Pith. sign in

REVIEW 4 major objections 5 minor 104 references

A Chain Is Only as Strong as Its Weakest Link: A Scoping Review of System Integration Audits in AI

T0 review · 4 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read This paper claims that AI auditing has a distinct, under-utilized target—system integration—and organizes the emerging practice into three integration sites and four audit functions.

desk verdict A genuinely useful taxonomy of integration-focused AI audits, wrapped in a scoping review whose corpus provenance needs to be opened up before the prevalence claims can be trusted. read the letter →

arxiv 2608.04921 v1 pith:5FICRNGV submitted 2026-08-05 cs.SE cs.AI

Leah Davis , Dominic Martin , AJung Moon This is my paper · ORCID
classification cs.SEcs.AI
keywords systemintegrationauditsAIauditingscopingreviewsitesriskexplorationproceduralregularitygeneral-purposesociotechnicalsystems
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This scoping review argues that AI audits looking beyond a single model—examining how components, environments, and whole systems interact—form an emerging but fragmented audit class. Reviewing 4,259 documents and retaining 58, the paper proposes a taxonomy of three integration sites (inter-component, system-environment, and multi-system) and four functions these audits serve (risk exploration, risk determination, coordination, procedural regularity). The authors claim that system integration is a distinct and under-utilized target for AI risk, that few current audits measure integration-specific qualities like compatibility, completeness, and oversight, and that access to information and resources is the strongest influence on audit design. If correct, this gives regulators and practitioners a structured way to demand audits that catch failures at the seams of AI systems rather than only inside models.

What carries the argument

The taxonomy of integration sites and audit functions is the central organizing device. It is built on a definition of systems as interconnected elements with a function or purpose and on the distinction between technical intermediaries (e.g., APIs, datasheets, simulators) and sociotechnical intermediaries (e.g., sandboxes, post-market surveillance, stakeholder feedback). The paper maps each site to a system boundary, a dominant intermediary type, typical roles, and targeted system types, and maps audit outputs to the four functions. This machinery lets the review translate scattered practices into a common language and reveal where coverage is missing.

What would settle it

Re-run the screening of the same 4,259 records with independent reviewers applying inclusion criterion 3, then compare the resulting corpus and the derived percentages; if the 58-document corpus cannot be reproduced or the three-site taxonomy does not stabilize across codings, the central claim would be undermined. A simpler check is to request the document-to-code mapping and audit the coding tree, which would be falsified if documents coded as multi-system integration do not consistently describe relationships between two or more AI systems.

Watch

Extended reading notes

Core claim

The central claim is that system integration audits exist in AI as an identifiable but immature practice, and that their conceptual core can be described by where they look and what they do. Where: inter-component integration (between pipeline elements such as data, model, output, and user interaction), system-environment integration (between one AI system and its deployment context or broader ecosystem), and multi-system integration (between two or more AI systems, typically upstream foundation models and downstream applications). What: four functions—risk exploration, risk determination, coordination, and procedural regularity—plus qualities specific to integration: compatibility (10.3% of the corpus), completeness (8.6%), and oversight (6.9%). The paper further claims that most of these audits fail traditional audit expectations: fewer than half name an auditor, most are not independent, few reference a standard, and access to information shapes design. The review concludes that integration should be prioritized as a core strategy for AI risk and that component-level evaluation alone cannot capture failures across components, environments, and systems.

Load-bearing premise

The whole corpus and taxonomy rest on the subjective judgment that a document 'examines the integration, interaction, or interconnections among AI systems beyond the model'; inter-coder agreement was measured only on the first 50 documents (82%), and the document-to-code mapping is not provided.

