Technology as Transformation: Persistence of Expertise, Fragility of Trust, and Erosion of Accountability in Air Cargo Operations

Technology as Transformation: Persistence of Expertise, Fragility of Trust, and Erosion of Accountability in Air Cargo Operations | 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 Research Article Technology as Transformation: Persistence of Expertise, Fragility of Trust, and Erosion of Accountability in Air Cargo Operations Tipavinee Suwanwong Rodbundith, Teeris Thepchalerm This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9160278/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 4 You are reading this latest preprint version Abstract As algorithmic systems increasingly mediate safety-critical work, questions arise about how expertise, trust, and accountability are changing. This study examines air cargo operations in Thailand, where professionals navigate varying levels of technological sophistication, from manual processes to advanced algorithmic systems. Through semi-structured interviews with six cargo professionals, analysed using the distributed cognition framework, the research reveals a central paradox: increasing technological sophistication demands greater human expertise whilst reducing organisational visibility. Three patterns emerge. Physical–digital gaps require continuous intervention yet remain organisationally invisible. Trust proves fragile, manifesting as calibrated scepticism, symbolic compliance, or axiological rejection. Answerability erodes as sophistication increases, with documentation functioning as a liability shield rather than a basis for learning. These findings demonstrate that the technology–expertise relationship is transformative rather than substitutive, and that institutional structures have not adapted to this transformation. The study contributes to sociotechnical scholarship by examining how distributed cognition operates across the technology adoption spectrum in safety-critical operational domains. Human-technology collaboration trust in automation distributed cognition invisible labour accountability 1. Introduction Algorithmic systems increasingly mediate professional work in safety-critical domains, fundamentally altering the nature of expertise, trust, and accountability (Novelli et al., 2024 ). Air cargo operations exemplify this transformation, functioning as complex sociotechnical systems in which cognitive work is distributed among human operators, algorithmic agents, and environmental structures. While organisations frequently implement these technologies under the assumption that they will eliminate human error and reduce labour (Mosier et al., 2017 ), evidence suggests that technology transforms rather than replaces work. As these systems advance from passive tools to active agents capable of determining critical parameters, they introduce fundamental tensions between the digital representation of logistics and the physical reality. The integration of these algorithmic agents precipitates significant challenges regarding sensemaking and accountability. Situated action theory posits that algorithms often fail to account for the contingencies of material reality, thereby forcing operators to bridge physical-digital mismatches. Furthermore, the lack of transparency in these systems makes it harder to calibrate trust, especially given the efficiency-thoroughness trade-offs that are common in high-pressure operations (Hollnagel, 2017 ). Importantly, even as decision-making authority is increasingly delegated to technologies, responsibility remains with human operators. This creates a moral crumple zone where individuals bear the repercussions of system failures (Elish, 2025 ). This discrepancy places professionals in a double bind of assuming responsibility without comprehension, a challenge exacerbated in local contexts such as Thailand, where systemic data-quality issues necessitate constant mediation between international standards and operational constraints (Tsolakis et al., 2023 ). Despite theoretical recognition, existing research on sensemaking, trust, and accountability largely examines these constructs in isolation and within single-technology contexts. No study systematically compares how these dimensions covary across levels of technological sophistication within the same operational domain. This gap matters because single-context studies cannot reveal whether challenges represent technology-specific artefacts or structural properties of human-technology collaboration. Cross-level comparison within a single domain controls for operational context while varying technological sophistication. This study addresses this gap by examining air cargo operations in which professionals use manual processes, basic software, and advanced algorithmic systems under identical operational constraints. 2. Theoretical Framework 2.1 Technology-Mediated Work in Air Cargo Operations Air cargo operations are complex socio-technical systems in which cognitive work is distributed across human operators, technological artefacts, and algorithmic agents (Stanton, Plant, et al., 2019 ). From package reception to aircraft loading, multiple processes require coordination between human judgment and technology-mediated decision support. While logistics providers have deployed algorithmic systems for optimization tasks for over a decade (Lau et al., 2009 ), contemporary implementations range from basic software to advanced algorithmic systems that actively determine critical operational parameters, such as the optimal centre of gravity and loading configurations (Desai et al., 2023 ). A recent study calls for empirical studies on how humans work with automation in organisational settings, particularly in high-stakes environments where system opacity and accountability pressures coexist (Lebovitz, Lifshitz-Assaf, & Levina, 2022 ). Evidence indicates significant variation in technology adoption across the air cargo industry. Despite the availability of advanced technology, adoption ranges from advanced decision support to manual operations with minimal technology. This gap, acknowledged in air cargo, remains underexplored in operational settings. This study addresses this gap by examining human-technology interaction across the technology adoption spectrum. Following the diffusion of innovations framework (Rogers et al., 2014 ) and the taxonomy of technology sophistication levels (Parasuraman & Riley, 1997 ), we conceptualise technology adoption as a continuum ranging from manual control to full automation. This framework enables a systematic examination of how technological sophistication affects three interrelated dimensions of work: collective sensemaking amid technological ambiguity, trust calibration across levels of technological sophistication, and accountability navigation in technology-augmented environments. 2.2 Sensemaking When Digital Meets Physical Reality Sensemaking occurs when established interpretive frameworks fail to explain observed phenomena (Turner et al., 2023 ; Weick et al., 2005 ). In technology-assisted cargo operations, ambiguity arises when outputs from digital systems conflict with physical observations. These conflicts manifest across all levels of technological sophistication, from advanced AI algorithmic clashes to manual booking inconsistencies. Despite technological differences, the fundamental challenge remains the same, that is, operators must reconcile digital representations with material reality. The divergence between human intuition and algorithmic logic requires a structured cognitive approach to resolve discrepancies during operational disruptions. Humans rely on intuitive pattern recognition, whereas software systems use analytical, rule-based processing (Hao et al., 2025 ). The data-frame model suggests that these conflicts force operators to choose between algorithmic frames that prioritise system data and experiential frames that privilege embodied knowledge (Klein et al., 2006 ). This process is inherently collective. When ambiguity arises, operators engage in discursive sensemaking through informal conversations and machine gossip, in which they discuss technology systems rather than merely operate them (Viktorelius & Larsson, 2025 ). Algorithms also transform work, as operators become implicit system trainers through data repair, fixing errors to align systems with local realities. This invisible labour is critical for operational resilience and safety (Dekker, 2016 ). Given the necessity of these human interventions in high-stakes environments, it is essential to examine the specific social mechanisms used to bridge the gap between data and the physical world. This study, therefore, introduces Research Question 1 (RQ1). RQ1 How do operators make sense of technological ambiguity when digital representations conflict with physical reality across different levels of technological sophistication? 2.3 Trust Development Under Operational Constraints Trust in systems develops through structured cognitive evaluation rather than mere emotion. Appropriate trust requires two forms of knowledge: procedural knowledge of system operation and principled understanding of the underlying logic (Oksman et al., 2024 ). Without principled knowledge, operators cannot achieve calibrated trust (Lee & See, 2004 ), risking either algorithmic bias or disuse. However, explainability does not always improve trust. Recent findings suggest explanations benefit experienced users while potentially confusing novices (Schemmer et al., 2022 ). Technical knowledge alone is insufficient. The concept of axiological trustworthiness grounds trust in value alignment (Afroogh et al., 2024 ). Operators evaluate whether technological recommendations align with professional norms and safety standards. Systems that prioritise throughput over load balancing may be rejected due to conflicting values rather than technical errors. This evaluation becomes more critical as technological sophistication increases. In the context of air cargo, where strict flight schedules and rapid turnaround times define the environment, trust remains fragile despite knowledge and value alignment. Single observed errors, such as a load-planning system miscalculating the centre of gravity or an automated pallet dimensioner producing inaccurate measurements, often lead to lasting algorithm aversion and a significant decline in trust (Wright et al., 2019 ). These challenges intensify under operational pressure. The efficiency thoroughness trade-off principle posits that time constraints force operators to sacrifice verification for efficiency. Furthermore, trust depends on organisational responsibility. Operators may exhibit low trust if they fear personal reprimand for following system recommendations within inadequate accountability frameworks (Gillespie, 2022 ). Calibration strategies vary across the adoption spectrum. Advanced algorithmic system users develop nuanced trust through experience, whereas basic technology users may treat technology as rough guidelines that require refinement. Understanding these patterns is essential for designing effective human technology partnerships. This leads to RQ2: RQ2 How is trust characterised across different levels of technology integration in high-stakes operational contexts? 2.4 The Accountability-Answerability Gap The integration of technologies into high-stakes cargo operations poses complex accountability challenges. Human operators can become moral crumple zones, absorbing legal liability for system failures they neither caused nor could have predicted (Elish, 2025 ). Although this phenomenon is most evident with opaque systems, similar patterns could appear throughout the technology spectrum. A key difference lies between liability, which refers to legal responsibility, and answerability, which involves the practical capacity to justify decisions (Novelli et al., 2024 ). This distinction is important across all levels of technological mediation, though the gap between liability and answerability may widen with technological sophistication. Responsibility depends on understanding how systems make decisions. When operators cannot explain their technical reasoning, a gap emerges between accountability and answerability, which varies with levels of technological sophistication (Coeckelbergh, 2020 ). Users of algorithmic systems face opacity that erodes answerability, while liability remains. Basic technology users share responsibility with transparent systems. Manual practitioners retain clear accountability, but transparency can still hinder accountability under time constraints. Errors in complex automated systems are often unforeseeable to operators who approve plans without understanding system logic, yet remain legally liable if incidents occur (Fraser & Suzor, 2025 ). This creates a professional double bind. Verifying system recommendations is often impossible due to opacity and time limits, yet blindly trusting them risks liability (Johnson & Powers, 2005 ). Operators sign off on system plans, cargo manifests, and routing to meet efficiency targets, holding final authority but lacking control or understanding of system reasoning. Research shows that this dynamic transforms work (Pachidi et al., 2021 ). Organisations may appear to adopt systems for legitimacy, while operators rely on human expertise for decisions. This symbolic adoption differs from substantive adoption, causing tensions between formal procedures and actual practices. Accountability varies across levels of technological sophistication, affecting operators' professional agency and identity. Key questions include whether operators are becoming strategic managers or mere risk buffers. Understanding how workers navigate efficiency pressures, liability, and their professional identity amid technological mediation is crucial for sustainable work design. RQ3 How do operators experience accountability when working with technology systems? 