifies and addresses potential risks associated with automation systems, thereby improving flight reliability and operational safety. The development of AI presents new opportunities for the design and optimization of aviation automation systems 48 . Specifically, as AI technology continues to evolve, the autonomy of cockpit automation systems is expected to improve, allowing for dynamic responses independent of pilot intervention. Research on automation surprise can in turn advance relevant technologies and support the development of AI systems with high transparency and explainability 9,31 . Furthermore, this research contributes to exploring AI’s role in aviation, promoting collaborative and cooperative intelligent systems, and ultimately supporting the innovation of trustworthy automation 9,49 . Looking ahead, it is essential to build a trustworthy relationship between humans and autonomous systems, thereby improving automation system design. Focusing on that relationship will help reveal deficiencies in current cockpit design as well as guide the development of future automation systems that are more human-centered 27,48 and more aligned with human cognitive and operational habits 50 . In turn, automation surprise and human-machine conflict will be reduced 51 , mitigating safety risks from a technical standpoint. Research on automation surprise sits at the intersection of aviation engineering, AI, human factors, and engineering psychology. This interdisciplinary effort will cultivate cross-disciplinary professionals with integrated knowledge and skills, fostering both academic integration and industry progress. Declarations Competing interests The authors declare no competing interests. Funding This work was supported by National Social Science Fund of China (19BSH038), and National Natural Science Fund of China (32000753), Author Contribution Lei Du,Xuqun You,And Xinye Wang. contributed to the conception and design the research; contributed to data acquisition;Lei Du ,Ying Li,Tao Wen,Mingliang Li and Yaoliang Wu contributed to data analysis and result interpretation.Lei Du,Qiang Wei and Xinye Wang drafed and revised the manuscript. Saifang Liu,Yuan Li and Ming Ji contributed to manuscript editing .All authors read and approved the final manuscript. Data Availability Data is provided within the manuscript or supplementary information files. References You, X., Ji, M. & Han, H. The Effects of Risk Perception and Flight Experience on Airline Pilots' Locus of Control with Regard to Safety Operation Behaviors. Accident Analysis & Prevention 57 , 131-139 (2013). https://doi.org:10.1016/j.aap.2013.03.036 IATA (International Air Transport Association). Safety Report 2021. Montreal: International Air Transport Association.( 2022). Billings, C. E. Human-Centered Aviation Automation: Principles and Guidelines. openalex (1996). Sarter, N. Investigating Mode Errors on Automated Flight Decks: Illustrating the Problem-Driven, Cumulative, and Interdisciplinary Nature of Human Factors Research. Human Factors 50 , 506-510 (2008). https://doi.org:10.1518/001872008x312233 Sarter, N. B., Mumaw, R. J. & Wickens, C. D. Pilots' Monitoring Strategies and Performance on Automated Flight Decks: An Empirical Study Combining Behavioral and Eye-Tracking Data. Human Factors 49 , 347.340-357.340 (2007). https://doi.org:10.1518/001872007x196685 Dekker, S. W. A. & Woods, D. D. MABA-MABA or Abracadabra? Progress on Human-Automation Co-ordination. Cognition, Technology & Work 4 (2002). https://doi.org:10.1007/s101110200022 Naweed, A. & Bye, R. Joint Cognitive Systems: Patterns in Cognitive Systems Engineering. Ergonomics 51 , 768-770 (2008). https://doi.org:10.1080/00140130701223774 Xinye, W., Yuan, L., Ming, C. & Xuqun, Y. The Detriments and Improvement of Automation Trust and Dependence to Aviation Safety. Advances in Psychological Science 25 , 1614-1614 (2017). https://doi.org:10.3724/sp.j.1042.2017.01614 European Aviation Safety Agency. (2020). Artificial Intelligence Roadmap: A human-centric approach to AI in aviation (Version 1.0). Cologne, G. E. A. S. A. ]Federal Aviation Administration. Operational use of flight path management systems: Final report of the performance-based Operations Aviation Rulemaking Commitee/Commercial Aviation Safety Team Flight Deck Automation Working Group. Washington: Author.(2013). Moriarty, D. Practical human factors for pilots. London, UK: Elsevier.(2015). Horstmann, G. Latency and Duration of the Action Interruption in Surprise. Cognition & emotion 20 , 242-273 (2006). https://doi.org:10.1080/02699930500262878 International Air Transport Association. (2019). Loss of Control In-Flight Accident Analysis Report Edition 2019. Retrieved September, 2021. from https://www.iata.org/contentassets/b6eb2adc248c484192101edd1ed36015/loc-i 2019.pdf . Boer, R. J. d. & Hurts, K. Automation Surprise. Aviation Psychology and Applied Human Factors 7 , 28-41 (2017). https://doi.org:10.1027/2192-0923/a000113 Trippe, J. & Mauro, R. Understanding Automation Surprise: Analysis of ASRS Reports. 18th International Symposium on Aviation Psychology (2015). Ji, M., You, X., Lan, J. & Yang, S. The Impact of Risk Tolerance, Risk Perception and Hazardous Attitude on Safety Operation among Airline Pilots in China. Safety Science 49 , 1412-1420 (2011). https://doi.org:10.1016/j.ssci.2011.06.007 Ji, M., Yang, C., Li, Y., Xu, Q. & He, R. The Influence of Trait Mindfulness on Incident Involvement among Chinese Airline Pilots: the Role of Risk Perception and Flight Experience. Journal of safety research 66 , 161-168 (2018). https://doi.org:10.1016/j.jsr.2018.07.005 Dekker, S. The Field Guide to Human Error Investigations. msra (2017). https://doi.org:10.4324/9781315202778 Xinye, W., Shu, L., Wenhao, C., Yangmei, L. & Xuqun, Y. Management of Startle and Surprise in Flight: Extended Landman Model. Journal of Psychological Science 44 (2021). https://doi.org:10.16719/j.cnki.1671-6981.20210621 European Aviation Safety Agency. (2018). Startle effect management. Report number: NLR-CR-2018-242. Cologne, G. E. A. S. A. Chen, Y. Y., Fang, W.N., Guo, B.Y., & Bao, H.F. (2023). Effect of task interruption on task performance and moderating effect of mental fatigue. Acta Psychologica Sinica, 55(1), 22-35. Baxter, G., Besnard, D. & Riley, D. Cognitive Mismatches in the Cockpit: Will They Ever Be A Thing of the Past? Applied ergonomics 38 (2007). https://doi.org:10.1016/j.apergo.2007.01.005 Ong, D. C., Zaki, J. & Goodman, N. D. Computational Models of Emotion Inference in Theory of Mind: A Review and Roadmap. Topics in Cognitive Science 11 , 338-357 (2019). https://doi.org:10.1111/tops.12371 Landman, A., Groen, E. L., Paassen, M. M. v., Bronkhorst, A. W. & Mulder, M. The Influence of Surprise on Upset Recovery Performance in Airline Pilots. International Journal of Aerospace Psychology 27 (2017). https://doi.org:10.1080/10508414.2017.1365610 Wen, H. et al. A Methodology to Assess Human-Automated System Conflict from Safety Perspective. Computers & Chemical Engineering 165 , 107939-107939 (2022). https://doi.org:10.1016/j.compchemeng.2022.107939 woods, d. d. & hollnagel, e. Joint cognitive systems : patterns in cognitive systems engineering. mag (2006). Grindley, B. et al. Avoiding Automation Surprise: Identifying Requirements to Support Pilot Intervention in Automated Uncrewed Aerial Vehicle (UAV) Flight. Applied Ergonomics 127 (2025). https://doi.org:10.1016/j.apergo.2025.104516 Ancel, E. & Shih, A. Modeling Increased Complexity and the Reliance on Automation: FLightdeck Automation Problems (FLAP) Model. semanticscholar Wang, X.Y., Wang, S., Lin, J. K., & Du, L. Causative analysis of two Boeing 737 MAX crashes: FLAP model based on human-automation interaction. Applied Psychology, 30, 483-494 (2024). https://doi.org:10.20058/j.cnki.CJAP.023101 European Aviation Safety Agency. Artificial Intelligence Roadmap: A human-centric approach to AI in aviation (Version 1.0). Cologne, Germany: European Aviation Safety Agency, (2020). Lyons, J. B., Clark, M. A., Wagner, A. R. & Schuelke, M. J. Certifiable Trust in Autonomous Systems: Making the Intractable Tangible. AI Magazine 38 (2017). https://doi.org:10.1609/aimag.v38i3.2717 Farjadian, A. B., Thomsen, B., Annaswamy, A. M. & Woods, D. D. Resilient Flight Control: an Architecture for Human Supervision of Automation. IEEE Transactions on Control Systems Technology 29 , 29-42 (2020). https://doi.org:10.1109/tcst.2019.2959542 Kinney, L. & O'Hare, D. Responding to an Unexpected In-Flight Event: Physiological Arousal, Information Processing, and Performance. Human Factors 62 (2020). https://doi.org:10.1177/0018720819854830 Liu, Y., Pan, Y. D., Li, M., Li, C. L., Wang, X.Y., & You, X.Q. Applications and prospects of near infrared spectroscopy in aeronautical psychology. Psychological Science 46(6) , 1518-1528. (2023). Li, R., Hosseini, H., Saggar, M., Balters, S. C. & Reiss, A. L. Current Opinions on the Present and Future Use of Functional Near-Infrared Spectroscopy in Psychiatry. Neurophotonics 10 (2023). https://doi.org:10.1117/1.nph.10.1.013505 Ji, M., Xie, X.D, & Qiu, Y . Application of eye tracking technology in cockpit ergonomics design. Packaging Engineering 42(18) , 84-93. (2021). Parasuraman, R. Neuroergonomics: Brain, Cognition, and Performance at Work. CURRENT DIRECTIONS IN PSYCHOLOGICAL SCIENCE 20 (2011). https://doi.org:10.1177/0963721411409176 Zhang, X. & Mahadevan, S. Bayesian Network Modeling of Accident Investigation Reports for Aviation Safety Assessment. Reliability Engineering & System Safety 209 , 107371-107371 (2020). https://doi.org:10.1016/j.ress.2020.107371 Wang, L., Zhang, J., Dong, C., Sun, H. & Ren, Y. A