Editorial extensions

If this is right

  • System integration should be treated as a distinct audit target, with dedicated measures for compatibility, completeness, and oversight rather than model-level metrics.
  • Regulators and audit standards bodies can use the three-site taxonomy to specify where audits must look and whom they must involve.
  • Because multi-system integration of general-purpose, generative, and multimodal AI appears in under 30% of the corpus, supply-chain audits of foundation models are a priority gap.
  • Institutionalizing these audits requires solving information and resource access, since nearly 75% of the audits use internal approaches constrained by access.
  • Audits that meet traditional expectations—independent external auditors, adherence to a standard, and system-level evidence—remain rare, so the field currently resembles evaluation more than formal auditing.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Editorial inference: an unstated consequence of the paper's taxonomy is that it provides a checklist for auditing the auditors—a regulator could ask any AI audit to specify its integration site(s) and functions, and treat silence as a coverage gap.
  • Editorial inference: the 'weakest link' framing suggests that integration audits could be prioritized by mapping dependency chains and auditing the most coupled, least visible intermediaries (e.g., APIs, vendor-supplied model updates) rather than all components equally.
  • Editorial inference: a testable extension would be to apply the three-site taxonomy to a sample of model-centric audits and count how many would fail to detect failures at system-environment or multi-system boundaries; the paper's own data on the rarity of integration-specific qualities predicts most would fail.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper presents a PRISMA-ScR scoping review of AI audits that treat system integration as a core evaluation target. From 4,259 screened documents, the authors analyze 58 included studies using reflexive thematic analysis, and derive a taxonomy of three integration sites (inter-component, system-environment, and multi-system), four audit functions (risk exploration, risk determination, coordination, and procedural regularity), and a set of integration-specific qualities (notably compatibility, completeness, and oversight). The paper argues that system integration is an emerging but fragmented and under-institutionalized audit target in AI, and it compares the corpus against traditional audit expectations such as independent auditors and standardization.

Significance. The review addresses a genuine gap in the AI auditing literature, which has largely focused on model-level evaluation. If the corpus is representative, the proposed taxonomy offers a useful conceptual framework for system-level assurance and connects AI auditing to established practices in safety-critical engineering. The paper is transparent about its search strings, inclusion criteria, and analytical procedures, and it provides a reproducible search protocol. However, the quantitative claims about the prevalence and fragmentation of integration audits depend on the representativeness of the corpus and on the reliability of screening and coding; those dependencies are currently not fully documented.

major comments (4)
  1. [Section 3.1 / Appendix B] The manuscript adds 77 documents from the authors' existing collection, but Appendix B lists included studies without any source field, so a reader cannot determine how many of the 58 included documents entered through the manual route. Because the prevalence figures in Section 4 (e.g., 87.9% engaging data/input, 65.5% conceptual proposals, 29.3% GPAI/generative, and the 10.3%, 8.6%, and 6.9% prevalences of the integration-specific qualities) are computed over this pool, the claims that integration audits are 'emerging' and 'under-utilized' may reflect selection rather than a property of the wider literature. Please add a source tag to each entry in Appendix B and provide a sensitivity analysis restricted to database-identified documents, or substantively temper the prevalence-based conclusions.
  2. [Section 3.1] Inter-coder agreement (82%) was measured only on the first 50 documents, yet the lead reviewer screened all 4,259 titles and abstracts, and inclusion criterion 3 ('system integration') is highly subjective. The consistency of screening for the remaining records is unverified, which threatens the validity of the corpus composition. Please report a second agreement sample (e.g., on a random set of documents, including excluded ones) or provide a full dual-screening procedure for ambiguous cases, and consider making the document-to-code mapping available as a supplementary table.
  3. [Section 3.1 vs. Appendix B] The search was conducted in November 2024, but the included list contains items dated 2025 and 2026 (e.g., DeGrave et al. 2025, Mökander et al. 2025, and Sisodia 2026). Please clarify which included documents were identified through the manual/expert route and how they satisfy the search-window criterion; otherwise the systematic search is not reproducible as described.
  4. [Section 4] The paper reports precise percentages for roles, qualities, and outputs (e.g., 44.8% using the term 'auditor', 69.0% assessing 'technical performance', and 53.4% producing quantitative metrics) but provides no inter-coder reliability for the coding of the full 58-document corpus. If these numbers are used as evidence for 'fragmentation' and under-utilization, please report agreement statistics for the full coding set, or label the percentages as exploratory and refrain from presenting them as field-prevalence estimates.
minor comments (5)
  1. [Section 2.2] The sentence 'open-source, upstream GPAI model providers are subject to surprisingly relaxed regulatory expectations' uses 'surprisingly' redundantly with the previous sentence; consider rewording.
  2. [Section 3.1] The description of the screening process is ambiguous: 'The lead reviewer screened the titles and abstracts of all documents ... with the help of two other reviewers, who reviewed the same.' Please clarify whether the two other reviewers screened all documents or only a subset.
  3. [Appendix A] The table row for 'Articles from experts' lists 'Reached out to experts in their domain and personal inputs' as the selection condition, which is inconsistent with the main-text wording that the 77 documents were 'manually added from the authors’ existing collection'; please reconcile these descriptions.
  4. [Table 3] In the last row, the 'System Type Targeted' cell reads 'Predominantly general-purpose, generative, and physical AI systems' and omits the final period used in the other rows; also define 'physical AI systems' in the text.
  5. [Section 4.3 / References] The citation 'Sghaier et al (2024)' is missing a period after 'al', and the reference list contains inconsistent formatting of author initials across entries; a careful copyedit of the references would improve readability.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the review's taxonomy is an inductive coding summary of its own corpus, and the one self-citation is background evidence, not a load-bearing reduction.