3. Methodology This study employs a qualitative research design to investigate distributed cognitive processes in human-technology interaction within air cargo operations. Given the complex socio-technical nature of phenomena involving sensemaking, trust, and accountability across diverse technological contexts, qualitative approaches prove essential for exploring participant lived experiences and organisational practices (Creswell & Poth, 2016 ). The study adopts the gate-to-gate distributed cognition framework, recognising that cognitive work in aviation is distributed across human agents, technological artefacts, and environmental structures rather than residing in individual minds. This framework accommodates diverse levels of technological sophistication and examines how cognitive distribution patterns vary with it. Data analysis employs Qualitative Content Analysis (Schreier, 2012 ), combining concept-driven deductive categories from the theoretical framework with data-driven inductive categories emerging from participant narratives. This hybrid approach enables a systematic examination of theoretically predicted patterns while remaining open to unexpected findings across the spectrum of technological sophistication. 3.1 Participants and Sampling Participants were selected through purposive sampling to represent diverse positions along the technology adoption spectrum. The sample consists of six air cargo operations professionals distributed across three adoption categories: advanced users, basic users, and primarily manual practitioners. This distribution captures meaningful variation in technological integration while achieving thematic saturation within each adoption category. All participants work in operational roles, such as load planning, ramp coordination, or operations supervision, with varying levels of experience with technology systems. Recruitment proceeded through professional industry networks, ensuring voluntary participation and informed consent. Although small, this sample size is appropriate for the study's goals. In qualitative research on technology adoption, six to twelve participants suffice with information-rich sampling and category saturation (Creswell & Poth, 2016 ). The two participants per level (advanced, basic, manual) enable comparisons within categories while preserving depth. This approach emphasises theoretical saturation over numerical representativeness (Schreier, 2012 ). Saturation occurred when no new codes or themes emerged, and cross-level patterns stabilised after analysing all six transcripts. 3.2 Data Collection Semi-structured, in-depth interviews lasting thirty to forty-five minutes were conducted via videoconference. The interview protocol explored three dimensions aligned with research questions. First, inspired by the Event Analysis of Systemic Teamwork methodology (Stanton, Salmon, & Walker, 2018 ), participants described the information sources they consult when resolving ambiguities and the communication patterns they use when technology outputs conflict with operational realities. This revealed variations in information-seeking strategies and collaborative sensemaking practices across levels of technological sophistication. Second, vignette-based scenarios probed trust calibration and accountability navigation. Scenarios presented situations involving data-system mismatches, organisational pressure under time constraints, and accountability for technology-generated decisions. These scenarios were designed to be applicable across levels of technological sophistication, examining how operators with different technological tools approach similar operational challenges. All interviews were conducted in Thai, audio recorded with written consent, and transcribed verbatim. 3.3 Data Analysis Transcribed data were analysed using Qualitative Content Analysis, with particular attention to variation across levels of technological sophistication. The coding frame combined deductive main categories from theoretical pillars, including Collective Sensemaking, Trust Calibration, and Accountability Navigation, with inductive subcategories emerging from data. Expected themes include discursive sensemaking practices (Viktorelius & Larsson, 2025 ), cognitive collision between human intuition and technological logic (Hao et al., 2025 ), axiological evaluation based on value alignment (Afroogh, Akbari, Malone, Kargar, & Alambeigi, 2024 ), data repair labour (Faraj et al., 2018 ; Pachidi et al., 2021 ), and answerability gaps where operators cannot explain decisions for which they remain liable (Coeckelbergh, 2020 ; Novelli et al., 2024 ). Analysis proceeded iteratively: first, coding within technology-level categories to identify level-specific patterns; then, conducting cross-level comparisons to identify common themes and systematic variations. The coding frame was refined iteratively to ensure that categories were mutually exclusive and exhaustive while capturing meaningful differences across the technology adoption spectrum. 3.4 Validity and Reliability Multiple validation strategies ensured rigour (Creswell & Poth, 2016 ; Schreier, 2012 ). Reflexivity was maintained through the researcher's journaling to monitor assumptions about technology adoption, effectiveness, and operator agency. Inter-rater reliability was established through double coding, in which a second trained analyst independently coded 20% of transcripts (two complete transcripts representing advanced algorithmic systems and manual operation contexts). Cohen's kappa coefficient was calculated to assess inter-rater agreement, yielding κ = 0.69, indicating substantial agreement beyond chance (Gisev et al., 2013 ). Discrepancies were resolved through dialogue and iterative refinement of coding definitions. 4. Results Analysis of six air cargo professionals revealed three overarching themes. Advanced users (P1, P4) employed technology-driven systems; basic users (P5, P6) relied on limited-functionality software, and minimal-technology users (P2, P3) worked primarily manually. Table 1 presents the coding framework. Table 1 Analysis results Theme Definition Representative Data THEME 1: Persistence of Human Expertise Physical-Digital Gap Systematic mismatches requiring intervention ‘If it's a garment, it bulges‘ (P1) Invisible Data Repair Continuous correction work ‘take the actual weight and close it in the system again.’ (P1) Embodied Knowledge Physical inspection as irreplaceable ‘AI can't walk and look’ (P3) THEME 2: Fragility of Trust Calibrated Scepticism Conditional trust despite experience ‘Trust it, but check’ (P4) Symbolic Trust Formal usage masking actual reliance ‘Don't trust it’ (P5) Axiological Rejection Value-based refusal ‘Don't trust 100 per cent on safety’ (P3) THEME 3: Erosion of Answerability Documentation Shield Records deflecting accountability ‘at least we have a record that the system warned’ (P1) Fear of Blame Individual accountability anxiety ‘Might get reprimanded’ (P6) 4.1 Persistence of Human Expertise 4.1.1 Physical-Digital Gap Physical-digital gaps manifested across all technology levels. Advanced users reported systematic calculation failures. P1 described garment cargo bulging beyond declared dimensions, with P4 reporting correctly dimensioned pallets failing to stack due to warping. Basic users experienced parallel gaps. P5 characterised inaccuracy as a Thailand-specific limitation, whereas P6 stated that actual weights frequently exceed system predictions, necessitating manual verification. Manual users avoided miscalculations through direct assessment. P2 explained that physical inspection reveals cargo conditions that systems cannot detect. 4.1.2 Invisible Data Repair Labour Physical-digital gaps generated continuous correction work, remaining organisationally invisible across technology levels. Advanced users performed algorithmic training through repeated corrections. P1 described operators taking the actual weight and re-entering it into the system, while P4 reported constant adjustments to align with physical reality. Basic users validated system outputs through physical verification. P5 noted that operations must be suspended and reviewed to verify that the build-up aligns with the plan, while P6 explained that system recommendations require verification before implementation. Manual users sustained operations through continuous assessment. P2 stated that every shipment requires a physical inspection, whereas P3 described repeatedly checking cargo security and balance during loading. 4.1.3 Embodied Knowledge as Irreplaceable Resource P3 stated, ‘AI can't come walk and look,’ emphasising physical presence as knowledge beyond algorithms. The verbs ‘walk’ and ‘look’ highlight irreplaceable embodied expertise. P2 noted that physical checks reveal cargo conditions that systems can't detect. Advanced users rely on physical inspection despite sophisticated algorithms. P1 and P4 said they maintain embodied expertise alongside algorithms, not replace it. Basic users favour physical knowledge over limited system capabilities. P5 said physical assessment provides information that quality systems can't, while P6 preferred physical inspection over system outputs. This reflects both operational necessity and a defence of professional identity. 4.2 Fragility of Trust 4.2.1 Calibrated Scepticism on patterns varied systematically among user levels. Advanced users exhibited nuanced scepticism: P4 trusted the system due to long-term use but still checked, whereas P1 believed the system was problem-free but acknowledged user-reported issues, revealing defensive trust. Basic users had different behaviours. P5 viewed systems as guidelines, with final judgement left to humans, whereas P6 followed procedures but doubted the accuracy of the systems. Manual users completely rejected algorithmic trust. P2 believed trust should only be placed in human judgement, while P3 outright refused to trust technology, especially regarding safety. 4.2.2 Symbolic Trust and Organisational Illusion Basic users demonstrated the largest divergence between organisational expectations and actual reliance. P5 stated, ‘The company asks us to use it as a guideline... basically, we don't trust it,‘ while organisations require system usage. This symbolic trust creates an organisational illusion in which management perceives technology-driven operations, while practitioners maintain human-centred decision-making. P1 noted pressures that force ‘use it completely or not at all’ choices, thereby eliminating calibration space. 4.2.3 Axiological Rejection Manual users rejected algorithmic trust based on principles of safety culture rather than performance assessment. P3 refused to rely completely on technology for safety-critical decisions, regardless of system reliability. This value-based rejection operated independently of empirical system performance, grounding distrust in professional ethics rather than pragmatic evaluation. System improvements cannot address axiologically grounded distrust because rejection stems from core safety principles rather than technical limitations. 4.3 Erosion of Answerability 4.3.1 Documentation as a Liability Shield Advanced users employ documentation defensively. P1's statement, ‘at least we have a record that the system warned,‘ frames documentation as liability protection rather than decision support. The qualifier ‘at least’ concedes insufficient understanding while asserting procedural compliance. P1 described mandatory warning compliance, under which operators may not question warnings but remain responsible for their accuracy. P4 observed that distributed authority spreads responsibility, yet individuals still retain accountability. 4.3.2 Fear of Blame Fear of individual blame manifested across technology levels, though mechanisms differed. Basic users expressed anxiety despite procedural compliance. P6 stated that they might be reprimanded even when following system recommendations, while P5 expressed concern about being held responsible for system errors. Advanced users anticipated accountability despite distributed authority. P1 noted that operators bear responsibility when systems provide incorrect recommendations, while P4 reported that investigations focus on human inputs rather than algorithmic limitations. Manual users faced direct accountability. P2 explained that they accepted full responsibility for decisions made through physical assessment, while P3 stated that clear accountability accompanies manual expertise. 