Method of Applying Flight Data to Evaluate Landing Operation Performance. Ergonomics 62 , 171-180 (2018). https://doi.org:10.1080/00140139.2018.1502806 FAA. Operational use of flight path management systems: Final report of the performance-based Operations Aviation Rulemaking Commitee/Commercial Aviation Safety Team Flight Deck Automation Working Group. Washington: Author. (2013). Wang, Z. & You, X. Models and influential factors of aeronautical decision making. Advances in Psychological Science 25 (2017). https://doi.org:10.3724/sp.J.1042.2017.02002 WANG, Y. & ZHAO, R. Construction of pilots' psychological competency model based on personality characteristics. Journal of Civil Aviation University of China 36 , 46-51 (2018). https://doi.org:10.3969/j.issn.1674-5590.2018.03.010 Macleod, N. Crew Resource Management Training: A Competence-based Approach for Airline Pilots. openalex (2021). Faros, B. The Paradox of Automation. Journal of Computing Sciences in Colleges (2016). https://doi.org:10.5555/2904298.2904320 Maurino, D. E. Human factors and aviation safety: What the in- dustry has, what the industry needs. . Ergonomics 43 , 952–959. (2000). Endsley, M. R. From here to autonomy: Lessons learned from human-automation research. Human Factors, 59 (1) , 5-27. (2017). Stowers, K. et al. A Framework to Guide the Assessment of Human–Machine Systems. Human Factors 59 , 172-188 (2017). https://doi.org:10.1177/0018720817695077 Wei, X. Seven essays on user-centered design: from automation to intelligent aircraft cockpits. Applied Psychology. 28(4) , 291-313. (2022). Dong Wenli, Fang Weining, Trust in Automation:Research Review and Future Perspectives. Zidonghua Xuebao/Acta Automatica Sinica , 1183-1200 (2021). https://doi.org:10.16383/j.aas.c200432 Xu Wei, G. L., Gao Zaifeng. Human-AI interaction: a new interdisciplinary field to realize the concept of human-centered AI. Journal of Intelligent Systems 16(4) , 605-621. (2021). Xu, W., Dainoff, M. J., Ge, L. & Gao, Z. Transitioning to Human Interaction with AI Systems: New Challenges and Opportunities for HCI Professionals to Enable Human-Centered AI. International Journal of Human-computer Interaction 39 (2023). https://doi.org:10.1080/10447318.2022.2041900 Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7171123","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":496914724,"identity":"efc7dad2-816d-4c61-aea5-1c2f2ca2859d","order_by":0,"name":"Lei Du","email":"","orcid":"","institution":"Shaanxi Normal University","correspondingAuthor":false,"prefix":"","firstName":"Lei","middleName":"","lastName":"Du","suffix":""},{"id":496914725,"identity":"b52b2a64-f6aa-44e7-aa65-953cda3f4c3c","order_by":1,"name":"Ying Li","email":"","orcid":"","institution":"Shaanxi Normal University","correspondingAuthor":false,"prefix":"","firstName":"Ying","middleName":"","lastName":"Li","suffix":""},{"id":496914726,"identity":"a620c159-5883-4327-954a-96e20a7d3afa","order_by":2,"name":"Xingye Wang","email":"","orcid":"","institution":"Shaanxi Normal University","correspondingAuthor":false,"prefix":"","firstName":"Xingye","middleName":"","lastName":"Wang","suffix":""},{"id":496914727,"identity":"6c791e5f-ff5d-42b2-b36d-3eb356bf19a7","order_by":3,"name":"Yangliang Wu","email":"","orcid":"","institution":"Shaanxi Normal University","correspondingAuthor":false,"prefix":"","firstName":"Yangliang","middleName":"","lastName":"Wu","suffix":""},{"id":496914728,"identity":"6b90c1d8-d4b9-4fb3-9474-b442266c8cb1","order_by":4,"name":"Mingliang Li","email":"","orcid":"","institution":"Shaanxi Normal University","correspondingAuthor":false,"prefix":"","firstName":"Mingliang","middleName":"","lastName":"Li","suffix":""},{"id":496914729,"identity":"61c2ad47-ae0a-4b65-aca0-d4077d4388b4","order_by":5,"name":"Tao Wen","email":"","orcid":"","institution":"Tibet Airlines Co., Ltd.","correspondingAuthor":false,"prefix":"","firstName":"Tao","middleName":"","lastName":"Wen","suffix":""},{"id":496914730,"identity":"64df31e6-9908-49e8-9039-0018f321dd53","order_by":6,"name":"Qiang Wei","email":"","orcid":"","institution":"Tibet Airlines Co., Ltd.","correspondingAuthor":false,"prefix":"","firstName":"Qiang","middleName":"","lastName":"Wei","suffix":""},{"id":496914731,"identity":"c855a424-b1d3-49ac-b6d0-abdb39f915c1","order_by":7,"name":"Saifang Liu","email":"","orcid":"","institution":"Shaanxi Normal University","correspondingAuthor":false,"prefix":"","firstName":"Saifang","middleName":"","lastName":"Liu","suffix":""},{"id":496914732,"identity":"1fe7cc38-3131-4d59-b585-84774ed92b6d","order_by":8,"name":"Ming Ji","email":"","orcid":"","institution":"Shaanxi Normal University","correspondingAuthor":false,"prefix":"","firstName":"Ming","middleName":"","lastName":"Ji","suffix":""},{"id":496914733,"identity":"6d6f2959-c841-447b-8a58-ac52659b0008","order_by":9,"name":"Yuan Li","email":"","orcid":"","institution":"Shaanxi Normal University","correspondingAuthor":false,"prefix":"","firstName":"Yuan","middleName":"","lastName":"Li","suffix":""},{"id":496914734,"identity":"642e9f0e-52d9-4caf-be42-76d9a34ae6b8","order_by":10,"name":"Xuqun You","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA2UlEQVRIiWNgGAWjYPCDCgk5eRK1nLEwNmwgSQdjW0UiwwECigxu5B68zbvjsJzB8cMPP/ycJ5HA2MD88NENPFokZ+QlW/OeOWxscCbNWLJ3m0QeOwObsXEOHi38Ejlm0rxthxM33GAwkGbcJlHM2MDDJo1PCxtCC/vn34xzJBIbDhDQgmQLj5k0YwMRWiR73hhbzm1LN5Y8k1Nm2XNMwtiwmYBfDI7nGN5422Ytx3f8+OYbP2rq5OTZmx8+xqcFBCQYGJqRuMwElEO11BGhbBSMglEwCkYsAABfREeiPjnHKAAAAABJRU5ErkJggg==","orcid":"","institution":"Shaanxi Normal University","correspondingAuthor":true,"prefix":"","firstName":"Xuqun","middleName":"","lastName":"You","suffix":""}],"badges":[],"createdAt":"2025-07-20 17:23:08","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7171123/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7171123/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":88900753,"identity":"23042382-baa9-4858-be7e-e92197ffde46","added_by":"auto","created_at":"2025-08-12 13:41:24","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":116514,"visible":true,"origin":"","legend":"\u003cp\u003eA320 cockpit control and display system\u003c/p\u003e\n\u003cp\u003eNote: MCP = mode control panel; PFD = primary flight display; ND= navigation display; MFD = multi-function display; CDU = control display unit\u003c/p\u003e","description":"","filename":"image1.png","url":"https://assets-eu.researchsquare.com/files/rs-7171123/v1/b5eadd410a92aab52bcac03a.png"},{"id":88901552,"identity":"cf6be6b6-6156-4fe6-aa73-bf3ba5b9000d","added_by":"auto","created_at":"2025-08-12 13:49:24","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":319391,"visible":true,"origin":"","legend":"\u003cp\u003eInput interfaces of flight control automation system and operation logic between systems\u003c/p\u003e\n\u003cp\u003eNote: MCP= mode control panel; LNAV = Lateral Navigation; VNAV = Vertical Navigation; AP = Autopilot; AT=Autothrottle; FMS = flight management system.\u003c/p\u003e","description":"","filename":"image2.png","url":"https://assets-eu.researchsquare.com/files/rs-7171123/v1/b2f0c5f6c057b9b29db1ef75.png"},{"id":88901554,"identity":"fa4fdfe1-2a16-4371-869e-0cbea0a430f5","added_by":"auto","created_at":"2025-08-12 13:49:24","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":26044,"visible":true,"origin":"","legend":"\u003cp\u003eMatching mental models with automated behavior\u003c/p\u003e\n\u003cp\u003eNote: Horizontal double arrows indicate that mental models match automated behavior and no automated surprises occur. Vertical dashed double arrows indicate a mismatch between mental models and automatic behavior, resulting in automation surprise.\u003c/p\u003e","description":"","filename":"image3.png","url":"https://assets-eu.researchsquare.com/files/rs-7171123/v1/d6bbda09a2ae9d6d4d2ec779.png"},{"id":88900763,"identity":"518498e6-bed7-462e-8786-7452fe222ce0","added_by":"auto","created_at":"2025-08-12 13:41:24","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":71642,"visible":true,"origin":"","legend":"\u003cp\u003eThe mechanism of automation surprises in FLAP model.\u003c/p\u003e","description":"","filename":"image4.png","url":"https://assets-eu.researchsquare.com/files/rs-7171123/v1/a5a8e023da1d2e467373b9dc.png"},{"id":88900759,"identity":"2ad69f3f-a606-4e69-9780-65b9897cc4d7","added_by":"auto","created_at":"2025-08-12 13:41:24","extension":"jpeg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":147754,"visible":true,"origin":"","legend":"\u003cp\u003eSearch criteria for automated surprise events in ASRS\u003c/p\u003e","description":"","filename":"image5.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7171123/v1/1ebf4101239e1a0743ba462b.jpeg"},{"id":88902766,"identity":"c9ae5176-c578-4d57-8e18-7c29c7e0a604","added_by":"auto","created_at":"2025-08-12 13:57:24","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":245738,"visible":true,"origin":"","legend":"\u003cp\u003eBayesian network model of severity formation mechanism of automation surprise results\u003c/p\u003e\n\u003cp\u003eNote: W = Warning and Alarm System, I = Electronic Flight Instrument System, F = Flight Management System; Graphs for Severity of Results\u003c/p\u003e\n\u003cp\u003eAccording to Table 1, the higher the value, the greater the severity.