full rationale

The derivation chain in this scoping review is descriptive rather than predictive: the corpus (n=58) is assembled by PRISMA-ScR screening (Section 3.1) and the taxonomy of integration sites, roles, qualities, and functions is produced by reflexive thematic analysis of those 58 documents. No equation is fitted and no parameter is estimated from a subset and then 'predicted' on a related quantity; the prevalence percentages in Section 4 are direct counts over the same corpus that was coded, and the paper does not claim them as out-of-sample predictions. The one self-citation (Rismani et al. 2025, which includes authors Davis and Moon) is used only as background evidence that component-level measures dominate ('over 57.3% of responsible AI measures target the model'), and the review's central taxonomy does not rest on that citation; the cited statistic is externally falsifiable and the same gap is independently visible in the review's own coding. The main structural concern is provenance: Section 3.1 reports 'an additional 77 documents were manually added from the authors' existing collection of AI auditing literature,' and Appendix B lists included studies without a source field, so the contribution of the non-systematic route to the final 58 cannot be audited. This is a selection-bias and reproducibility limitation, not a circular derivation: the findings remain descriptive of whatever corpus was assembled, and the screening criterion 'Relevance to system integration' does not by itself entail the three-site taxonomy, which is an inductive coding outcome rather than a consequence of the inclusion criteria. No circular step meeting the quoted-evidence standard was found.

Assumptions & free parameters 0 free parameters · 4 assumptions · 0 invented entities

This review is a qualitative synthesis. It has no fitted parameters or invented physical entities. The central claims rest on the adopted definition of system integration, the transferability of aerospace audit concepts to AI, the adequacy of the search strategy, and the stability of the thematic categories, all of which are domain assumptions made by the authors.

assumptions (4)
  • domain assumption The definition of system integration as breakdowns in the coordinated functioning of interconnected components, subsystems, and the actors responsible for them is appropriate for AI systems.
    Section 2.1 adopts Madni and Sievers' definition from systems engineering and applies it to AI auditing; this frames what counts as an integration audit and what is excluded.
  • domain assumption AI systems are sufficiently analogous to safety-critical engineered systems (e.g., aerospace) that integration audit concepts transfer.
    Section 1 and 2 use Boeing 737 MAX incidents to motivate the review and Section 2.1 compares aerospace audit conditions to AI; transferability is assumed, not demonstrated.
  • domain assumption The search queries and database selection capture the relevant literature on AI auditing.
    Section 3.1 and Appendix A show the queries; because 'system integration auditing' is not a used term, the authors rely on broad queries and manual screening, assuming this recovers the relevant corpus.
  • domain assumption Reflexive thematic analysis yields stable and meaningful categories.
    Section 3.2 describes the approach; the reliability of the final taxonomy is assumed rather than tested beyond the first 50 documents.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Chain Is Only as Strong as Its Weakest Link: A Scoping Review of System Integration Audits in AI." pith.science (2026). https://pith.science/paper/5FICRNGV

@misc{pith2026260804921,
  author       = {Pith},
  title        = {Pith review of: A Chain Is Only as Strong as Its Weakest Link: A Scoping Review of System Integration Audits in AI},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5FICRNGV}},
  note         = {Machine review of arXiv:2608.04921}
}
read the original abstract