5. Discussion 5.1 Theoretical Contributions The results support Situated Action Theory, which views technology-generated plans as flexible resources rather than strict scripts (Napoleone et al., 2023 ; Suchman, 2007 ). Our findings show that the difficulty of operational gaps rises with the level of technology. Basic users mostly fix clear database errors. In contrast, advanced users must address subtle blind spots in which the software fails to account for physical properties. This confirms that as systems become more rigid, operators need deeper knowledge to resolve conflicts. Both require continuous correction work, demonstrating that expertise shifts rather than disappears. This suggests the relationship between technology and expertise is transformative rather than substitutive (Berente et al., 2021 ). The transformation is paradoxical: more advanced technology requires more sophisticated human intervention, not less. P1's description of garments that categorically bulge, yet systems sometimes fail to account for this, reveals linguistic masking of structural limitations through the framing of randomness. Operators describe predictable material behaviour failures as occasional system oversights, obscuring the fundamental gap between algorithmic calculation and physical reality. The persistence of correction work demonstrates the continued need for expertise, often unseen, representing invisible labour (Star & Strauss, 2016 ), in which cycles are regarded as part of system training rather than as crucial expertise. Pachidi et al. ( 2021 ) noted that operators assume the role of algorithm trainers; however, our findings indicate that this shift occurs without professional acknowledgment or role redefinition. Organisations treat correction as a temporary inefficiency requiring elimination rather than a permanent requirement demanding recognition. When such expertise is unrecognised, organisations tend to set unrealistic expectations for efficiency while simultaneously devaluing the human contributions that make algorithmic systems function. Consequently, performance metrics presume seamless data integration, which forces workers to absorb the temporal cost of reconciling digital errors without official credit. This structural misalignment leads management to misattribute skilled intervention to operational friction, thereby penalising operators for the very competence that prevents systemic failure. In addition, embodied knowledge highlights the limitations of replacing human skill with algorithms (Stanton et al., 2019 ), functioning both as essential operational knowledge enabling cargo verification algorithms to perform and as a safeguard for professional identity against perceived displacement. Findings addressing how trust develops fragility under operational constraints complicate the calibration model through three mechanisms (Lee & See, 2004 ). Initially, system design constraints enforce binary functionality, meaning algorithmic outputs are either fully implemented or not at all. Instead of relying solely on binary trust, operators navigate these limitations by symbolically adhering to system requirements, with actual decisions being driven by human judgement and experience (Pachidi et al., 2021 ). Organisations require system usage for technological legitimacy, operators comply symbolically, while actual decisions rest on human judgement. This creates an organisational illusion in which management perceives technology-driven operations, while practitioners maintain human-centred decision-making. When symbolic and substantive trust diverge under these constraints, organisations lack an accurate understanding of decision-making processes, thereby increasing risk exposure during system failures when assumed algorithmic safeguards prove illusory. Second, axiological rejection operates independently of performance (Afroogh et al., 2024 ). P3's categorical refusal to ‘trust 100 per cent’ in matters of safety reflects professional values that prioritise human primacy in high-stakes decisions (Dekker, 2016 ) rather than algorithm aversion. Complex system failures arise from small deviations that accumulate over time; maintaining human verification despite system sophistication provides crucial layers of defence regardless of system reliability. Third, accountability asymmetry hinders the stabilisationn of trust. Algorithms give recommendations but bear no liability for failure. This leads operators to adopt a defensive stance, rejecting valid outputs not because of doubt, but because they can't justify the algorithm's logic when errors occur. As a result, trust depends not on system reliability but on the operator's ability to justify decisions to superiors. Our findings demonstrate that technology erodes answerability, effectively extending the moral crumple zone. This occurs through distinct mechanisms where documentation serves as a liability shield rather than a tool for explanation, providing procedural proof without substantive justification. Furthermore, while decision-making authority is distributed between humans and machines, blame remains concentrated on the individual. This disparity causes anxiety, as operators fear reprimand for uncontrollable system outputs. Although they have formal authority, their understanding of how to justify their decisions is lacking. This traps them in a double bind of responsibility without comprehension (Johnson & Powers, 2005 ; Novelli et al., 2024 ). The consequences are twofold: organisationally, the inability to explain decisions precludes learning from failures, while individually, professional identity devolves from expert to system manager in the absence of institutional support, leaving operators to navigate role ambiguity alone. 5.2 Practical Implications Organisations must recognise data correction work in formal job descriptions, performance evaluations, and compensation structures. Current invisibility systematically devalues expertise sustaining algorithmic functionality. When correction work lacks institutional recognition, organisations develop unrealistic expectations of efficiency, while skilled operators exit roles that fail to acknowledge their actual responsibilities. This produces two failures. First, the quality of correction work declines as expertise attrition accelerates. Second, system failures increase because operators who perform unrecognised work lack incentives to maintain correction standards. Beyond recognising correction work, verification protocols must accommodate operational time constraints rather than impose binary trust requirements. The current design assumes operators possess deliberative space for calibration. Operational reality eliminates this space, forcing complete reliance or complete rejection. Both extremes compromise safety. Blind trust allows uncorrected errors to propagate. Complete rejection wastes system capabilities and increases manual workload. Integrating protocols directly into workflows, rather than treating verification as a deviation from standard procedures, enables graduated reliance appropriate to operational constraints. Trust calibration becomes meaningless without corresponding accountability alignment. Accountability structures currently assign operators responsibility for algorithmic outputs they cannot explain or override. This produces three problems. Investigations focus on operator inputs rather than system design limitations, preventing organisational learning. Operators experience professional anxiety from bearing liability without corresponding authority. Documentation shifts from decision explanation to liability protection, obscuring actual reasoning when failures occur. Regulatory frameworks should require accountability, the distribution of decision-making authority, and the alignment of decision-making authority with responsibilities. When systems make critical determinations, designers share liability in proportion to their control over the algorithmic logic. When operators bear full responsibility, they receive override authority commensurate with assigned accountability. These structural changes must account for fundamental information boundaries that constrain algorithmic systems. Embodied knowledge provides information that algorithmic systems cannot capture. Physical cargo assessment reveals material properties and condition characteristics that are difficult to digitise. Organisations that position physical expertise as obsolete lose essential capabilities for detecting when algorithmic outputs diverge from material reality. System design should support human-algorithm collaboration, leveraging complementary capabilities rather than assuming complete digitisationn is achievable. Career pathways must value physical expertise alongside data analysis competencies. Current approaches that treat embodied knowledge as transient skill, requiring its elimination through automation, ignore persistent information boundaries that constrain the algorithmic capture of such knowledge. 6. Conclusion, Limitations, and Future Directions Technology adoption transforms rather than eliminates human work, creates trust dilemmas rather than resolving uncertainties, and obscures rather than clarifies responsibility. Physical-digital gaps require continuous intervention; correction work remains invisible yet essential; and embodied knowledge is irreplaceable. Trust requires active construction, in which calibrated scepticism represents achievement, symbolic trust masks authority, and axiological rejection operates independently of performance. Answerability erosion creates vulnerabilities through documentation shields, distributed authority, and unrecognised transformation. As technology sophistication increases, the requirements for human expertise intensify, while organisational visibility decreases. Sustainable collaboration requires acknowledging essential expertise, supporting appropriate calibration, and establishing frameworks matching actual authority. Study limitations include a small sample size (n = 6), a focus on theoretical saturation rather than statistical breadth, a geographic concentration that emphasises contextual variables, and a cross-sectional design that captures a single point in time. Despite these, the study offers valuable insights. Participant 5's view of dimensional inaccuracy as region-specific indicates that local conditions influence technology adoption, challenging assumptions about the universal effectiveness of automated systems. Future research should include comparative analyses across high-stakes areas, longitudinal studies of adaptation, and ethnographies to explore hidden labour. Developing accountability frameworks is vital to ensure accountability. Solving the paradox in which automation increases expertise demands yet reduces visibility requires institutional innovation to meet the demand for new competencies. Declarations Funding: The authors did not receive any funding support for the submitted work. No funding was received for conducting this study. Competing Interests: The authors have no competing interests to declare that are relevant to the content of this article. The authors have no relevant financial or non-financial interests to disclose. Ethics Approval: This study was conducted in accordance with ethical standards for research involving human participants. All procedures were approved by the author’s ethics committee. All participants provided informed written consent prior to data collection and were assured of confidentiality and voluntary participation. Data Availability: The data that support the findings of this study consist of interview transcripts collected from individual participants. These data are not publicly available due to participant privacy and confidentiality constraints. The data cannot be fully anonymised without compromising the richness required for qualitative analysis. Researchers who wish to discuss access conditions may contact the corresponding author. Author Contribution T.P.: Conceptualization, Investigation, Data curation, Validation, Writing – original draft, Supervision. T.T.: Conceptualization, Methodology, Data curation, Formal analysis, Validation, Writing – original draft. References Afroogh, S., Akbari, A., Malone, E., Kargar, M., & Alambeigi, H. (2024). Trust in AI: Progress, challenges, and future directions. Humanities and Social Sciences Communications, 11 (1), 1568. In. Berente, N., Gu, B., Recker, J., & Santhanam, R. (2021). Managing artificial intelligence. In (Vol. 45, pp. 1433-1450): Management Information Systems Research Center, University of Minnesota. Coeckelbergh, M. (2020). Artificial intelligence, responsibility attribution, and a relational justification of explainability. Science and Engineering Ethics, 26 (4), 2051-2068. Creswell, J. W., & Poth, C. N. (2016). Qualitative Inquiry and Research Design: Choosing Among Five Approaches : SAGE