\u003c/p\u003e","description":"","filename":"image6.png","url":"https://assets-eu.researchsquare.com/files/rs-7171123/v1/b823e05f8aae256be0b8ace0.png"},{"id":88901556,"identity":"2304ed18-e0af-4db1-8d15-6087bc3a6326","added_by":"auto","created_at":"2025-08-12 13:49:24","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":52339,"visible":true,"origin":"","legend":"\u003cp\u003eAutomation surprise human-computer interaction model\u003c/p\u003e","description":"","filename":"image7.png","url":"https://assets-eu.researchsquare.com/files/rs-7171123/v1/42fe24a50bef69bd52037d19.png"},{"id":88900765,"identity":"bf4e668a-2d90-450d-9c03-8687843b0c7b","added_by":"auto","created_at":"2025-08-12 13:41:24","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":102827,"visible":true,"origin":"","legend":"\u003cp\u003eHuman-computer interaction model of commercial aircraft\u003c/p\u003e\n\u003cp\u003eASRS = Aviation Safety Reporting System; FLAP = Aviation Safety Autonomous Reporting System\u003c/p\u003e\n\u003cp\u003eFLightdeck Automation Problems Flight cockpit automation problem model; W = warning and alarm system, I = electrical Sub-flight instrument system, F = flight management system; s = scenario simulation training, k = theoretical knowledge training, n = extraTraining; AT = Autothrottle; LLM = Large Language Model.\u003c/p\u003e","description":"","filename":"image8.png","url":"https://assets-eu.researchsquare.com/files/rs-7171123/v1/9b598513de587c35d3fe2c4f.png"},{"id":91303594,"identity":"56eb9f5d-dc0d-42bf-b015-33b7c71c9dda","added_by":"auto","created_at":"2025-09-15 06:17:02","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1635034,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7171123/v1/b6638346-d1b6-410e-a9e7-2a8e1423e9fb.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"The Impact of Automation Systems on Commercial Aircraft: A Human-Machine Interaction Perspective","fulltext":[{"header":"Introduction","content":"\u003cp\u003eIn the field of aviation, safety has always been a major concern. With substantial advancements in automation and aircraft manufacturing technologies, modern aircraft and avionics have become increasingly sophisticated and reliable. As a result, pilots have transitioned from direct operators of aircraft to comprehensive system managers. This shift in pilot role has likely introduced new error patterns\u003csup\u003e1\u003c/sup\u003e. According to the International Air Transport Association's safety report on the causes of aviation accidents from 2017 to 2021, manual operation or flight control accounted for 33% of crew-related errors, ranking first among human factors \u003csup\u003e2\u003c/sup\u003e. The transformation of pilot responsibilities that has brought new error patterns poses additional training requirements\u003csup\u003e3-5\u003c/sup\u003e. In addition, it has increased cognitive and attentional demands, thereby prolonging the time pilots need to detect and correct errors\u003csup\u003e6\u003c/sup\u003e.When an automated system does not perform as expected or fails to operate entirely, an automation surprise may occur in which pilots are surprised by the system's actual behavior\u003csup\u003e7\u003c/sup\u003e. Due to the increasing role of automation management in pilot duties, pilots tend to develop trust in and reliance on automation through prolonged interaction, leading to a degradation of manual flying skills and the emergence of automation surprises\u003csup\u003e8\u003c/sup\u003e. While previous research has identified some factors influencing automation surprise, the specific mechanisms of automation surprise formation in commercial aviation and the associated human-machine interaction (HMI)remain underexplored.\u003c/p\u003e\n\u003cp\u003eWith the rapid development of artificial intelligence (AI), AI-enabled automation is expected to evolve from assisting pilots to collaborating with them, and eventually to operating autonomously\u003csup\u003e9\u003c/sup\u003e. In this transition, AI-driven automation systems must be transparent, presenting essential information to help pilots understand system behaviors. Large Language Models (LLMs), trained on aviation-specific knowledge, can serve to communicate and explain system behavior, thereby aligning intentions between pilots and automated systems. This approach helps eliminate the occurrence of automation surprises from a design standpoint and reduces their likelihood through improved human-machine interaction.\u003c/p\u003e"},{"header":"Introduction","content":"\u003cp\u003eMajor Automation Equipment:\u003c/p\u003e\n\u003cp\u003eAccording to a report by the Federal Aviation Administration \u003csup\u003e10\u003c/sup\u003e, Modern commercial aircraft cockpits are commonly equipped with the following automation systems:\u003c/p\u003e\n\u003col\u003e\n\u003cli\u003e\n\u003cp\u003eA powerful Flight Management System (FMS), also known as a Flight Management Computer (FMC), which is used to enhance flight path management;\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eAdvanced navigation systems, such as Area Navigation (RNAV), Required Navigation Performance (RNP), and the Global Positioning System (GPS);\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eMore sophisticated digital display systems integrated into cockpit components, such as the Primary Flight Display (PFD), Navigation Display (ND), Mode Control Panel (MCP), Multi-Function Display (MFD), and Control Display Unit (CDU).\u003c/p\u003e\n\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003eThese automation systems are illustrated in Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eMain Functions of Automation Implementation:\u003c/p\u003e\n\u003cp\u003ePilots typically configure combinations of these automation systems to achieve the desired flight performance. The configurations are generally standardized and exhibit inherent logical relationships among the systems (see Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e). For example, Moriarty \u003csup\u003e11\u003c/sup\u003eidentified several high-level automated flight control modes, including the following:\u003c/p\u003e\n\u003col\u003e\n\u003cli\u003e\n\u003cp\u003eAutopilot (AP) or Autothrottle (AT) engagement, combined with manual selection of Heading (HDG), Very High Frequency Omnidirectional Range (VOR), or Localizer (LOC) navigation modes\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eAP or AT engagement in conjunction with FMS guidance, in which case the aircraft can follow a predefined route with minimal pilot intervention—typically corresponding to the cruise phase of a flight.\u003c/p\u003e\n\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003eHowever,the benefits of these automation systems are not always easily realized. The introduction of complex automated systems requires pilots to spend a significant amount of time monitoring them to ensure timely detection and resolution of anomalies. This shift in pilot responsibilities has introduced new patterns of error and increased cognitive, attentional, and training demands \u003csup\u003e3–5\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eIn this study, we manually mined case data from NASA's Aviation Safety Reporting System (ASRS). Elements corresponding to the automation surprise mechanism in the FLight deck Automation Problems (FLAP) model were manually labeled and categorized. The FLAP model examines not only the interaction of humans and automated systems when they are functioning properly, but also includes crew defects, and specifically examines scenarios such as automated system failures and avionics system anomalies.Based on this categorization, a Bayesian Network was constructed to model the formation mechanism of automation surprise. The backpropagation algorithm was then applied to identify the key factors leading to severe consequences and the most influential contributors to automation surprise. A human-machine interaction model for commercial aircraft was ultimately developed. This model helps bridge the gap between the severity of automation incidents and related aspects such as flight management systems, mental models, and decision-making coordination from the perspective of human-machine interaction.\u003c/p\u003e\n\u003cp\u003eAutomation Surprise:\u003c/p\u003e\n\u003cp\u003eSurprise is typically a reaction to events that deviate from prior expectations \u003csup\u003e12\u003c/sup\u003e. Automation surprise, in particular, thus occurs when a malfunction in an automated system leads to a mode transition, or when an inappropriate automation mode is activated by the pilot, causing the system to disengage without explicit pilot command. In such scenarios, pilots may struggle to comprehend the automated system's behavior, find it difficult to regain control of the aircraft, and, in extreme cases, lose control in flight (LOC-I)\u003csup\u003e13\u003c/sup\u003e .\u003c/p\u003e\n\u003cp\u003eNumerous studies, particularly those based on questionnaire surveys, have identified various direct and indirect factors influencing automation surprise\u003csup\u003e14\u003c/sup\u003e. Trippe and Mauro \u003csup\u003e15\u003c/sup\u003ecategorized the contributing factors to automation surprise into three broad domains:\u003c/p\u003e\n\u003col\u003e\n\u003cli\u003e\n\u003cp\u003eAutomation system factors: level of automation, system failure, system design\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003ePilot-related factors: flight experience, risk perception, mindfulness level\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eInterface factors: false or unclear display of information\u003c/p\u003e\n\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003eThe occurrence of automation surprise is closely related to pilot characteristics such as experience, risk perception, and mindfulness\u003csup\u003e16,17\u003c/sup\u003e. Pilots respond to automation surprise based on their knowledge of aircraft automation and operational experience.