As AI systems become increasingly integrated into diverse interfaces and applications, model-centric audits are insufficient to address risks arising from interactions among system components and deployment environments. System integration has long been central to software audits in safety-critical domains such as aerospace. However, its role in AI auditing remains underexplored. Scanning through 4,259 documents, we present a scoping review of AI audits that treat system integration as a core tenet of evaluation (n = 58). Using reflexive thematic analysis, we analyze their elements, actors, enablers, and constraints. We find that the corpus represents an emerging yet still fragmented form of AI auditing: few existing measures target integration-specific risks; large gaps remain in meeting traditional audit expectations; and access to necessary information and resources significantly influences audit design. Nonetheless, integration can be categorized across three sites (inter-component, system-environment, and multi-system), each serving the functions of risk exploration, risk determination, coordination, and procedural regularity. Deviating from other types of evaluations, these audits assess qualities specific to system integration, including compatibility, completeness, and oversight. This review calls on the AI community to prioritize system integration as a core strategy for addressing AI risk, and to develop audit practices capable of capturing failures across components, environments, and systems beyond the reach of component-level evaluation.

Figures

Figures reproduced from arXiv: 2608.04921 by the authors.

Figure 1
Figure 1. The Preferred Reporting Items for Systematic reviews and Meta-Analyses extension for Scoping Reviews flow chart [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Three system integration sites are identified in this review: (1) inter-component integration (solid lines), (2) system [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

104 extracted references · 57 canonical work pages

  1. [1]

    and Agirre, I

    Abella, J. and Agirre, I. and Bui, T. and Geujen, F. and Giordana, G. and Donzella, C. and Dallara, V. , year =. 2025 28th

    2025

  2. [2]

    Nature Scientific Reports , author =

    Clinical audit of an. Nature Scientific Reports , author =. 2024 , pages =. doi:10.1038/s41598-024-69314-6 , number =

    work page doi:10.1038/s41598-024-69314-6 2024

  3. [3]

    and Klaes, M

    Adler, R. and Klaes, M. , editor =. Assurance Cases as Foundation Stone for Auditing AI-Enabled and Autonomous Systems: Workshop Results and Political Recommendations for Action from the ExamAI Project , doi =. 2022 , pages =

    2022

  4. [4]

    Alagarswamy, D. P. and Berghoff, C. and Danos, V. and Langer, F. and Markert, T. and Schneider, G. and Twickel, A. von and Woitschek, F. , year =. Towards. doi:10.5220/0011619500003405 , booktitle =

    work page doi:10.5220/0011619500003405

  5. [5]

    Anthropic's

    Srishti Singh Sisodia , file =. Anthropic's. 2026 , journal =

    2026

  6. [6]

    Safety-Critical Systems , author =

    The. Safety-Critical Systems , author =

  7. [7]

    Frontiers in Digital Health , author =

    An integration engineering framework for machine learning in healthcare , volume =. Frontiers in Digital Health , author =. 2022 , pages =

    2022

  8. [9]

    and Bautista, J

    Barnard, P. and Bautista, J. R. and Krook, J. and Liu, A. and Menéndez, H. and Schmidt, A. and Sookoor, T. , year =. doi:10.1145/3597512.3597522 , booktitle =

    work page arXiv

Show all 104 references
  1. [10]

    and Waltl, B

    Becker, N. and Waltl, B. , editor =. Auditing and. 2022 , pages =. doi:10.1007/978-3-031-06018-2_20 , booktitle =

    2022 doi

  2. [11]

    2024 , doi =

    Pragmatic Auditing: A Pilot-Driven Approach for Auditing Machine Learning Systems , author =. 2024 , doi =

    2024

  3. [12]

    and Shekman, D

    Benthall, S. and Shekman, D. , year =. Designing. doi:10.1145/3617694.3623230 , journal =

  4. [13]

    , year =

    Bharadhwaj, H. , year =. Auditing. Proceedings of the 5th

  5. [14]

    2016 , journal =

    Evaluation of General-Purpose Artificial Intelligence: Why, What & How , author =. 2016 , journal =

    2016

  6. [15]

    and Steed, R

    Birhane, A. and Steed, R. and Ojewale, V. and Vecchione, B. and Raji, I. D. , year =. 2024

    2024

  7. [16]

    and Getir Yaman, S

    Boltz, N. and Getir Yaman, S. and Inverardi, P. and Lemos, R. and Landuyt, D. and Zisman, A. , year =. Human empowerment in self-adaptive socio-technical systems , doi =. Proceedings of the 19th