Publications. Dekker, S. (2016). Drift into failure: From hunting broken components to understanding complex systems : CRC press. Desai, J., Srivathsan, S., Lai, W. Y., Li, L., & Yu, C. (2023). An optimization-based decision support tool for air cargo loading. Computers & Industrial Engineering, 175 , 108816. doi:https://doi.org/10.1016/j.cie.2022.108816 Elish, M. C. (2025). Moral crumple zones: cautionary tales in human–robot interaction. In Robot Law: Volume II (pp. 83-105): Edward Elgar Publishing. Faraj, S., Pachidi, S., & Sayegh, K. (2018). Working and organising in the age of the learning algorithm. Information and Organization, 28 (1), 62-70. Fraser, H. L., & Suzor, N. P. (2025). Locating fault for AI harms: a systems theory of foreseeability, reasonable care and causal responsibility in the AI value chain. Law, Innovation and Technology, 17 (1), 103-138. Gillespie, T. (2022). Building trust and responsibility into autonomous human-machine teams. Frontiers in Physics, 10 , 942245. Gisev, N., Bell, J. S., & Chen, T. F. (2013). Interrater agreement and interrater reliability: key concepts, approaches, and applications. Research in Social and Administrative Pharmacy, 9 (3), 330-338. Hao, X., Demir, E., & Eyers, D. (2025). Beyond human-in-the-loop: Sensemaking between artificial intelligence and human intelligence collaboration. Sustainable Futures, 10 , 101152. Hollnagel, E. (2017). The ETTO principle: efficiency-thoroughness trade-off: why things that go right sometimes go wrong : CRC press. Johnson, D. G., & Powers, T. M. (2005). Computer systems and responsibility: A normative look at technological complexity. Ethics and information technology, 7 (2), 99-107. Klein, G., Moon, B., & Hoffman, R. R. (2006). Making sense of sensemaking 1: Alternative perspectives. IEEE intelligent systems, 21 (4), 70-73. Lau, H. C. W., Chan, T. M., Tsui, W. T., Ho, G. T. S., & Choy, K. L. (2009). An AI approach for optimizing multi-pallet loading operations. Expert Systems with Applications, 36 (3, Part 1), 4296-4312. doi:https://doi.org/10.1016/j.eswa.2008.03.024 Lebovitz, S., Lifshitz-Assaf, H., & Levina, N. (2022). To engage or not to engage with AI for critical judgments: How professionals deal with opacity when using AI for medical diagnosis. Organization science, 33 (1), 126-148. Lee, J. D., & See, K. A. (2004). Trust in automation: Designing for appropriate reliance. Human factors, 46 (1), 50-80. Mosier, K. L., Fischer, U., Burian, B. K., & Kochan, J. A. (2017). Autonomous, context-sensitive, task management systems and decision support tools I: Human-autonomy teaming fundamentals and state of the art. Napoleone, A., Andersen, A.-L., Brunoe, T. D., & Nielsen, K. (2023). Towards human-centric reconfigurable manufacturing systems: Literature review of reconfigurability enablers for reduced reconfiguration effort and classification frameworks. Journal of Manufacturing Systems, 67 , 23-34. Novelli, C., Taddeo, M., & Floridi, L. (2024). Accountability in artificial intelligence: What it is and how it works. Ai & Society, 39 (4), 1871-1882. Oksman, V., Kulju, M., & Kramar, V. (2024). Exploring the Factors Influencing Public Acceptance and Intention to Use Automated Drone Delivery Systems: A Trust Theory Approach . Pachidi, S., Berends, H., Faraj, S., & Huysman, M. (2021). Make way for the algorithms: Symbolic actions and change in a regime of knowing. Organization science, 32 (1), 18-41. Parasuraman, R., & Riley, V. (1997). Humans and automation: Use, misuse, disuse, abuse. Human factors, 39 (2), 230-253. Rogers, E. M., Singhal, A., & Quinlan, M. M. (2014). Diffusion of innovations. In An integrated approach to communication theory and research (pp. 432-448): Routledge. Schemmer, M., Hemmer, P., Nitsche, M., Kühl, N., & Vössing, M. (2022). A meta-analysis of the utility of explainable artificial intelligence in human-AI decision-making. Paper presented at the Proceedings of the 2022 AAAI/ACM Conference on AI, Ethics, and Society. Schreier, M. (2012). Qualitative Content Analysis in Practice : SAGE Publications. Stanton, N. A., Plant, K. L., Revell, K. M., Griffin, T. G., Moffat, S., & Stanton, M. (2019). Distributed cognition in aviation operations: a gate-to-gate study with implications for distributed crewing. Ergonomics, 62 (2), 138-155. Stanton, N. A., Salmon, P., & Walker, G. H. (2018). Systems thinking in practice: Applications of the event analysis of systemic teamwork method : CRC Press. Star, S. L., & Strauss, A. (2016). 19 Layers of Silence, Arenas of Voice: The Ecology of Visible and Invisible Work. Boundary Objects and Beyond: Working with Leigh Star , 351. Suchman, L. A. (2007). Human-machine reconfigurations: Plans and situated actions : Cambridge university press. Tsolakis, N., Schumacher, R., Dora, M., & Kumar, M. (2023). Artificial intelligence and blockchain implementation in supply chains: a pathway to sustainability and data monetisation? Annals of Operations Research, 327 (1), 157-210. Turner, J. R., Allen, J., Hawamdeh, S., & Mastanamma, G. (2023). The multifaceted sensemaking theory: A systematic literature review and content analysis on sensemaking. Systems, 11 (3), 145. Viktorelius, M., & Larsson, S. (2025). Talking about machines: discursive sensemaking and workplace learning in the age of AI. Studies in Continuing Education, 47 (3), 634-652. Weick, K. E., Sutcliffe, K. M., & Obstfeld, D. (2005). Organising and the process of sensemaking. Organization science, 16 (4), 409-421. Wright, J. L., Chen, J. Y., & Lakhmani, S. G. (2019). Agent transparency and reliability in human–robot interaction: The influence on user confidence and perceived reliability. IEEE Transactions on Human-Machine Systems, 50 (3), 254-263. Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Review Version 1 posted Reviewers invited by journal 30 Mar, 2026 Editor assigned by journal 23 Mar, 2026 Submission checks completed at journal 23 Mar, 2026 First submitted to journal 18 Mar, 2026 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-9160278","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":614532341,"identity":"53ec991c-2004-4dcc-8a56-a8917b8569c4","order_by":0,"name":"Tipavinee Suwanwong Rodbundith","email":"","orcid":"","institution":"Mae Fah Luang University","correspondingAuthor":false,"prefix":"","firstName":"Tipavinee","middleName":"Suwanwong","lastName":"Rodbundith","suffix":""},{"id":614532342,"identity":"a74cd250-d631-4272-9d9f-a9ea24b8e7c1","order_by":1,"name":"Teeris Thepchalerm","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA4UlEQVRIiWNgGAWjYLCCBCA2YGA+wFAB4bPhVc0D1SJhwMCWwHAGLMZMhBYGsBYeA+K02LM3H/vw4A9DnTl7z9cNB9sY5Pkb+I89wGsLz7HkGYltDBKWPWe33QBqMZxxgJndAK8WiRxjhsQGoMNu5G67/bGNgXED0GESeLXIv//MkPAHqOX+m2cgW+wJa5HgYWZIYAPZwsMG0pJIWMuZNKDD2iQkN5xJM7tx4JxE8ozDzGZ4tbC3H37M+OOPDb/B8cPPbhwos7Htb298hlcLFEggMZiJUD8KRsEoGAWjAD8AAJ3/RGPshVs1AAAAAElFTkSuQmCC","orcid":"","institution":"Mae Fah Luang University","correspondingAuthor":true,"prefix":"","firstName":"Teeris","middleName":"","lastName":"Thepchalerm","suffix":""}],"badges":[],"createdAt":"2026-03-18 14:08:08","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9160278/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9160278/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":105954805,"identity":"cb83f47d-4878-4d59-a40b-7f91027fb658","added_by":"auto","created_at":"2026-04-01 19:43:09","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":791169,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9160278/v1/d0f2ece9-4088-4a64-b01c-a7bdd10ab20d.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Technology as Transformation: Persistence of Expertise, Fragility of Trust, and Erosion of Accountability in Air Cargo Operations","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eAlgorithmic systems increasingly mediate professional work in safety-critical domains, fundamentally altering the nature of expertise, trust, and accountability (Novelli et al., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Air cargo operations exemplify this transformation, functioning as complex sociotechnical systems in which cognitive work is distributed among human operators, algorithmic agents, and environmental structures. While organisations frequently implement these technologies under the assumption that they will eliminate human error and reduce labour (Mosier et al., \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2017\u003c/span\u003e), evidence suggests that technology transforms rather than replaces work. As these systems advance from passive tools to active agents capable of determining critical parameters, they introduce fundamental tensions between the digital representation of logistics and the physical reality.\u003c/p\u003e \u003cp\u003eThe integration of these algorithmic agents precipitates significant challenges regarding sensemaking and accountability. Situated action theory posits that algorithms often fail to account for the contingencies of material reality, thereby forcing operators to bridge physical-digital mismatches. Furthermore, the lack of transparency in these systems makes it harder to calibrate trust, especially given the efficiency-thoroughness trade-offs that are common in high-pressure operations (Hollnagel, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). Importantly, even as decision-making authority is increasingly delegated to technologies, responsibility remains with human operators. This creates a moral crumple zone where individuals bear the repercussions of system failures (Elish, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). This discrepancy places professionals in a double bind of assuming responsibility without comprehension, a challenge exacerbated in local contexts such as Thailand, where systemic data-quality issues necessitate constant mediation between international standards and operational constraints (Tsolakis et al., \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eDespite theoretical recognition, existing research on sensemaking, trust, and accountability largely examines these constructs in isolation and within single-technology contexts. No study systematically compares how these dimensions covary across levels of technological sophistication within the same operational domain. This gap matters because single-context studies cannot reveal whether challenges represent technology-specific artefacts or structural properties of human-technology collaboration. Cross-level comparison within a single domain controls for operational context while varying technological sophistication. This study addresses this gap by examining air cargo operations in which professionals use manual processes, basic software, and advanced algorithmic systems under identical operational constraints.\u003c/p\u003e"},{"header":"2. Theoretical Framework","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e2.1 Technology-Mediated Work in Air Cargo Operations\u003c/h2\u003e \u003cp\u003eAir cargo operations are complex socio-technical systems in which cognitive work is distributed across human operators, technological artefacts, and algorithmic agents (Stanton, Plant, et al., \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). From package reception to aircraft loading, multiple processes require coordination between human judgment and technology-mediated decision support. While logistics providers have deployed algorithmic systems for optimization tasks for over a decade (Lau et al., \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2009\u003c/span\u003e), contemporary implementations range from basic software to advanced algorithmic systems that actively determine critical operational parameters, such as the optimal centre of gravity and loading configurations (Desai et al., \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eA recent study calls for empirical studies on how humans work with automation in organisational settings, particularly in high-stakes environments where system opacity and accountability pressures coexist (Lebovitz, Lifshitz-Assaf, \u0026amp; Levina, \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Evidence indicates significant variation in technology adoption across the air cargo industry. Despite the availability of advanced technology, adoption ranges from advanced decision support to manual operations with minimal technology. This gap, acknowledged in air cargo, remains underexplored in operational settings.\u003c/p\u003e \u003cp\u003eThis study addresses this gap by examining human-technology interaction across the technology adoption spectrum. Following the diffusion of innovations framework (Rogers et al., \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2014\u003c/span\u003e) and the taxonomy of technology sophistication levels (Parasuraman \u0026amp; Riley, \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e1997\u003c/span\u003e), we conceptualise technology adoption as a continuum ranging from manual control to full automation. This framework enables a systematic examination of how technological sophistication affects three interrelated dimensions of work: collective sensemaking amid technological ambiguity, trust calibration across levels of technological sophistication, and accountability navigation in technology-augmented environments.