\u003c/p\u003e\n\u003cp\u003eThreat to Aviation Safety:\u003c/p\u003e\n\u003cp\u003eDe Boer and Hurts\u003csup\u003e14\u003c/sup\u003eclassified the outcomes of automation surprise events into different severity levels (as shown in Table\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e). They found that most events fell into levels 1–3, with only a few rated at levels 4–5. Moreover, most incidents did not result in aircraft damage, and very few involved physical injury or fatalities among crew or passengers\u003csup\u003e15\u003c/sup\u003e.\u003c/p\u003e\n\u003cdiv class=\"gridtable\"\u003e\n\u003cdiv class=\"colspec\" align=\"left\"\u003e\u0026nbsp;\u003c/div\u003e\n\u003cdiv class=\"colspec\" align=\"left\"\u003e\u0026nbsp;\u003c/div\u003e\n\u003ctable id=\"Tab1\" border=\"1\"\u003e\u003ccaption\u003e\n\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\n\u003cdiv class=\"CaptionContent\"\u003e\n\u003cp\u003eSeverity classification of automation surprise consequences\u003c/p\u003e\n\u003c/div\u003e\n\u003c/caption\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\"\u003e\n\u003cp\u003eseverity rating\u003c/p\u003e\n\u003c/th\u003e\u003cth align=\"left\"\u003e\n\u003cp\u003eType of consequence\u003c/p\u003e\n\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003e1\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003eNo result\u003c/p\u003e\n\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003e2\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003eReduced automation\u003c/p\u003e\n\u003cp\u003eReuse automated deviation from FMS program after calibration\u003c/p\u003e\n\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003e3\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003eManual takeover/correction\u003c/p\u003e\n\u003cp\u003ePrepare to land and resume flight\u003c/p\u003e\n\u003cp\u003eSignificant increase in vectoring/air control assistance workloads\u003c/p\u003e\n\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003e4\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003eSpeed deviation is significant\u003c/p\u003e\n\u003cp\u003eSerious deviation from course/path\u003c/p\u003e\n\u003cp\u003eSerious deviation from altitude\u003c/p\u003e\n\u003cp\u003eReporting of security reports/technical logs\u003c/p\u003e\n\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003e5\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003eUnstable flight\u003c/p\u003e\n\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003e6\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003eOrganic damage\u003c/p\u003e\n\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003ctfoot\u003e\u003ctr\u003e\u003ctd colspan=\"2\"\u003eNote: The higher the severity level, the higher the hazard to aviation safety.\u003c/td\u003e\u003c/tr\u003e\u003c/tfoot\u003e\u003c/table\u003e\n\u003c/div\u003e\n\u003cp\u003eNonetheless, from an aviation safety perspective, automation surprise represents at least a potential threat. Dekker\u003csup\u003e18\u003c/sup\u003eemphasized that automation surprise becomes evident only when pilots notice the system behaving in an unexpected manner—but by that point, significant consequences may have already occurred. Automation surprise is considered a specific manifestation of the broader \"surprise effect\" \u003csup\u003e19\u003c/sup\u003e. Such events can narrow a pilot’s attention, increase their cognitive workload, reduce their situational awareness, and even temporarily compromise their control of the aircraft\u003csup\u003e20\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eFurthermore, handling automation surprises can interrupt a pilot's ongoing tasks, reducing the attention allocated to the primary task. Furthermore, irrelevant information introduced by the interruption may interfere with working memory. Fatigue exacerbates these effects, negatively impacting attention, memory, and task performance\u003csup\u003e21\u003c/sup\u003e. As a result, automation surprise can significantly impair a pilot’s ability to effectively operate the aircraft.\u003c/p\u003e\n\u003cp\u003eIn addition, automation surprise may disrupt air traffic operations, as controllers need to redirect other aircraft to accommodate the affected crew's handling of unexpected aircraft behavior\u003csup\u003e15\u003c/sup\u003e. In summary, automation surprise remains an issue warranting ongoing attention \u003csup\u003e22\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eMan-Machine Conflicts:\u003c/p\u003e\n\u003cp\u003eMental models are typically developed through training and operational experience. These models are used to reason about and control the underlying processes of socio-technical systems \u003csup\u003e23\u003c/sup\u003e. In aviation, a cognitive mismatch arises when a pilot’s mental model of the automation does not accurately reflect the system’s actual behavior \u003csup\u003e19,22,24\u003c/sup\u003e. Cognitive mismatches can take various forms, such as mode confusion, in which pilots believe the automation is operating in one mode when it is actually functioning in another\u003csup\u003e25\u003c/sup\u003e. Under such conditions, automation surprises occur\u003csup\u003e26,27\u003c/sup\u003e. Correspondingly, Dekker\u003csup\u003e18\u003c/sup\u003edefines automation surprise as a mismatch between the operator’s mental model (e.g., expectations or intentions) and the actual behavior of the automation (see Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e). From this perspective, automation surprise is closely related to human-machine conflict.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAutomated FLAP Model:\u003c/p\u003e\n\u003cp\u003eAncel and Shih \u003csup\u003e28\u003c/sup\u003eproposed the FLAP model based on automation-related aviation accidents and the human-machine interaction literature. The FLAP model provides a top-down, systematic framework for simulating automation-related problems in commercial flight operations. For example, Wang et al.\u003csup\u003e29\u003c/sup\u003eemployed the FLAP model to successfully trace the root causes of two Boeing 737 MAX crash accidents. The model includes a mechanism of automation surprise (AS), as shown in Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003e. When pilots lack system-related knowledge and a comprehensive mental model of automation, automated behavior may be poorly understood and automation surprises are more likely to occur\u003csup\u003e4\u003c/sup\u003e. Thus, inadequate decision making and degraded monitoring awareness may follow. Reduced situational monitoring can further lead to decision failures. Unexpected automation behaviors may become the source of crew confusion or judgment errors, ultimately resulting in automation surprise\u003csup\u003e27\u003c/sup\u003e.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAs discussed above, most research on automation surprise has focused on its negative impact on aviation safety. In particular, much attention has been paid to the factors contributing to automation surprise. However, apart from the FLAP model proposed by Ancel and Shih\u003csup\u003e28\u003c/sup\u003e, few studies have systematically examined the formation mechanism of automation surprise and its implications for safety. Furthermore, the aviation accident datasets used to construct and validate the FLAP model remain very limited. Therefore, it is essential to mine additional aviation case data to validate the automation surprise formation mechanism illustrated in Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003e.\u003c/p\u003e\n\u003cp\u003eAutomated AI Enabling Psychology:\u003c/p\u003e\n\u003cp\u003eWith the rapid development of artificial intelligence (AI), automation is expected to evolve from assisting pilots, to collaborating with them, and, ultimately, to operating autonomously \u003csup\u003e30\u003c/sup\u003e. During this transition, AI-enabled automation must exhibit greater transparency, providing pilots with the essential information needed to understand system behaviors. Research into automation surprise can also help advance intelligent system technologies, particularly in the development of AI systems with high transparency and explainability\u003csup\u003e30,31\u003c/sup\u003e. Large Language Models (LLMs), trained on aviation-specific knowledge, can interpret and explain automation behavior to help align pilot and system intentions\u003csup\u003e32\u003c/sup\u003e. This design approach can significantly reduce the likelihood of automation surprises from the outset.\u003c/p\u003e\n\u003cp\u003eBuilding on conceptual models of surprise and startle during flight, some researchers have investigated pilots’ emotional responses to unexpected engine failures using physiological indicators such as heart rate, skin conductance, and eye tracking \u003csup\u003e24,33\u003c/sup\u003e. However, few studies have explored the neural activity of pilots during automation surprise events. Liu et al. \u003csup\u003e34\u003c/sup\u003e highlighted the advantages of functional near-infrared spectroscopy (fNIRS) in aviation psychology research. fNIRS allows for continuous, non-invasive, and portable measurement of hemodynamic changes related to brain function\u003csup\u003e35\u003c/sup\u003e, enabling inferences about a pilot’s cognitive state during flight.\u003c/p\u003e\n\u003cp\u003eThus, fNIRS enables the observation of neural responses to automation surprise and the measurement of mental workload. Additionally, eye-tracking technology has already seen widespread use in cockpit human factors design\u003csup\u003e36\u003c/sup\u003e. Future studies may adopt neuroergonomics approaches by combining near-infrared and eye-tracking technologies in full-flight simulators\u003csup\u003e37\u003c/sup\u003e. The combination will allow researchers to examine the neural impacts of automation surprise on pilots and to assess the effectiveness of enhanced training and AI-empowered system design in mitigating automation surprise.