  8. [17]

    and Brown, N

    Brohan, A. and Brown, N. and Carbajal, J. and Chebotar, Y. and Chen, X. and Choromanski, K. and Zitkovich, B. , year =. Rt-2:. doi:https://arxiv.org/abs/2307.15818 , note =

    arXiv

  9. [18]

    Proceedings of the ACM on Human-Computer Interaction , author =

    Bargaining with the. Proceedings of the ACM on Human-Computer Interaction , author =. 2022 , pages =. doi:10.1145/3570601 , number =

    2022 doi

  10. [19]

    , year =

    Carey, S. , year =. Regulating. doi:10.1017/err.2025.10040 , journal =

    2025

  11. [20]

    2024 , pages =

    Proceedings of the AAAI Conference on Artificial Intelligence , author =. 2024 , pages =. doi:10.1609/aaai.v38i21.30346 , number =

    2024 doi

  12. [21]

    Information , author =

    Large language models in medical chatbots: opportunities, challenges, and the need to address. Information , author =. 2025 , pages =

    2025

  13. [22]

    and Alexander, E

    Correll, M. and Alexander, E. and Albers, D. and Sarikaya, A. and Gleicher, M. , year =. Navigating reductionism and holism in evaluation , doi =. Proceedings of the

  14. [24]

    Nature Biomedical Engineering , author =

    Auditing the inference processes of medical-image classifiers by leveraging generative. Nature Biomedical Engineering , author =. 2025 , pages =. doi:10.1038/s41551-023-01160-9 , number =

    2025 doi

  15. [25]

    and Li, S

    Dong, T. and Li, S. and Chen, G. and Xue, M. and Zhu, H. , year =. doi:10.14722/ndss.2023.241012 , booktitle =

    2023

  16. [26]

    Information & Communications Technology Law , author =

    Upstream, downstream, and in between: navigating the. Information & Communications Technology Law , author =. 2026 , pages =. doi:10.1080/13600834.2026.2624923 , language =

    2026

  17. [27]

    2023 , note =

    Advancing. 2023 , note =. doi:10.48550/arXiv.2312.00044 , author =

    2023 doi arXiv

  18. [28]

    Nature Machine Intelligence , author =

    Governing. Nature Machine Intelligence , author =. 2021 , pages =

    2021

  19. [29]

    and Kalff, Y

    Fernsel, L. and Kalff, Y. and Simbeck, K. , year =. Where. doi:10.5220/0012689800003693 , booktitle =

    doi

  20. [30]

    Ferrell, U. D. and Anderegg, A. H.A. , year =. doi:10.1109/DASC50938.2020.9256608 , booktitle =

    2020

  21. [31]

    and Oostdyk, R

    Ferrell, B. and Oostdyk, R. , year =. Modeling and performance considerations for automated fault isolation in complex systems , booktitle =

  22. [32]

    2024 , note =

    Auditing the. 2024 , note =. doi:10.48550/arXiv.2404.03086 , author =

    2024 doi arXiv

  23. [33]

    and Martín Zamorano, M

    Galdon Clavell, G. and Martín Zamorano, M. and Castillo, C. and Smith, O. and Matic, A. , year =. Auditing. doi:10.1145/3375627.3375852 , booktitle =

  24. [34]

    and Wan, S

    Gan, W. and Wan, S. and Philip, S. Y. , year =. Model-as-a-service (. 2023

    2023

  25. [35]

    and Kitroeff, N

    Gelles, D. and Kitroeff, N. , year =

  26. [37]

    2023 , journal =

    Co-audit:. 2023 , journal =. doi:10.48550/arXiv.2310.01297 , author =

    2023 doi arXiv

  27. [39]

    2022 , note =

    System. 2022 , note =. doi:10.48550/arXiv.2203.04754 , author =

    2022 doi arXiv

  28. [40]

    Hauer, M. P. and Adler, R. and Zweig, K. , year =. Assuring. doi:10.1109/ICSTW52544.2021.00029 , booktitle =

    2021

  29. [41]

    and Martens, C

    Helmer, L. and Martens, C. and Wegener, D. and Akila, M. and Becker, D. and Abbas, S. , year =. Towards. doi:10.1145/3643691.3648584 , booktitle =

  30. [42]

    2025 , pages =

    Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society , author =. 2025 , pages =. doi:10.1609/aies.v8i2.36628 , number =