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e2.2 Sensemaking When Digital Meets Physical Reality\u003c/h2\u003e \u003cp\u003eSensemaking occurs when established interpretive frameworks fail to explain observed phenomena (Turner et al., \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Weick et al., \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2005\u003c/span\u003e). In technology-assisted cargo operations, ambiguity arises when outputs from digital systems conflict with physical observations. These conflicts manifest across all levels of technological sophistication, from advanced AI algorithmic clashes to manual booking inconsistencies. Despite technological differences, the fundamental challenge remains the same, that is, operators must reconcile digital representations with material reality. The divergence between human intuition and algorithmic logic requires a structured cognitive approach to resolve discrepancies during operational disruptions.\u003c/p\u003e \u003cp\u003eHumans rely on intuitive pattern recognition, whereas software systems use analytical, rule-based processing (Hao et al., \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). The data-frame model suggests that these conflicts force operators to choose between algorithmic frames that prioritise system data and experiential frames that privilege embodied knowledge (Klein et al., \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2006\u003c/span\u003e). This process is inherently collective. When ambiguity arises, operators engage in discursive sensemaking through informal conversations and machine gossip, in which they discuss technology systems rather than merely operate them (Viktorelius \u0026amp; Larsson, \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eAlgorithms also transform work, as operators become implicit system trainers through data repair, fixing errors to align systems with local realities. This invisible labour is critical for operational resilience and safety (Dekker, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). Given the necessity of these human interventions in high-stakes environments, it is essential to examine the specific social mechanisms used to bridge the gap between data and the physical world. This study, therefore, introduces Research Question 1 (RQ1).\u003c/p\u003e \u003cp\u003e \u003cstrong\u003eRQ1\u003c/strong\u003e \u003cp\u003eHow do operators make sense of technological ambiguity when digital representations conflict with physical reality across different levels of technological sophistication?\u003c/p\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003e2.3 Trust Development Under Operational Constraints\u003c/h2\u003e \u003cp\u003eTrust in systems develops through structured cognitive evaluation rather than mere emotion. Appropriate trust requires two forms of knowledge: procedural knowledge of system operation and principled understanding of the underlying logic (Oksman et al., \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Without principled knowledge, operators cannot achieve calibrated trust (Lee \u0026amp; See, \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2004\u003c/span\u003e), risking either algorithmic bias or disuse. However, explainability does not always improve trust. Recent findings suggest explanations benefit experienced users while potentially confusing novices (Schemmer et al., \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2022\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eTechnical knowledge alone is insufficient. The concept of axiological trustworthiness grounds trust in value alignment (Afroogh et al., \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Operators evaluate whether technological recommendations align with professional norms and safety standards. Systems that prioritise throughput over load balancing may be rejected due to conflicting values rather than technical errors. This evaluation becomes more critical as technological sophistication increases.\u003c/p\u003e \u003cp\u003eIn the context of air cargo, where strict flight schedules and rapid turnaround times define the environment, trust remains fragile despite knowledge and value alignment. Single observed errors, such as a load-planning system miscalculating the centre of gravity or an automated pallet dimensioner producing inaccurate measurements, often lead to lasting algorithm aversion and a significant decline in trust (Wright et al., \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). These challenges intensify under operational pressure. The efficiency thoroughness trade-off principle posits that time constraints force operators to sacrifice verification for efficiency. Furthermore, trust depends on organisational responsibility. Operators may exhibit low trust if they fear personal reprimand for following system recommendations within inadequate accountability frameworks (Gillespie, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2022\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eCalibration strategies vary across the adoption spectrum. Advanced algorithmic system users develop nuanced trust through experience, whereas basic technology users may treat technology as rough guidelines that require refinement. Understanding these patterns is essential for designing effective human technology partnerships. This leads to RQ2:\u003c/p\u003e \u003cp\u003e \u003cstrong\u003eRQ2\u003c/strong\u003e \u003cp\u003eHow is trust characterised across different levels of technology integration in high-stakes operational contexts?\u003c/p\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003e2.4 The Accountability-Answerability Gap\u003c/h2\u003e \u003cp\u003eThe integration of technologies into high-stakes cargo operations poses complex accountability challenges. Human operators can become moral crumple zones, absorbing legal liability for system failures they neither caused nor could have predicted (Elish, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Although this phenomenon is most evident with opaque systems, similar patterns could appear throughout the technology spectrum. A key difference lies between liability, which refers to legal responsibility, and answerability, which involves the practical capacity to justify decisions (Novelli et al., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). This distinction is important across all levels of technological mediation, though the gap between liability and answerability may widen with technological sophistication.\u003c/p\u003e \u003cp\u003eResponsibility depends on understanding how systems make decisions. When operators cannot explain their technical reasoning, a gap emerges between accountability and answerability, which varies with levels of technological sophistication (Coeckelbergh, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). Users of algorithmic systems face opacity that erodes answerability, while liability remains. Basic technology users share responsibility with transparent systems. Manual practitioners retain clear accountability, but transparency can still hinder accountability under time constraints.\u003c/p\u003e \u003cp\u003eErrors in complex automated systems are often unforeseeable to operators who approve plans without understanding system logic, yet remain legally liable if incidents occur (Fraser \u0026amp; Suzor, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). This creates a professional double bind. Verifying system recommendations is often impossible due to opacity and time limits, yet blindly trusting them risks liability (Johnson \u0026amp; Powers, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2005\u003c/span\u003e). Operators sign off on system plans, cargo manifests, and routing to meet efficiency targets, holding final authority but lacking control or understanding of system reasoning.\u003c/p\u003e \u003cp\u003eResearch shows that this dynamic transforms work (Pachidi et al., \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Organisations may appear to adopt systems for legitimacy, while operators rely on human expertise for decisions. This symbolic adoption differs from substantive adoption, causing tensions between formal procedures and actual practices. Accountability varies across levels of technological sophistication, affecting operators' professional agency and identity. Key questions include whether operators are becoming strategic managers or mere risk buffers. Understanding how workers navigate efficiency pressures, liability, and their professional identity amid technological mediation is crucial for sustainable work design.\u003c/p\u003e \u003cp\u003e \u003cstrong\u003eRQ3\u003c/strong\u003e \u003cp\u003eHow do operators experience accountability when working with technology systems?\u003c/p\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"3. Methodology","content":"\u003cp\u003eThis study employs a qualitative research design to investigate distributed cognitive processes in human-technology interaction within air cargo operations. Given the complex socio-technical nature of phenomena involving sensemaking, trust, and accountability across diverse technological contexts, qualitative approaches prove essential for exploring participant lived experiences and organisational practices (Creswell \u0026amp; Poth, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). The study adopts the gate-to-gate distributed cognition framework, recognising that cognitive work in aviation is distributed across human agents, technological artefacts, and environmental structures rather than residing in individual minds. This framework accommodates diverse levels of technological sophistication and examines how cognitive distribution patterns vary with it.\u003c/p\u003e \u003cp\u003eData analysis employs Qualitative Content Analysis (Schreier, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2012\u003c/span\u003e), combining concept-driven deductive categories from the theoretical framework with data-driven inductive categories emerging from participant narratives. This hybrid approach enables a systematic examination of theoretically predicted patterns while remaining open to unexpected findings across the spectrum of technological sophistication.\u003c/p\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003e3.1 Participants and Sampling\u003c/h2\u003e \u003cp\u003eParticipants were selected through purposive sampling to represent diverse positions along the technology adoption spectrum. The sample consists of six air cargo operations professionals distributed across three adoption categories: advanced users, basic users, and primarily manual practitioners. This distribution captures meaningful variation in technological integration while achieving thematic saturation within each adoption category. All participants work in operational roles, such as load planning, ramp coordination, or operations supervision, with varying levels of experience with technology systems. Recruitment proceeded through professional industry networks, ensuring voluntary participation and informed consent.\u003c/p\u003e \u003cp\u003eAlthough small, this sample size is appropriate for the study's goals. In qualitative research on technology adoption, six to twelve participants suffice with information-rich sampling and category saturation (Creswell \u0026amp; Poth, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). The two participants per level (advanced, basic, manual) enable comparisons within categories while preserving depth. This approach emphasises theoretical saturation over numerical representativeness (Schreier, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2012\u003c/span\u003e). Saturation occurred when no new codes or themes emerged, and cross-level patterns stabilised after analysing all six transcripts.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003e3.2 Data Collection\u003c/h2\u003e \u003cp\u003eSemi-structured, in-depth interviews lasting thirty to forty-five minutes were conducted via videoconference. The interview protocol explored three dimensions aligned with research questions. First, inspired by the Event Analysis of Systemic Teamwork methodology (Stanton, Salmon, \u0026amp; Walker, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2018\u003c/span\u003e), participants described the information sources they consult when resolving ambiguities and the communication patterns they use when technology outputs conflict with operational realities. This revealed variations in information-seeking strategies and collaborative sensemaking practices across levels of technological sophistication.\u003c/p\u003e \u003cp\u003eSecond, vignette-based scenarios probed trust calibration and accountability navigation. Scenarios presented situations involving data-system mismatches, organisational pressure under time constraints, and accountability for technology-generated decisions. These scenarios were designed to be applicable across levels of technological sophistication, examining how operators with different technological tools approach similar operational challenges.