\u003c/p\u003e\n\n\n\n\n\n\n\n\n\n\n\n\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\n\n\n\n"},{"header":"Data and Methods","content":"\u003cp\u003eAutomation Surprise Formation Mechanism Based on ASRS Data Mining\u003c/p\u003e\u003cp\u003eWe constructed a Bayesian network based on case reports related to automation surprise from the Aviation Safety Reporting System (ASRS) database. the Bayesian network has an important advantage: when additional information (e.g., data, evidence) on some random variables is available, the Bayesian network enables real-time updating of system state predictions through forward-looking forecasts and backward-tracing analysis. The network allows us to gain deeper insight into the mechanisms underlying automation surprise, identify the key factors contributing to severe consequences, and distinguish between different types of automation systems (e.g., warning and alerting systems, electronic flight instrument systems, and flight management systems). Typical automation surprise scenarios associated with each system type were summarized to inform model development.\u003c/p\u003e\u003cp\u003eCase Data Filtering and Marking\u003c/p\u003e\u003cp\u003eThe data for this study were retrieved from the ASRS database (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://asrs.arc.nasa.gov/search/database.html\u003c/span\u003e\u003c/span\u003e). The criteria for the search, which was conducted on March 1, 2022, are shown in Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003e. The selected time window spanned from January 2010 to December 2021. The operational type was restricted to Part 121 operations, which sets the standard for commercial aircraft opration. The human factor category was limited to human-machine interaction. The text query used root terms related to surprise: “surpri%”, “freez%”, “perplex%”, “confuse%”, “shock%”, and “unexpect%”, in which “%” includes all word derivatives to ensure comprehensive retrieval. The direct search yielded 305 reports, which were then downloaded for further processing.\u003c/p\u003e\u003cp\u003e\u0026nbsp;\u003c/p\u003e\u003cp\u003eSubsequently, a five-member research team—consisting of three doctoral students in aviation psychology (each of whom is also a captain of a commercial aircraft), one avionics engineer, and one human factors expert—individually read and discussed each incident report. Incidents that mentioned automation, autopilot, autothrottle, flight management system (FMS), flight management computer (FMC), mode control panel (MCP), lateral navigation (LNAV), vertical navigation (VNAV), or their abbreviations were retained. Information related to air traffic controllers (ATCs), passengers, cabin crew, dispatchers, maintenance personnel, ground handlers, or other unrelated individuals was excluded. In total, 281 automation surprise events were identified for further analysis.\u003c/p\u003e\u003cp\u003eThe five-member team then labeled each case based on factors associated with automation surprise in the FLAP model, such as the following:\u003c/p\u003e\u003cul\u003e\n\u003cli\u003e\n\u003cp\u003eThe specific automation system involved in the incident\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eWhether the pilot's mental model of the system was sufficient or lacking\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eWhether the pilot’s monitoring awareness was high or low prior to the incident\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eWhether the decision-making quality was high or low\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eThe severity of the consequences after the surprise occurred\u003c/p\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\u003cp\u003eThis labeling process prepared the dataset for the subsequent Bayesian network modeling.\u003c/p\u003e\u003cp\u003eBayesian Network Model Construction\u003c/p\u003e\u003cp\u003eBased on the labeled case data and the automation surprise formation mechanism in the FLAP model, a data-driven Bayesian network was constructed\u003csup\u003e38\u003c/sup\u003e (see Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e6\u003c/span\u003e). The backpropagation algorithm was then applied to identify which factors or combinations of factors contributed to different levels of outcome severity.\u003c/p\u003e"},{"header":"Results Analysis","content":"\u003cp\u003eDistribution of Surprise Events\u003c/p\u003e\u003cp\u003eThe distribution of pilot-reported surprise events by manufacturer was as follows: Airbus (29%), Boeing (30%), Bombardier (20%), Embraer (10%), and McDonnell Douglas (5%) (see Table\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e). A chi-square test indicated a significant difference in the distribution of automation surprise across aircraft types (χ²(4) = 78.398, \u003cem\u003ep\u003c/em\u003e \u0026lt; .001). However, current data suggest that all types of commercial aircraft are susceptible to automation surprise and that such events may negatively impact safe operations. Understandably, as Boeing and Airbus dominate modern commercial fleets, these two manufacturers collectively accounted for 59% of all surprise events.\u003c/p\u003e\u003cdiv class=\"gridtable\"\u003e\n\u003cdiv class=\"colspec\" align=\"left\"\u003e\u0026nbsp;\u003c/div\u003e\n\u003ctable id=\"Tab2\" border=\"1\"\u003e\u003ccaption\u003e\n\u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\n\u003cdiv class=\"CaptionContent\"\u003e\n\u003cp\u003eIncidence of automated pilot surprise effects by aircraft type\u003c/p\u003e\n\u003c/div\u003e\n\u003c/caption\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\"\u003e\n\u003cp\u003eAircraft Manufacturer\u003c/p\u003e\n\u003c/th\u003e\u003cth align=\"left\"\u003e\n\u003cp\u003eAircraft type\u003c/p\u003e\n\u003c/th\u003e\u003cth align=\"left\"\u003e\n\u003cp\u003eN\u003c/p\u003e\n\u003c/th\u003e\u003cth align=\"left\"\u003e\n\u003cp\u003eN\u003c/p\u003e\n\u003c/th\u003e\u003cth align=\"left\"\u003e\n\u003cp\u003e%\u003c/p\u003e\n\u003c/th\u003e\u003cth colspan=\"2\" align=\"left\"\u003e\n\u003cp\u003eCum.%\u003c/p\u003e\n\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003eAirbus\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003eA300\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003e7\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003e82\u003c/p\u003e\n\u003c/td\u003e\u003ctd colspan=\"2\" align=\"left\"\u003e\n\u003cp\u003e30\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003e30\u003c/p\u003e\n\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003eA320 Family\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003e35\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd colspan=\"2\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003eA330\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003e14\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd colspan=\"2\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003eA340\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003e15\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd colspan=\"2\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003eA350\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003e11\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd colspan=\"2\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003eBoeing\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003eB727\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003e3\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003e86\u003c/p\u003e\n\u003c/td\u003e\u003ctd colspan=\"2\" align=\"left\"\u003e\n\u003cp\u003e31\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003e61\u003c/p\u003e\n\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003eB737 Classic\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003e24\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd colspan=\"2\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003eB737 NG\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003e19\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd colspan=\"2\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003eB757\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003e5\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd colspan=\"2\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003eB767\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003e8\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd colspan=\"2\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003eB777\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003e16\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd colspan=\"2\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003eB787\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003e11\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd colspan=\"2\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003eBombardier\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003eCRJ Series\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003e33\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003e57\u003c/p\u003e\n\u003c/td\u003e\u003ctd colspan=\"2\" align=\"left\"\u003e\n\u003cp\u003e20\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003e81\u003c/p\u003e\n\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003eDash 8 