    2025 doi

  31. [43]

    Therapeutic Innovation & Regulatory Science , author =

    Current application of the medical device single audit program (. Therapeutic Innovation & Regulatory Science , author =. 2024 , pages =

    2024

  32. [44]

    International

    ISO , url =. International. Medical devices — Quality Management Systems — Requirements for Regulatory Purposes (ISO Standard , year =

  33. [45]

    International

    ISO , url =. International. Software engineering — Software Product Quality Requirements and Evaluation (SQuaRE) — Data quality model (ISO Standard , year =

  34. [46]

    and Sapiezynski, P

    Kaplan, L. and Sapiezynski, P. , year =. Comprehensively. doi:10.1145/3589335.3651260 , booktitle =

  35. [47]

    Journal of Intelligence , author =

    Systematizing. Journal of Intelligence , author =. doi:10.3390/jintelligence9030046 , number =

    doi

  36. [48]

    2023 , annote =

    Fostering transparent medical image. 2023 , annote =. doi:10.1101/2023.06.07.23291119 , author =

    2023 doi

  37. [49]

    Proceedings of the ACM on Human-Computer Interaction , author =

    Give. Proceedings of the ACM on Human-Computer Interaction , author =. 2022 , pages =. doi:10.1145/3555149 , number =

    2022 doi

  38. [50]

    The Review of Socionetwork Strategies , author =

    Towards a. The Review of Socionetwork Strategies , author =. 2023 , pages =. doi:10.1007/s12626-023-00148-w , number =

    2023 doi

  39. [51]

    IEEE Internet Computing , author =

    Common. IEEE Internet Computing , author =. 2024 , pages =. doi:10.1109/MIC.2024.3377170 , number =

    2024

  40. [52]

    Royal Society Open Science , author =

    Towards algorithm auditing: managing legal, ethical and technological risks of. Royal Society Open Science , author =. 2024 , pages =. doi:10.1098/rsos.230859 , abstract =

    2024 doi

  41. [54]

    Proceedings of the ACM on Human-Computer Interaction , author =

    End-. Proceedings of the ACM on Human-Computer Interaction , author =. 2022 , pages =. doi:10.1145/3555625 , number =

    2022 doi

  42. [55]

    Proceedings of the ACM on Human-Computer Interaction , author =

    Sociotechnical. Proceedings of the ACM on Human-Computer Interaction , author =. 2023 , pages =. doi:10.1145/3610209 , number =

    2023 doi

  43. [56]

    Enterprise Risk Management , author =

    Examining risk management failures:. Enterprise Risk Management , author =. 2023 , pages =

    2023

  44. [57]

    Reliability Engineering & System Safety , author =

    Metrics for evaluating the performance of complex engineering system health monitoring models , volume =. Reliability Engineering & System Safety , author =. 2022 , pages =

    2022

  45. [58]

    Columbia Journal of Transnational Law , author =

    More. Columbia Journal of Transnational Law , author =

  46. [59]

    The Lancet Digital Health , author =

    The medical algorithmic audit , volume =. The Lancet Digital Health , author =. 2022 , pages =. doi:10.1016/S2589-7500(22)00003-6 , number =

    2022 doi

  47. [60]

    IEEE Transactions on Dependable and Secure Computing , author =

    The “. IEEE Transactions on Dependable and Secure Computing , author =. 2024 , pages =. doi:10.1109/TDSC.2024.3367737 , number =

    2024

  48. [61]

    Systems Engineering , author =

    Systems integration:. Systems Engineering , author =. 2014 , pages =

    2014

  49. [62]

    Systems Engineering , author =

    System of systems integration:. Systems Engineering , author =. 2014 , pages =

    2014

  50. [63]

    Academic Radiology , author =

    The. Academic Radiology , author =. 2020 , pages =. doi:10.1016/j.acra.2019.09.009 , number =

    2020 doi

  51. [64]

    and Ahmed, R

    Malunga, M. and Ahmed, R. and Raynard, A. and Stoyanova, M. and Chisala, K. , file =. Key lessons from the. Loss Prevention Bulletin , volume =

  52. [65]

    , year =

    Meadows, D. , year =. Thinking in systems:

  53. [66]

    and Seabridge, A

    Moir, I. and Seabridge, A. , year =. Aircraft

  54. [67]