\u003c/p\u003e \u003cp\u003eAll interviews were conducted in Thai, audio recorded with written consent, and transcribed verbatim.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003e3.3 Data Analysis\u003c/h2\u003e \u003cp\u003eTranscribed data were analysed using Qualitative Content Analysis, with particular attention to variation across levels of technological sophistication. The coding frame combined deductive main categories from theoretical pillars, including Collective Sensemaking, Trust Calibration, and Accountability Navigation, with inductive subcategories emerging from data. Expected themes include discursive sensemaking practices (Viktorelius \u0026amp; Larsson, \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2025\u003c/span\u003e), cognitive collision between human intuition and technological logic (Hao et al., \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2025\u003c/span\u003e), axiological evaluation based on value alignment (Afroogh, Akbari, Malone, Kargar, \u0026amp; Alambeigi, \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2024\u003c/span\u003e), data repair labour (Faraj et al., \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Pachidi et al., \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2021\u003c/span\u003e), and answerability gaps where operators cannot explain decisions for which they remain liable (Coeckelbergh, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Novelli et al., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eAnalysis proceeded iteratively: first, coding within technology-level categories to identify level-specific patterns; then, conducting cross-level comparisons to identify common themes and systematic variations. The coding frame was refined iteratively to ensure that categories were mutually exclusive and exhaustive while capturing meaningful differences across the technology adoption spectrum.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003e3.4 Validity and Reliability\u003c/h2\u003e \u003cp\u003eMultiple validation strategies ensured rigour (Creswell \u0026amp; Poth, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Schreier, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2012\u003c/span\u003e). Reflexivity was maintained through the researcher's journaling to monitor assumptions about technology adoption, effectiveness, and operator agency.\u003c/p\u003e \u003cp\u003eInter-rater reliability was established through double coding, in which a second trained analyst independently coded 20% of transcripts (two complete transcripts representing advanced algorithmic systems and manual operation contexts). Cohen's kappa coefficient was calculated to assess inter-rater agreement, yielding κ\u0026thinsp;=\u0026thinsp;0.69, indicating substantial agreement beyond chance (Gisev et al., \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2013\u003c/span\u003e). Discrepancies were resolved through dialogue and iterative refinement of coding definitions.\u003c/p\u003e \u003c/div\u003e"},{"header":"4. Results","content":"\u003cp\u003eAnalysis of six air cargo professionals revealed three overarching themes. Advanced users (P1, P4) employed technology-driven systems; basic users (P5, P6) relied on limited-functionality software, and minimal-technology users (P2, P3) worked primarily manually. Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e presents the coding framework.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eAnalysis results\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTheme\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDefinition\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRepresentative Data\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colspan=\"3\" nameend=\"c3\" namest=\"c1\"\u003e \u003cp\u003eTHEME 1: Persistence of Human Expertise\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePhysical-Digital Gap\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSystematic mismatches requiring intervention\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u0026lsquo;If it's a garment, it bulges\u0026lsquo; (P1)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eInvisible Data Repair\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eContinuous correction work\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u0026lsquo;take the actual weight and close it in the system again.\u0026rsquo; (P1)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eEmbodied Knowledge\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePhysical inspection as irreplaceable\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u0026lsquo;AI can't walk and look\u0026rsquo; (P3)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c3\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eTHEME 2: Fragility of Trust\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCalibrated Scepticism\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eConditional trust despite experience\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u0026lsquo;Trust it, but check\u0026rsquo; (P4)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSymbolic Trust\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eFormal usage masking actual reliance\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u0026lsquo;Don't trust it\u0026rsquo; (P5)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAxiological Rejection\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eValue-based refusal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u0026lsquo;Don't trust 100 per cent on safety\u0026rsquo; (P3)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c3\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eTHEME 3: Erosion of Answerability\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDocumentation Shield\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRecords deflecting accountability\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u0026lsquo;at least we have a record that the system warned\u0026rsquo; (P1)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFear of Blame\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eIndividual accountability anxiety\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u0026lsquo;Might get reprimanded\u0026rsquo; (P6)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003e4.1 Persistence of Human Expertise\u003c/h2\u003e \u003cdiv id=\"Sec14\" class=\"Section3\"\u003e \u003ch2\u003e4.1.1 Physical-Digital Gap\u003c/h2\u003e \u003cp\u003ePhysical-digital gaps manifested across all technology levels. Advanced users reported systematic calculation failures. P1 described garment cargo bulging beyond declared dimensions, with P4 reporting correctly dimensioned pallets failing to stack due to warping. Basic users experienced parallel gaps. P5 characterised inaccuracy as a Thailand-specific limitation, whereas P6 stated that actual weights frequently exceed system predictions, necessitating manual verification. Manual users avoided miscalculations through direct assessment. P2 explained that physical inspection reveals cargo conditions that systems cannot detect.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section3\"\u003e \u003ch2\u003e4.1.2 Invisible Data Repair Labour\u003c/h2\u003e \u003cp\u003ePhysical-digital gaps generated continuous correction work, remaining organisationally invisible across technology levels. Advanced users performed algorithmic training through repeated corrections. P1 described operators taking the actual weight and re-entering it into the system, while P4 reported constant adjustments to align with physical reality. Basic users validated system outputs through physical verification. P5 noted that operations must be suspended and reviewed to verify that the build-up aligns with the plan, while P6 explained that system recommendations require verification before implementation. Manual users sustained operations through continuous assessment. P2 stated that every shipment requires a physical inspection, whereas P3 described repeatedly checking cargo security and balance during loading.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section3\"\u003e \u003ch2\u003e4.1.3 Embodied Knowledge as Irreplaceable Resource\u003c/h2\u003e \u003cp\u003eP3 stated, \u0026lsquo;AI can't come walk and look,\u0026rsquo; emphasising physical presence as knowledge beyond algorithms. The verbs \u0026lsquo;walk\u0026rsquo; and \u0026lsquo;look\u0026rsquo; highlight irreplaceable embodied expertise. P2 noted that physical checks reveal cargo conditions that systems can't detect. Advanced users rely on physical inspection despite sophisticated algorithms. P1 and P4 said they maintain embodied expertise alongside algorithms, not replace it. Basic users favour physical knowledge over limited system capabilities. P5 said physical assessment provides information that quality systems can't, while P6 preferred physical inspection over system outputs. This reflects both operational necessity and a defence of professional identity.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003e4.2 Fragility of Trust\u003c/h2\u003e \u003cdiv id=\"Sec18\" class=\"Section3\"\u003e \u003ch2\u003e4.2.1 Calibrated Scepticism\u003c/h2\u003e \u003cp\u003eon patterns varied systematically among user levels. Advanced users exhibited nuanced scepticism: P4 trusted the system due to long-term use but still checked, whereas P1 believed the system was problem-free but acknowledged user-reported issues, revealing defensive trust. Basic users had different behaviours. P5 viewed systems as guidelines, with final judgement left to humans, whereas P6 followed procedures but doubted the accuracy of the systems. Manual users completely rejected algorithmic trust. P2 believed trust should only be placed in human judgement, while P3 outright refused to trust technology, especially regarding safety.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec19\" class=\"Section3\"\u003e \u003ch2\u003e4.2.2 Symbolic Trust and Organisational Illusion\u003c/h2\u003e \u003cp\u003eBasic users demonstrated the largest divergence between organisational expectations and actual reliance. P5 stated, \u0026lsquo;The company asks us to use it as a guideline... basically, we don't trust it,\u0026lsquo; while organisations require system usage. This symbolic trust creates an organisational illusion in which management perceives technology-driven operations, while practitioners maintain human-centred decision-making. P1 noted pressures that force \u0026lsquo;use it completely or not at all\u0026rsquo; choices, thereby eliminating calibration space.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec20\" class=\"Section3\"\u003e \u003ch2\u003e4.2.3 Axiological Rejection\u003c/h2\u003e \u003cp\u003eManual users rejected algorithmic trust based on principles of safety culture rather than performance assessment. P3 refused to rely completely on technology for safety-critical decisions, regardless of system reliability. This value-based rejection operated independently of empirical system performance, grounding distrust in professional ethics rather than pragmatic evaluation. System improvements cannot address axiologically grounded distrust because rejection stems from core safety principles rather than technical limitations.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec21\" class=\"Section2\"\u003e \u003ch2\u003e4.3 Erosion of Answerability\u003c/h2\u003e \u003cdiv id=\"Sec22\" class=\"Section3\"\u003e \u003ch2\u003e4.3.1 Documentation as a Liability Shield\u003c/h2\u003e \u003cp\u003eAdvanced users employ documentation defensively. P1's statement, \u0026lsquo;at least we have a record that the system warned,\u0026lsquo; frames documentation as liability protection rather than decision support. The qualifier \u0026lsquo;at least\u0026rsquo; concedes insufficient understanding while asserting procedural compliance. P1 described mandatory warning compliance, under which operators may not question warnings but remain responsible for their accuracy. P4 observed that distributed authority spreads responsibility, yet individuals still retain accountability.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec23\" class=\"Section3\"\u003e \u003ch2\u003e4.3.2 Fear of Blame\u003c/h2\u003e \u003cp\u003eFear of individual blame manifested across technology levels, though mechanisms differed. Basic users expressed anxiety despite procedural compliance. P6 stated that they might be reprimanded even when following system recommendations, while P5 expressed concern about being held responsible for system errors. Advanced users anticipated accountability despite distributed authority. P1 noted that operators bear responsibility when systems provide incorrect recommendations, while P4 reported that investigations focus on human inputs rather than algorithmic limitations. Manual users faced direct accountability. P2 explained that they accepted full responsibility for decisions made through physical assessment, while P3 stated that clear accountability accompanies manual expertise.