Q400\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003e24\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd colspan=\"2\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003eEmbraer\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003eEMB 135/145\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003e17\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003e28\u003c/p\u003e\n\u003c/td\u003e\u003ctd colspan=\"2\" align=\"left\"\u003e\n\u003cp\u003e10\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003e91\u003c/p\u003e\n\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003eEMB 170/175/190/195\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003e11\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd colspan=\"2\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003eMcDonnel Douglass\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003eMD-11\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003e13\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003e13\u003c/p\u003e\n\u003c/td\u003e\u003ctd colspan=\"2\" align=\"left\"\u003e\n\u003cp\u003e5\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003e96\u003c/p\u003e\n\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003eOthers\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003e15\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003e15\u003c/p\u003e\n\u003c/td\u003e\u003ctd colspan=\"2\" align=\"left\"\u003e\n\u003cp\u003e5\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003e100\u003c/p\u003e\n\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003eTotal\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003e281\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\n\u003cp\u003e281\u003c/p\u003e\n\u003c/td\u003e\u003ctd colspan=\"2\" align=\"left\"\u003e\n\u003cp\u003e100\u003c/p\u003e\n\u003c/td\u003e\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003ctfoot\u003e\u003ctr\u003e\u003ctd colspan=\"7\"\u003eNote: Total percentage may not equal 100% due to rounding of values. Others indicates aircraft of unknown type.\u003c/td\u003e\u003c/tr\u003e\u003c/tfoot\u003e\u003c/table\u003e\n\u003cp\u003e\u0026nbsp;Bayesian Network Model\u003c/p\u003e\n\u003c/div\u003e\u003cp\u003eBased on the analysis of the Bayesian network model, high-severity outcomes were found to be primarily associated with the flight management system, pilot mental models, and decision-making quality\u003csup\u003e38\u003c/sup\u003e. For example, failures in the FMS may lead to hazardous or high-severity aircraft behavior, such as steep pitch or roll angles or altitude loss. In contrast, low-severity outcomes were often linked to warning systems and monitoring awareness. Specifically, many alerts generated by warning systems turn out to be false alarms, posing minimal risk after verification\u003csup\u003e39\u003c/sup\u003e. These observations are consistent with previous exploratory analyses of ASRS data.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eInteraction Failures\u003c/p\u003e\n\u003cp\u003eWith ongoing technological advancements, various new automation technologies have been integrated into commercial aircraft. Does higher automation necessarily lead to safer flights? The answer is not always in the affirmative. According to an FAA\u003csup\u003e40\u003c/sup\u003einvestigation, approximately one-fourth of aviation accidents or serious incidents were associated with flight crew performance during abnormal situations. In such cases, pilots may be diverted when attempting to understand the abnormal aircraft state and respond to automation surprise, thereby degrading their situational awareness.\u003c/p\u003e\n\u003cp\u003ePilots make aviation decisions by interacting with the automation interface when current flight conditions deviate from their expectations\u003csup\u003e41\u003c/sup\u003e. For instance, the fatal crashes of Lion Air Flight JT610 (October 2018) and Ethiopian Airlines Flight ET302 (March 2019)\u0026mdash;both involving the Boeing 737 MAX\u0026mdash;were attributed to unexpected behavior from the Maneuvering Characteristics Augmentation System (MCAS). Investigations revealed that the MCAS's operational logic exceeded pilots' understanding of the 737 series\u0026rsquo; flight control system. This mismatch triggered automation surprise, resulted in human-automation interaction failure, and ultimately led to the loss of control \u003csup\u003e20,29\u003c/sup\u003e. These cases highlight the potential safety risks inherent in human-automation interaction.\u003c/p\u003e\n\u003cp\u003eAutomatic Surprise Human-Computer Interaction Model\u003c/p\u003e\n\u003cp\u003eBased on abnormal reports and subsequent data analysis, the research group developed a human-machine interaction (HMI) model for automation surprise (see Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e7\u003c/span\u003e). Specifically, pilots with stronger risk perception are more likely to detect abnormalities. A \u0026ldquo;+\u0026rdquo; indicates a reinforcing relationship, while a \u0026ldquo;\u0026minus;\u0026rdquo; indicates a weakening one. Experienced pilots typically show higher tolerance toward abnormal conditions caused by automation due to their past exposure.\u003c/p\u003e\n\u003cp\u003ePsychological competence refers to the congruence between a pilot\u0026rsquo;s mental health and adaptability and their role's psychological and occupational requirements\u003csup\u003e42\u003c/sup\u003e. Pilots with good psychological competence are more likely to remain calm and focused under stress, better equipping them to handle automation surprise. In contrast, other pilots may be more prone to emotional reactivity, such as anxiety and agitation. Meanwhile, mindfulness can help pilots regulate their basic psychological needs\u003csup\u003e17\u003c/sup\u003e, reducing their susceptibility to the psychological impact of surprise.\u003c/p\u003e\n\u003cp\u003eWhen experiencing surprise during flight, if pilots judge the event\u0026mdash;based on experience and standard operating procedures\u0026mdash;to be a false alarm, no action is required other than monitoring. However, if the event impacts flight safety, partial disengagement of automation and manual control may be necessary. Thus, to maintain safety, it is essential to repeatedly train for surprise scenarios in simulators. Additionally, Crew Resource Management (CRM) training can integrate surprise scenarios to help reduce psychological responses to automation surprise through improved communication and coordination\u003csup\u003e43\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eImportantly, the factors explored in this study do not act in isolation; rather, they interact dynamically. To ensure safe flight operations, pilots must continuously monitor their internal and external environments, collect relevant information, rapidly assess risks, and make timely decisions\u003csup\u003e8\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eCommercial Aircraft Human-Computer Interaction Model\u003c/p\u003e\n\u003cp\u003eIn summary, although the original intent of automation design is to enhance efficiency, reduce workload, and improve safety, it has introduced a range of unforeseen operational risks. These risks include the degraded monitoring and situational awareness of pilots as well as their declining manual flying skills. Post-\u0026ldquo;black swan\u0026rdquo; phenomena such as the \u0026ldquo;lumberjack effect,\u0026rdquo; a dramatic increase in workload, and automation surprise are additional risks. These phenomena are often referred to as the \"paradox of automation\"\u003csup\u003e44\u003c/sup\u003e, the root cause of which lies largely in the failure of human-machine interaction.\u003c/p\u003e\n\u003cp\u003eHistorically, research on pilot error has often been fragmented and reductionist, attributing errors solely to either the technology (automation) or the pilot (human). These attributions do not consider the interactive nature of errors arising from the interplay between pilots and automated systems\u003csup\u003e45\u003c/sup\u003e. This perspective is gradually evolving. In recent years, researchers have adopted top-down, systems-thinking approaches to model factors influencing human-automation interaction and to understand their underlying mechanisms\u003csup\u003e46,47\u003c/sup\u003e. Inspired by such work, and drawing upon the model developed in this study, we propose the human-machine interaction model for commercial aircraft shown in Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e8\u003c/span\u003e.\u003c/p\u003e\n\u003cp\u003eThis study makes several key contributions. First, the study conducts a multidimensional exploration of automation surprise formation. By combining ASRS data mining with Bayesian network modeling, the study reveals the mechanism behind automation surprise, offering a new perspective on human-machine interaction challenges in aviation automation systems. Second, the study provides a systematic assessment of training impacts. Specifically, this research evaluates how enhanced training methods affect automation surprise, providing scientific guidance for flight training innovations and improving pilots\u0026rsquo; capacity to handle unexpected automation behaviors. Third, the study conducts an exploration of explainability in aviation AI. By investigating explainable aviation-focused Large Language Models (LLMs) and their role in mitigating automation surprise, the study contributes to expanding the application of AI in aviation, offering new ideas to improve automation transparency and interpretability. Finally, the study explores how system transparency influences automation surprise, providing important design guidance for user-friendly automation systems and fostering innovation in human-centered interface design.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eBy constructing a human-machine interaction model for commercial aircraft, this study identifies and addresses potential risks associated with automation systems, thereby improving flight reliability and operational safety. The development of AI presents new opportunities for the design and optimization of aviation automation systems\u003csup\u003e48\u003c/sup\u003e. Specifically, as AI technology continues to evolve, the autonomy of cockpit automation systems is expected to improve, allowing for dynamic responses independent of pilot intervention. Research on automation surprise can in turn advance relevant technologies and support the development of AI systems with high transparency and explainability\u003csup\u003e9,31\u003c/sup\u003e. Furthermore, this research contributes to exploring AI\u0026rsquo;s role in aviation, promoting collaborative and cooperative intelligent systems, and ultimately supporting the innovation of trustworthy automation \u003csup\u003e9,49\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eLooking ahead, it is essential to build a trustworthy relationship between humans and autonomous systems, thereby improving automation system design. Focusing on that relationship will help reveal deficiencies in current cockpit design as well as guide the development of future automation systems that are more human-centered\u003csup\u003e27,48\u003c/sup\u003eand more aligned with human cognitive and operational habits\u003csup\u003e50\u003c/sup\u003e. In turn, automation surprise and human-machine conflict will be reduced\u003csup\u003e51\u003c/sup\u003e, mitigating safety risks from a technical standpoint.\u003c/p\u003e\u003cp\u003eResearch on automation surprise sits at the intersection of aviation engineering, AI, human factors, and engineering psychology. This interdisciplinary effort will cultivate cross-disciplinary professionals with integrated knowledge and skills, fostering both academic integration and industry progress.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003ch2\u003eCompeting interests\u003c/h2\u003e\u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e\u003c/p\u003e\u003ch2\u003eFunding\u003c/h2\u003e\u003cp\u003eThis work was supported by National Social Science Fund of China (19BSH038), and National Natural Science Fund of China (32000753),\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eLei Du,Xuqun You,And Xinye Wang. contributed to the conception and design the research; contributed to data acquisition;Lei Du ,Ying Li,Tao Wen,Mingliang Li and Yaoliang Wu contributed to data analysis and result interpretation.Lei Du,Qiang Wei and Xinye Wang drafed and revised the manuscript. Saifang Liu,Yuan Li and Ming Ji contributed to manuscript editing .All authors read and approved the final manuscript.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eData is provided within the manuscript or supplementary information files.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eYou, X., Ji, M. \u0026amp; Han, H. The Effects of Risk Perception and Flight Experience on Airline Pilots' Locus of Control with Regard to Safety Operation Behaviors. \u003cem\u003eAccident Analysis \u0026amp;amp; Prevention\u003c/em\u003e\u003cstrong\u003e57\u003c/strong\u003e, 131-139 (2013). https://doi.org:10.1016/j.aap.2013.03.036\u003c/li\u003e\n\u003cli\u003eIATA (International Air Transport Association). Safety Report 2021. Montreal: International Air Transport Association.( 2022).\u003c/li\u003e\n\u003cli\u003eBillings, C. E. Human-Centered Aviation Automation: Principles and Guidelines. \u003cem\u003eopenalex\u003c/em\u003e (1996).\u003c/li\u003e\n\u003cli\u003eSarter, N. Investigating Mode Errors on Automated Flight Decks: Illustrating the Problem-Driven, Cumulative, and Interdisciplinary Nature of Human Factors Research. \u003cem\u003eHuman Factors\u003c/em\u003e\u003cstrong\u003e50\u003c/strong\u003e, 506-510 (2008). https://doi.org:10.1518/001872008x312233\u003c/li\u003e\n\u003cli\u003eSarter, N. B., Mumaw, R. J. \u0026amp; Wickens, C. D. Pilots' Monitoring Strategies and Performance on Automated Flight Decks: An Empirical Study Combining Behavioral and Eye-Tracking Data. \u003cem\u003eHuman Factors\u003c/em\u003e\u003cstrong\u003e49\u003c/strong\u003e, 347.340-357.340 (2007). https://doi.org:10.1518/001872007x196685\u003c/li\u003e\n\u003cli\u003eDekker, S. W. A. \u0026amp; Woods, D. D. MABA-MABA or Abracadabra? Progress on Human-Automation Co-ordination. \u003cem\u003eCognition, Technology \u0026amp; Work\u003c/em\u003e\u003cstrong\u003e4\u003c/strong\u003e (2002). https://doi.org:10.1007/s101110200022\u003c/li\u003e\n\u003cli\u003eNaweed, A. \u0026amp; Bye, R. Joint Cognitive Systems: Patterns in Cognitive Systems Engineering. \u003cem\u003eErgonomics\u003c/em\u003e\u003cstrong\u003e51\u003c/strong\u003e, 768-770 (2008). https://doi.org:10.1080/00140130701223774\u003c/li\u003e\n\u003cli\u003eXinye, W., Yuan, L., Ming, C. \u0026amp; Xuqun, Y. The Detriments and Improvement of Automation Trust and Dependence to Aviation Safety. \u003cem\u003eAdvances in Psychological Science\u003c/em\u003e\u003cstrong\u003e25\u003c/strong\u003e, 1614-1614 (2017). https://doi.org:10.3724/sp.j.1042.2017.01614\u003c/li\u003e\n\u003cli\u003eEuropean Aviation Safety Agency. (2020). Artificial Intelligence Roadmap: A human-centric approach to AI in aviation (Version 1.0). Cologne, G. E. A. S. A.\u003c/li\u003e\n\u003cli\u003e ]Federal Aviation Administration. Operational use of flight path management systems: Final report of the performance-based Operations Aviation Rulemaking Commitee/Commercial Aviation Safety Team Flight Deck Automation Working Group. Washington: Author.(2013).\u003c/li\u003e\n\u003cli\u003e Moriarty, D. Practical human factors for pilots. London, UK: Elsevier.(2015).\u003c/li\u003e\n\u003cli\u003e Horstmann, G. Latency and Duration of the Action Interruption in Surprise. \u003cem\u003eCognition \u0026amp; emotion\u003c/em\u003e\u003cstrong\u003e20\u003c/strong\u003e, 242-273 (2006). https://doi.org:10.1080/02699930500262878\u003c/li\u003e\n\u003cli\u003e \u003cem\u003eInternational Air Transport Association. (2019). Loss of Control In-Flight Accident Analysis Report Edition 2019. Retrieved September, 2021. from https://www.iata.org/contentassets/b6eb2adc248c484192101edd1ed36015/loc-i 2019.pdf .\u003c/em\u003e\u0026nbsp;\u003c/li\u003e\n\u003cli\u003e Boer, R. J. d. \u0026amp; Hurts, K. Automation Surprise. \u003cem\u003eAviation Psychology and Applied Human Factors\u003c/em\u003e\u003cstrong\u003e7\u003c/strong\u003e, 28-41 (2017). https://doi.org:10.1027/2192-0923/a000113\u003c/li\u003e\n\u003cli\u003e Trippe, J. \u0026amp; Mauro, R. Understanding Automation Surprise: Analysis of ASRS Reports. \u003cem\u003e18th International Symposium on Aviation Psychology\u003c/em\u003e (2015).\u003c/li\u003e\n\u003cli\u003e Ji, M., You, X., Lan, J. \u0026amp; Yang, S. The Impact of Risk Tolerance, Risk Perception and Hazardous Attitude on Safety Operation among Airline Pilots in China. \u003cem\u003eSafety Science\u003c/em\u003e\u003cstrong\u003e49\u003c/strong\u003e, 1412-1420 (2011). https://doi.org:10.1016/j.ssci.2011.06.007\u003c/li\u003e\n\u003cli\u003e Ji, M., Yang, C., Li, Y., Xu, Q. \u0026amp; He, R. The Influence of Trait Mindfulness on Incident Involvement among Chinese Airline Pilots: the Role of Risk Perception and Flight Experience. \u003cem\u003eJournal of safety research\u003c/em\u003e\u003cstrong\u003e66\u003c/strong\u003e, 161-168 (2018). https://doi.org:10.1016/j.jsr.2018.07.005\u003c/li\u003e\n\u003cli\u003e Dekker, S. The Field Guide to Human Error Investigations. \u003cem\u003emsra\u003c/em\u003e (2017). https://doi.org:10.4324/9781315202778\u003c/li\u003e\n\u003cli\u003e Xinye, W., Shu, L., Wenhao, C., Yangmei, L. \u0026amp; Xuqun, Y. Management of Startle and Surprise in Flight: Extended Landman Model. \u003cem\u003eJournal of Psychological Science\u003c/em\u003e\u003cstrong\u003e44\u003c/strong\u003e (2021). https://doi.org:10.16719/j.cnki.1671-6981.20210621\u003c/li\u003e\n\u003cli\u003e European Aviation Safety Agency. (2018). Startle effect management. Report number: NLR-CR-2018-242. Cologne, G. E. A. S. A.\u003c/li\u003e\n\u003cli\u003e Chen, Y. Y., Fang, W.N., Guo, B.Y., \u0026amp; Bao, H.F. (2023). Effect of task interruption on task performance and moderating effect of mental fatigue. Acta Psychologica Sinica, 55(1), 22-35.\u003c/li\u003e\n\u003cli\u003e Baxter, G., Besnard, D. \u0026amp; Riley, D. Cognitive Mismatches in the Cockpit: Will They Ever Be A Thing of the Past? \u003cem\u003eApplied ergonomics\u003c/em\u003e\u003cstrong\u003e38\u003c/strong\u003e (2007). https://doi.org:10.1016/j.apergo.2007.01.005\u003c/li\u003e\n\u003cli\u003e Ong, D. C., Zaki, J. \u0026amp; Goodman, N. D. Computational Models of Emotion Inference in Theory of Mind: A Review and Roadmap. \u003cem\u003eTopics in Cognitive Science\u003c/em\u003e\u003cstrong\u003e11\u003c/strong\u003e, 338-357 (2019). https://doi.org:10.1111/tops.12371\u003c/li\u003e\n\u003cli\u003e Landman, A., Groen, E. L., Paassen, M. M. v., Bronkhorst, A. W. \u0026amp; Mulder, M. The Influence of Surprise on Upset Recovery Performance in Airline Pilots. \u003cem\u003eInternational Journal of Aerospace Psychology\u003c/em\u003e\u003cstrong\u003e27\u003c/strong\u003e (2017). https://doi.org:10.1080/10508414.2017.1365610\u003c/li\u003e\n\u003cli\u003e Wen, H.\u003cem\u003e et al.