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e"},{"header":"5. Discussion","content":"\u003cdiv id=\"Sec25\" class=\"Section2\"\u003e \u003ch2\u003e5.1 Theoretical Contributions\u003c/h2\u003e \u003cp\u003eThe results support Situated Action Theory, which views technology-generated plans as flexible resources rather than strict scripts (Napoleone et al., \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Suchman, \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2007\u003c/span\u003e). Our findings show that the difficulty of operational gaps rises with the level of technology. Basic users mostly fix clear database errors. In contrast, advanced users must address subtle blind spots in which the software fails to account for physical properties. This confirms that as systems become more rigid, operators need deeper knowledge to resolve conflicts. Both require continuous correction work, demonstrating that expertise shifts rather than disappears. This suggests the relationship between technology and expertise is transformative rather than substitutive (Berente et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). The transformation is paradoxical: more advanced technology requires more sophisticated human intervention, not less. P1's description of garments that categorically bulge, yet systems sometimes fail to account for this, reveals linguistic masking of structural limitations through the framing of randomness. Operators describe predictable material behaviour failures as occasional system oversights, obscuring the fundamental gap between algorithmic calculation and physical reality.\u003c/p\u003e \u003cp\u003eThe persistence of correction work demonstrates the continued need for expertise, often unseen, representing invisible labour (Star \u0026amp; Strauss, \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2016\u003c/span\u003e), in which cycles are regarded as part of system training rather than as crucial expertise. Pachidi et al. (\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2021\u003c/span\u003e) noted that operators assume the role of algorithm trainers; however, our findings indicate that this shift occurs without professional acknowledgment or role redefinition. Organisations treat correction as a temporary inefficiency requiring elimination rather than a permanent requirement demanding recognition. When such expertise is unrecognised, organisations tend to set unrealistic expectations for efficiency while simultaneously devaluing the human contributions that make algorithmic systems function. Consequently, performance metrics presume seamless data integration, which forces workers to absorb the temporal cost of reconciling digital errors without official credit. This structural misalignment leads management to misattribute skilled intervention to operational friction, thereby penalising operators for the very competence that prevents systemic failure. In addition, embodied knowledge highlights the limitations of replacing human skill with algorithms (Stanton et al., \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2019\u003c/span\u003e), functioning both as essential operational knowledge enabling cargo verification algorithms to perform and as a safeguard for professional identity against perceived displacement.\u003c/p\u003e \u003cp\u003eFindings addressing how trust develops fragility under operational constraints complicate the calibration model through three mechanisms (Lee \u0026amp; See, \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2004\u003c/span\u003e). Initially, system design constraints enforce binary functionality, meaning algorithmic outputs are either fully implemented or not at all. Instead of relying solely on binary trust, operators navigate these limitations by symbolically adhering to system requirements, with actual decisions being driven by human judgement and experience (Pachidi et al., \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Organisations require system usage for technological legitimacy, operators comply symbolically, while actual decisions rest on human judgement. This creates an organisational illusion in which management perceives technology-driven operations, while practitioners maintain human-centred decision-making.\u003c/p\u003e \u003cp\u003eWhen symbolic and substantive trust diverge under these constraints, organisations lack an accurate understanding of decision-making processes, thereby increasing risk exposure during system failures when assumed algorithmic safeguards prove illusory. Second, axiological rejection operates independently of performance (Afroogh et al., \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). P3's categorical refusal to \u0026lsquo;trust 100 per cent\u0026rsquo; in matters of safety reflects professional values that prioritise human primacy in high-stakes decisions (Dekker, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2016\u003c/span\u003e) rather than algorithm aversion. Complex system failures arise from small deviations that accumulate over time; maintaining human verification despite system sophistication provides crucial layers of defence regardless of system reliability. Third, accountability asymmetry hinders the stabilisationn of trust. Algorithms give recommendations but bear no liability for failure. This leads operators to adopt a defensive stance, rejecting valid outputs not because of doubt, but because they can't justify the algorithm's logic when errors occur. As a result, trust depends not on system reliability but on the operator's ability to justify decisions to superiors.\u003c/p\u003e \u003cp\u003eOur findings demonstrate that technology erodes answerability, effectively extending the moral crumple zone. This occurs through distinct mechanisms where documentation serves as a liability shield rather than a tool for explanation, providing procedural proof without substantive justification. Furthermore, while decision-making authority is distributed between humans and machines, blame remains concentrated on the individual. This disparity causes anxiety, as operators fear reprimand for uncontrollable system outputs. Although they have formal authority, their understanding of how to justify their decisions is lacking. This traps them in a double bind of responsibility without comprehension (Johnson \u0026amp; Powers, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2005\u003c/span\u003e; Novelli et al., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). The consequences are twofold: organisationally, the inability to explain decisions precludes learning from failures, while individually, professional identity devolves from expert to system manager in the absence of institutional support, leaving operators to navigate role ambiguity alone.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec26\" class=\"Section2\"\u003e \u003ch2\u003e5.2 Practical Implications\u003c/h2\u003e \u003cp\u003eOrganisations must recognise data correction work in formal job descriptions, performance evaluations, and compensation structures. Current invisibility systematically devalues expertise sustaining algorithmic functionality. When correction work lacks institutional recognition, organisations develop unrealistic expectations of efficiency, while skilled operators exit roles that fail to acknowledge their actual responsibilities. This produces two failures. First, the quality of correction work declines as expertise attrition accelerates. Second, system failures increase because operators who perform unrecognised work lack incentives to maintain correction standards.\u003c/p\u003e \u003cp\u003eBeyond recognising correction work, verification protocols must accommodate operational time constraints rather than impose binary trust requirements. The current design assumes operators possess deliberative space for calibration. Operational reality eliminates this space, forcing complete reliance or complete rejection. Both extremes compromise safety. Blind trust allows uncorrected errors to propagate. Complete rejection wastes system capabilities and increases manual workload. Integrating protocols directly into workflows, rather than treating verification as a deviation from standard procedures, enables graduated reliance appropriate to operational constraints.\u003c/p\u003e \u003cp\u003eTrust calibration becomes meaningless without corresponding accountability alignment. Accountability structures currently assign operators responsibility for algorithmic outputs they cannot explain or override. This produces three problems. Investigations focus on operator inputs rather than system design limitations, preventing organisational learning. Operators experience professional anxiety from bearing liability without corresponding authority. Documentation shifts from decision explanation to liability protection, obscuring actual reasoning when failures occur. Regulatory frameworks should require accountability, the distribution of decision-making authority, and the alignment of decision-making authority with responsibilities. When systems make critical determinations, designers share liability in proportion to their control over the algorithmic logic. When operators bear full responsibility, they receive override authority commensurate with assigned accountability.\u003c/p\u003e \u003cp\u003eThese structural changes must account for fundamental information boundaries that constrain algorithmic systems. Embodied knowledge provides information that algorithmic systems cannot capture. Physical cargo assessment reveals material properties and condition characteristics that are difficult to digitise. Organisations that position physical expertise as obsolete lose essential capabilities for detecting when algorithmic outputs diverge from material reality. System design should support human-algorithm collaboration, leveraging complementary capabilities rather than assuming complete digitisationn is achievable. Career pathways must value physical expertise alongside data analysis competencies. Current approaches that treat embodied knowledge as transient skill, requiring its elimination through automation, ignore persistent information boundaries that constrain the algorithmic capture of such knowledge.\u003c/p\u003e \u003c/div\u003e"},{"header":"6. Conclusion, Limitations, and Future Directions","content":"\u003cp\u003eTechnology adoption transforms rather than eliminates human work, creates trust dilemmas rather than resolving uncertainties, and obscures rather than clarifies responsibility. Physical-digital gaps require continuous intervention; correction work remains invisible yet essential; and embodied knowledge is irreplaceable. Trust requires active construction, in which calibrated scepticism represents achievement, symbolic trust masks authority, and axiological rejection operates independently of performance. Answerability erosion creates vulnerabilities through documentation shields, distributed authority, and unrecognised transformation. As technology sophistication increases, the requirements for human expertise intensify, while organisational visibility decreases. Sustainable collaboration requires acknowledging essential expertise, supporting appropriate calibration, and establishing frameworks matching actual authority.\u003c/p\u003e \u003cp\u003eStudy limitations include a small sample size (n\u0026thinsp;=\u0026thinsp;6), a focus on theoretical saturation rather than statistical breadth, a geographic concentration that emphasises contextual variables, and a cross-sectional design that captures a single point in time. Despite these, the study offers valuable insights. Participant 5's view of dimensional inaccuracy as region-specific indicates that local conditions influence technology adoption, challenging assumptions about the universal effectiveness of automated systems. Future research should include comparative analyses across high-stakes areas, longitudinal studies of adaptation, and ethnographies to explore hidden labour. Developing accountability frameworks is vital to ensure accountability. Solving the paradox in which automation increases expertise demands yet reduces visibility requires institutional innovation to meet the demand for new competencies.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eFunding:\u003c/strong\u003e The authors did not receive any funding support for the submitted work. No funding was received for conducting this study.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting Interests:\u003c/strong\u003e The authors have no competing interests to declare that are relevant to the content of this article. The authors have no relevant financial or non-financial interests to disclose.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics Approval:\u003c/strong\u003e This study was conducted in accordance with ethical standards for research involving human participants. All procedures were approved by the author\u0026rsquo;s ethics committee. All participants provided informed written consent prior to data collection and were assured of confidentiality and voluntary participation.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData Availability:\u003c/strong\u003e The data that support the findings of this study consist of interview transcripts collected from individual participants. These data are not publicly available due to participant privacy and confidentiality constraints. The data cannot be fully anonymised without compromising the richness required for qualitative analysis. Researchers who wish to discuss access conditions may contact the corresponding author.\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eT.P.: Conceptualization, Investigation, Data curation, Validation, Writing \u0026ndash; original draft, Supervision. T.T.: Conceptualization, Methodology, Data curation, Formal analysis, Validation, Writing \u0026ndash; original draft.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eAfroogh, S., Akbari, A., Malone, E., Kargar, M., \u0026amp; Alambeigi, H. (2024). Trust in AI: Progress, challenges, and future directions. Humanities and Social Sciences Communications, 11 (1), 1568. In.\u003c/li\u003e\n\u003cli\u003eBerente, N., Gu, B., Recker, J., \u0026amp; Santhanam, R. (2021). Managing artificial intelligence. In (Vol. 45, pp. 1433-1450): Management Information Systems Research Center, University of Minnesota.\u003c/li\u003e\n\u003cli\u003eCoeckelbergh, M. (2020). Artificial intelligence, responsibility attribution, and a relational justification of explainability. \u003cem\u003eScience and Engineering Ethics, 26\u003c/em\u003e(4), 2051-2068. \u003c/li\u003e\n\u003cli\u003eCreswell, J. W., \u0026amp; Poth, C. N. (2016). \u003cem\u003eQualitative Inquiry and Research Design: Choosing Among Five Approaches\u003c/em\u003e: SAGE Publications.\u003c/li\u003e\n\u003cli\u003eDekker, S. (2016). \u003cem\u003eDrift into failure: From hunting broken components to understanding complex systems\u003c/em\u003e: CRC press.\u003c/li\u003e\n\u003cli\u003eDesai, J., Srivathsan, S., Lai, W. Y., Li, L., \u0026amp; Yu, C. (2023). An optimization-based decision support tool for air cargo loading. \u003cem\u003eComputers \u0026amp; Industrial Engineering, 175\u003c/em\u003e, 108816. doi:https://doi.org/10.1016/j.cie.2022.108816\u003c/li\u003e\n\u003cli\u003eElish, M. C. (2025). Moral crumple zones: cautionary tales in human\u0026ndash;robot interaction. In \u003cem\u003eRobot Law: Volume II\u003c/em\u003e (pp. 83-105): Edward Elgar Publishing.\u003c/li\u003e\n\u003cli\u003eFaraj, S., Pachidi, S., \u0026amp; Sayegh, K. (2018). Working and organising in the age of the learning algorithm. \u003cem\u003eInformation and Organization, 28\u003c/em\u003e(1), 62-70. \u003c/li\u003e\n\u003cli\u003eFraser, H. L., \u0026amp; Suzor, N. P. (2025). Locating fault for AI harms: a systems theory of foreseeability, reasonable care and causal responsibility in the AI value chain. \u003cem\u003eLaw, Innovation and Technology, 17\u003c/em\u003e(1), 103-138. \u003c/li\u003e\n\u003cli\u003eGillespie, T. (2022). Building trust and responsibility into autonomous human-machine teams. \u003cem\u003eFrontiers in Physics, 10\u003c/em\u003e, 942245. \u003c/li\u003e\n\u003cli\u003eGisev, N., Bell, J. S., \u0026amp; Chen, T. F. (2013). Interrater agreement and interrater reliability: key concepts, approaches, and applications. \u003cem\u003eResearch in Social and Administrative Pharmacy, 9\u003c/em\u003e(3), 330-338. \u003c/li\u003e\n\u003cli\u003eHao, X., Demir, E., \u0026amp; Eyers, D. (2025). Beyond human-in-the-loop: Sensemaking between artificial intelligence and human intelligence collaboration. \u003cem\u003eSustainable Futures, 10\u003c/em\u003e, 101152. \u003c/li\u003e\n\u003cli\u003eHollnagel, E. (2017). \u003cem\u003eThe ETTO principle: efficiency-thoroughness trade-off: why things that go right sometimes go wrong\u003c/em\u003e: CRC press.\u003c/li\u003e\n\u003cli\u003eJohnson, D. G., \u0026amp; Powers, T. M. (2005). Computer systems and responsibility: A normative look at technological complexity. \u003cem\u003eEthics and information technology, 7\u003c/em\u003e(2), 99-107. \u003c/li\u003e\n\u003cli\u003eKlein, G., Moon, B., \u0026amp; Hoffman, R. R. (2006). Making sense of sensemaking 1: Alternative perspectives. \u003cem\u003eIEEE intelligent systems, 21\u003c/em\u003e(4), 70-73. \u003c/li\u003e\n\u003cli\u003eLau, H. C. W., Chan, T. M., Tsui, W. T., Ho, G. T. S., \u0026amp; Choy, K. L. (2009). An AI approach for optimizing multi-pallet loading operations. \u003cem\u003eExpert Systems with Applications, 36\u003c/em\u003e(3, Part 1), 4296-4312. doi:https://doi.org/10.1016/j.eswa.2008.03.024\u003c/li\u003e\n\u003cli\u003eLebovitz, S., Lifshitz-Assaf, H., \u0026amp; Levina, N. (2022). To engage or not to engage with AI for critical judgments: How professionals deal with opacity when using AI for medical diagnosis. \u003cem\u003eOrganization science, 33\u003c/em\u003e(1), 126-148. \u003c/li\u003e\n\u003cli\u003eLee, J. D., \u0026amp; See, K. A. (2004). Trust in automation: Designing for appropriate reliance. \u003cem\u003eHuman factors, 46\u003c/em\u003e(1), 50-80. \u003c/li\u003e\n\u003cli\u003eMosier, K. L., Fischer, U., Burian, B. K., \u0026amp; Kochan, J. A. (2017). Autonomous, context-sensitive, task management systems and decision support tools I: Human-autonomy teaming fundamentals and state of the art. \u003c/li\u003e\n\u003cli\u003eNapoleone, A., Andersen, A.-L., Brunoe, T. D., \u0026amp; Nielsen, K. (2023). Towards human-centric reconfigurable manufacturing systems: Literature review of reconfigurability enablers for reduced reconfiguration effort and classification frameworks. \u003cem\u003eJournal of Manufacturing Systems, 67\u003c/em\u003e, 23-34. \u003c/li\u003e\n\u003cli\u003eNovelli, C., Taddeo, M., \u0026amp; Floridi, L. (2024). Accountability in artificial intelligence: What it is and how it works. \u003cem\u003eAi \u0026amp; Society, 39\u003c/em\u003e(4), 1871-1882. \u003c/li\u003e\n\u003cli\u003eOksman, V., Kulju, M., \u0026amp; Kramar, V. (2024). \u003cem\u003eExploring the Factors Influencing Public Acceptance and Intention to Use Automated Drone Delivery Systems: A Trust Theory Approach\u003c/em\u003e.\u003c/li\u003e\n\u003cli\u003ePachidi, S., Berends, H., Faraj, S., \u0026amp; Huysman, M. (2021). Make way for the algorithms: Symbolic actions and change in a regime of knowing. \u003cem\u003eOrganization science, 32\u003c/em\u003e(1), 18-41. \u003c/li\u003e\n\u003cli\u003eParasuraman, R., \u0026amp; Riley, V. (1997). Humans and automation: Use, misuse, disuse, abuse. \u003cem\u003eHuman factors, 39\u003c/em\u003e(2), 230-253. \u003c/li\u003e\n\u003cli\u003eRogers, E. M., Singhal, A., \u0026amp; Quinlan, M. M. (2014). Diffusion of innovations. In \u003cem\u003eAn integrated approach to communication theory and research\u003c/em\u003e (pp. 432-448): Routledge.\u003c/li\u003e\n\u003cli\u003eSchemmer, M., Hemmer, P., Nitsche, M., K\u0026uuml;hl, N., \u0026amp; V\u0026ouml;ssing, M. (2022). \u003cem\u003eA meta-analysis of the utility of explainable artificial intelligence in human-AI decision-making.\u003c/em\u003e Paper presented at the Proceedings of the 2022 AAAI/ACM Conference on AI, Ethics, and Society.\u003c/li\u003e\n\u003cli\u003eSchreier, M. (2012). \u003cem\u003eQualitative Content Analysis in Practice\u003c/em\u003e: SAGE Publications.\u003c/li\u003e\n\u003cli\u003eStanton, N. A., Plant, K. L., Revell, K. M., Griffin, T. G., Moffat, S., \u0026amp; Stanton, M. (2019). Distributed cognition in aviation operations: a gate-to-gate study with implications for distributed crewing. \u003cem\u003eErgonomics, 62\u003c/em\u003e(2), 138-155. \u003c/li\u003e\n\u003cli\u003eStanton, N. A., Salmon, P., \u0026amp; Walker, G. H. (2018). \u003cem\u003eSystems thinking in practice: Applications of the event analysis of systemic teamwork method\u003c/em\u003e: CRC Press.\u003c/li\u003e\n\u003cli\u003eStar, S. L., \u0026amp; Strauss, A. (2016). 19 Layers of Silence, Arenas of Voice: The Ecology of Visible and Invisible Work. \u003cem\u003eBoundary Objects and Beyond: Working with Leigh Star\u003c/em\u003e, 351. \u003c/li\u003e\n\u003cli\u003eSuchman, L. A. (2007). \u003cem\u003eHuman-machine reconfigurations: Plans and situated actions\u003c/em\u003e: Cambridge university press.\u003c/li\u003e\n\u003cli\u003eTsolakis, N., Schumacher, R., Dora, M., \u0026amp; Kumar, M. (2023). Artificial intelligence and blockchain implementation in supply chains: a pathway to sustainability and data monetisation? \u003cem\u003eAnnals of Operations Research, 327\u003c/em\u003e(1), 157-210. \u003c/li\u003e\n\u003cli\u003eTurner, J. R., Allen, J., Hawamdeh, S., \u0026amp; Mastanamma, G. (2023). The multifaceted sensemaking theory: A systematic literature review and content analysis on sensemaking. \u003cem\u003eSystems, 11\u003c/em\u003e(3), 145. \u003c/li\u003e\n\u003cli\u003eViktorelius, M., \u0026amp; Larsson, S. (2025). Talking about machines: discursive sensemaking and workplace learning in the age of AI. \u003cem\u003eStudies in Continuing Education, 47\u003c/em\u003e(3), 634-652. \u003c/li\u003e\n\u003cli\u003eWeick, K. E., Sutcliffe, K. M., \u0026amp; Obstfeld, D. (2005). Organising and the process of sensemaking. \u003cem\u003eOrganization science, 16\u003c/em\u003e(4), 409-421. \u003c/li\u003e\n\u003cli\u003eWright, J. L., Chen, J. Y., \u0026amp; Lakhmani, S. G. (2019). Agent transparency and reliability in human\u0026ndash;robot interaction: The influence on user confidence and perceived reliability. \u003cem\u003eIEEE Transactions on Human-Machine Systems, 50\u003c/em\u003e(3), 254-263.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"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":"cognition-technology-and-work","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"ctwo","sideBox":"Learn more about [Cognition, Technology \u0026 Work](http://link.springer.com/journal/10111)","snPcode":"10111","submissionUrl":"https://submission.nature.com/new-submission/10111/3","title":"Cognition, Technology \u0026 Work","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Human-technology collaboration, trust in automation, distributed cognition, invisible labour, accountability","lastPublishedDoi":"10.21203/rs.3.rs-9160278/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9160278/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eAs algorithmic systems increasingly mediate safety-critical work, questions arise about how expertise, trust, and accountability are changing. This study examines air cargo operations in Thailand, where professionals navigate varying levels of technological sophistication, from manual processes to advanced algorithmic systems. Through semi-structured interviews with six cargo professionals, analysed using the distributed cognition framework, the research reveals a central paradox: increasing technological sophistication demands greater human expertise whilst reducing organisational visibility. Three patterns emerge. Physical\u0026ndash;digital gaps require continuous intervention yet remain organisationally invisible. Trust proves fragile, manifesting as calibrated scepticism, symbolic compliance, or axiological rejection. Answerability erodes as sophistication increases, with documentation functioning as a liability shield rather than a basis for learning. These findings demonstrate that the technology\u0026ndash;expertise relationship is transformative rather than substitutive, and that institutional structures have not adapted to this transformation. The study contributes to sociotechnical scholarship by examining how distributed cognition operates across the technology adoption spectrum in safety-critical operational domains.\u003c/p\u003e","manuscriptTitle":"Technology as Transformation: Persistence of Expertise, Fragility of Trust, and Erosion of Accountability in Air Cargo Operations","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-04-01 19:42:59","doi":"10.21203/rs.3.rs-9160278/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"reviewersInvited","content":"","date":"2026-03-30T10:18:48+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-03-23T05:06:29+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-03-23T05:06:06+00:00","index":"","fulltext":""},{"type":"submitted","content":"Cognition, Technology \u0026 Work","date":"2026-03-18T13:51:00+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"cognition-technology-and-work","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"ctwo","sideBox":"Learn more about [Cognition, Technology \u0026 Work](http://link.springer.com/journal/10111)","snPcode":"10111","submissionUrl":"https://submission.nature.com/new-submission/10111/3","title":"Cognition, Technology \u0026 Work","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"120b0fef-55c3-4225-82a3-2d720372695a","owner":[],"postedDate":"April 1st, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-04-01T19:43:00+00:00","versionOfRecord":[],"versionCreatedAt":"2026-04-01 19:42:59","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9160278","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9160278","identity":"rs-9160278","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}