\u003c/em\u003e A Methodology to Assess Human-Automated System Conflict from Safety Perspective. \u003cem\u003eComputers \u0026amp; Chemical Engineering\u003c/em\u003e\u003cstrong\u003e165\u003c/strong\u003e, 107939-107939 (2022). https://doi.org:10.1016/j.compchemeng.2022.107939\u003c/li\u003e\n\u003cli\u003e woods, d. d. \u0026amp; hollnagel, e. Joint cognitive systems : patterns in cognitive systems engineering. \u003cem\u003emag\u003c/em\u003e (2006).\u003c/li\u003e\n\u003cli\u003e Grindley, B.\u003cem\u003e et al.\u003c/em\u003e Avoiding Automation Surprise: Identifying Requirements to Support Pilot Intervention in Automated Uncrewed Aerial Vehicle (UAV) Flight. \u003cem\u003eApplied Ergonomics\u003c/em\u003e\u003cstrong\u003e127\u003c/strong\u003e (2025). https://doi.org:10.1016/j.apergo.2025.104516\u003c/li\u003e\n\u003cli\u003e Ancel, E. \u0026amp; Shih, A. Modeling Increased Complexity and the Reliance on Automation: FLightdeck Automation Problems (FLAP) Model. \u003cem\u003esemanticscholar\u003c/em\u003e\u0026nbsp;\u003c/li\u003e\n\u003cli\u003e Wang, X.Y., Wang, S., Lin, J. K., \u0026amp; Du, L. Causative analysis of two Boeing 737 MAX crashes: FLAP model based on human-automation interaction. \u003cem\u003eApplied Psychology, \u003c/em\u003e30, 483-494 (2024). https://doi.org:10.20058/j.cnki.CJAP.023101\u003c/li\u003e\n\u003cli\u003e European Aviation Safety Agency. Artificial Intelligence Roadmap: A human-centric approach to AI in aviation (Version 1.0). Cologne, Germany: European Aviation Safety Agency, (2020).\u003c/li\u003e\n\u003cli\u003e Lyons, J. B., Clark, M. A., Wagner, A. R. \u0026amp; Schuelke, M. J. Certifiable Trust in Autonomous Systems: Making the Intractable Tangible. \u003cem\u003eAI Magazine\u003c/em\u003e\u003cstrong\u003e38\u003c/strong\u003e (2017). https://doi.org:10.1609/aimag.v38i3.2717\u003c/li\u003e\n\u003cli\u003e Farjadian, A. B., Thomsen, B., Annaswamy, A. M. \u0026amp; Woods, D. D. Resilient Flight Control: an Architecture for Human Supervision of Automation. \u003cem\u003eIEEE Transactions on Control Systems Technology\u003c/em\u003e\u003cstrong\u003e29\u003c/strong\u003e, 29-42 (2020). https://doi.org:10.1109/tcst.2019.2959542\u003c/li\u003e\n\u003cli\u003e Kinney, L. \u0026amp; O'Hare, D. Responding to an Unexpected In-Flight Event: Physiological Arousal, Information Processing, and Performance. \u003cem\u003eHuman Factors\u003c/em\u003e\u003cstrong\u003e62\u003c/strong\u003e (2020). https://doi.org:10.1177/0018720819854830\u003c/li\u003e\n\u003cli\u003e Liu, Y., Pan, Y. D., Li, M., Li, C. L., Wang, X.Y., \u0026amp; You, X.Q. Applications and prospects of near infrared spectroscopy in aeronautical psychology.\u003cem\u003e Psychological Science\u003c/em\u003e\u003cstrong\u003e46(6)\u003c/strong\u003e, 1518-1528. (2023).\u003c/li\u003e\n\u003cli\u003e Li, R., Hosseini, H., Saggar, M., Balters, S. C. \u0026amp; Reiss, A. L. Current Opinions on the Present and Future Use of Functional Near-Infrared Spectroscopy in Psychiatry. \u003cem\u003eNeurophotonics\u003c/em\u003e\u003cstrong\u003e10\u003c/strong\u003e (2023). https://doi.org:10.1117/1.nph.10.1.013505\u003c/li\u003e\n\u003cli\u003e Ji, M., Xie, X.D, \u0026amp; Qiu, Y . Application of eye tracking technology in cockpit ergonomics design.\u003cem\u003e Packaging Engineering\u003c/em\u003e\u003cstrong\u003e42(18)\u003c/strong\u003e, 84-93. (2021).\u003c/li\u003e\n\u003cli\u003e Parasuraman, R. Neuroergonomics: Brain, Cognition, and Performance at Work. \u003cem\u003eCURRENT DIRECTIONS IN PSYCHOLOGICAL SCIENCE\u003c/em\u003e\u003cstrong\u003e20\u003c/strong\u003e (2011). https://doi.org:10.1177/0963721411409176\u003c/li\u003e\n\u003cli\u003e Zhang, X. \u0026amp; Mahadevan, S. Bayesian Network Modeling of Accident Investigation Reports for Aviation Safety Assessment. \u003cem\u003eReliability Engineering \u0026amp; System Safety\u003c/em\u003e\u003cstrong\u003e209\u003c/strong\u003e, 107371-107371 (2020). https://doi.org:10.1016/j.ress.2020.107371\u003c/li\u003e\n\u003cli\u003e Wang, L., Zhang, J., Dong, C., Sun, H. \u0026amp; Ren, Y. A Method of Applying Flight Data to Evaluate Landing Operation Performance. \u003cem\u003eErgonomics\u003c/em\u003e\u003cstrong\u003e62\u003c/strong\u003e, 171-180 (2018). https://doi.org:10.1080/00140139.2018.1502806\u003c/li\u003e\n\u003cli\u003e FAA. Operational use of flight path management systems: Final report of the performance-based Operations Aviation Rulemaking Commitee/Commercial Aviation Safety Team Flight Deck Automation Working Group. Washington: Author. (2013).\u003c/li\u003e\n\u003cli\u003e Wang, Z. \u0026amp; You, X. Models and influential factors of aeronautical decision making. \u003cem\u003eAdvances in Psychological Science\u003c/em\u003e\u003cstrong\u003e25\u003c/strong\u003e (2017). https://doi.org:10.3724/sp.J.1042.2017.02002\u003c/li\u003e\n\u003cli\u003e WANG, Y. \u0026amp; ZHAO, R. Construction of pilots' psychological competency model based on personality characteristics. \u003cem\u003eJournal of Civil Aviation University of China\u003c/em\u003e\u003cstrong\u003e36\u003c/strong\u003e, 46-51 (2018). https://doi.org:10.3969/j.issn.1674-5590.2018.03.010\u003c/li\u003e\n\u003cli\u003e Macleod, N. Crew Resource Management Training: A Competence-based Approach for Airline Pilots. \u003cem\u003eopenalex\u003c/em\u003e (2021).\u003c/li\u003e\n\u003cli\u003e Faros, B. The Paradox of Automation. \u003cem\u003eJournal of Computing Sciences in Colleges\u003c/em\u003e (2016). https://doi.org:10.5555/2904298.2904320\u003c/li\u003e\n\u003cli\u003e Maurino, D. E. Human factors and aviation safety: What the in- dustry has, what the industry needs. .\u003cem\u003e Ergonomics\u003c/em\u003e\u003cstrong\u003e 43\u003c/strong\u003e, 952\u0026ndash;959. (2000).\u003c/li\u003e\n\u003cli\u003e Endsley, M. R. From here to autonomy: Lessons learned from human-automation research.\u003cem\u003e Human Factors,\u003c/em\u003e\u003cstrong\u003e 59 (1)\u003c/strong\u003e, 5-27. (2017).\u003c/li\u003e\n\u003cli\u003e Stowers, K.\u003cem\u003e et al.\u003c/em\u003e A Framework to Guide the Assessment of Human\u0026ndash;Machine Systems. \u003cem\u003eHuman Factors\u003c/em\u003e\u003cstrong\u003e59\u003c/strong\u003e, 172-188 (2017). https://doi.org:10.1177/0018720817695077\u003c/li\u003e\n\u003cli\u003e Wei, X. Seven essays on user-centered design: from automation to intelligent aircraft cockpits. \u003cem\u003eApplied Psychology.\u003c/em\u003e\u003cstrong\u003e28(4)\u003c/strong\u003e, 291-313. (2022).\u003c/li\u003e\n\u003cli\u003e Dong Wenli, Fang Weining, Trust in Automation:Research Review and Future Perspectives.\u003cem\u003eZidonghua Xuebao/Acta Automatica Sinica\u003c/em\u003e, 1183-1200 (2021). https://doi.org:10.16383/j.aas.c200432\u003c/li\u003e\n\u003cli\u003e Xu Wei, G. L., Gao Zaifeng. Human-AI interaction: a new interdisciplinary field to realize the concept of human-centered AI. \u003cem\u003eJournal of Intelligent Systems\u003c/em\u003e\u003cstrong\u003e 16(4)\u003c/strong\u003e, 605-621. (2021).\u003c/li\u003e\n\u003cli\u003e Xu, W., Dainoff, M. J., Ge, L. \u0026amp; Gao, Z. Transitioning to Human Interaction with AI Systems: New Challenges and Opportunities for HCI Professionals to Enable Human-Centered AI. \u003cem\u003eInternational Journal of Human-computer Interaction\u003c/em\u003e\u003cstrong\u003e39\u003c/strong\u003e (2023). https://doi.org:10.1080/10447318.2022.2041900\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Automated systems, Commercial aircraft, Automation Surprise, Aviation accident data, Models, Human-computer interaction","lastPublishedDoi":"10.21203/rs.3.rs-7171123/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7171123/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eSince the introduction of automation systems in commercial aircraft, flight safety has improved significantly. However, pilots’ increasing reliance on these systems has led to a decline in their manual flying skills. When automated systems display unexpected behaviors, pilots experience a substantial increase in workload, which may lead to additional unforeseen automation surprises and pose serious risks to flight safety. This study explores the mechanisms underlying automation surprises by mining aviation accident databases and constructing a Bayesian network model. Based on this model, the key factors contributing to automation surprises are further explored, and a human-machine interaction model for commercial aircraft is developed. The findings indicate that building an interaction model between pilots and automated systems helps pilots develop better automation safety awareness. Moreover, “human-centered artificial intelligence” can assist in mitigating safety issues in human-machine interactions from a system design perspective.\u003c/p\u003e","manuscriptTitle":"The Impact of Automation Systems on Commercial Aircraft: A Human-Machine Interaction Perspective","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-08-12 13:41:19","doi":"10.21203/rs.3.rs-7171123/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"35604fdd-6861-4dc0-8dff-ff3674c635a7","owner":[],"postedDate":"August 12th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":52783390,"name":"Physical sciences/Engineering"},{"id":52783391,"name":"Physical sciences/Mathematics and computing"},{"id":52783392,"name":"Health sciences/Risk factors"}],"tags":[],"updatedAt":"2025-09-15T06:08:44+00:00","versionOfRecord":[],"versionCreatedAt":"2025-08-12 13:41:19","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7171123","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7171123","identity":"rs-7171123","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}