    A Blueprint for Auditing Generative

    M. A Blueprint for Auditing Generative. 2025 , booktitle =

    2025

  55. [68]

    2023 , pages =

    AI and Ethics , author =. 2023 , pages =. doi:10.1007/s43681-022-00171-7 , number =

    2023 doi

  56. [69]

    and Sinatra, R

    Moreau, T. and Sinatra, R. and Sekara, V. , year =. Failing. doi:10.1145/3630106.3658906 , booktitle =

  57. [70]

    and Markovic, M

    Naja, I. and Markovic, M. and Edwards, P. and Cottrill, C. , editor =. A. 2021 , pages =. doi:10.1007/978-3-030-77385-4_10 , booktitle =

    2021 doi

  58. [71]

    and Markovic, M

    Naja, I. and Markovic, M. and Edwards, P. and Pang, W. and Cottrill, C. and Williams, R. , year =. Using. doi:10.1109/ACCESS.2022.3188967 , journal =

    2022

  59. [72]

    , year =

    Nartey, J. , year =. The

  60. [73]

    and Kitroeff, N

    Nicas, J. and Kitroeff, N. and Gelles, D. and Glanz, J. , year =. Boeing built deadly assumptions into 737

  61. [74]

    The Lancet Digital Health , author =

    Validation and algorithmic audit of a deep learning system for the detection of proximal femoral fractures in patients in the emergency department:. The Lancet Digital Health , author =. 2022 , pages =. doi:10.1016/S2589-7500(22)00004-8 , number =

    2022 doi

  62. [75]

    and Fehr, J

    Oala, L. and Fehr, J. and Gilli, L. and Balachandran, P. and Leite, A. W. and Calderon-Ramirez, S. and Li, D. X. and Nobis, G. and Alvarado, E. A.M. and Jaramillo-Gutierrez, G. and Matek, C. and Shroff, A. and Kherif, F. and Sanguinetti, B. and Wiegand, T. , year =. From

  63. [76]

    2022 , note =

    Auditing. 2022 , note =. doi:10.48550/arXiv.2211.08742 , author =

    2022 doi arXiv

  64. [77]

    and West, J

    Pettet, G. and West, J. and Robert, D. and Khetani, A. and Kumar, S. and Golla, S. and Lavis, R. , year =. A retrospective audit of an. BJR. doi:10.1093/bjro/tzae033 , number =

    doi

  65. [78]

    and Fleiß, J

    Polzer, A. and Fleiß, J. and Ebner, T. and Kainz, P. and Koeth, C. and Thalmann, S. , year =. Validation of. doi:10.24251/HICSS.2022.186 , booktitle =

    2022 doi

  66. [79]

    Sustainability , author =

    Enhancing aerospace industry efficiency and sustainability:. Sustainability , author =. 2023 , pages =

    2023

  67. [80]

    Journal of Aerospace Information Systems , author =

    Software error incident categorizations in aerospace , volume =. Journal of Aerospace Information Systems , author =. 2024 , pages =

    2024

  68. [82]

    Rajabalinejad, Mohammad , month = may, year =. Safe. 2019 14th. doi:10.1109/SYSOSE.2019.8753867 , abstract =

    2019

  69. [83]

    and Dobbe, R

    Rakova, B. and Dobbe, R. , year =. Algorithms as. doi:10.1145/3593013.3594014 , booktitle =

  70. [84]

    Official Journal of the European Union , author =

    Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence and amending Regulations (EC) No 300/2008, (EU) No 167/2013, (EU) No 168/2013, (EU) 2018/858, (EU) 2018/1139 and (EU) 2019/2144 and D...

    2024

  71. [85]

    Graves, S

    Rep. Graves, S. , year =. H

  72. [86]

    Data Mining and Knowledge Discovery , author =

    An external stability audit framework to test the validity of personality prediction in. Data Mining and Knowledge Discovery , author =. 2022 , pages =. doi:10.1007/s10618-022-00861-0 , number =

    2022 doi

  73. [87]

    Rico, C. V. , year =. General-. The

  74. [88]

    Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society , author =

    Measuring. Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society , author =. 2025 , pages =

    2025

  75. [89]

    The Guardian , author =

    Boeing puts cost of 737. The Guardian , author =

  76. [90]

    , year =

    Schulz, M. , year =. Logic of consequences and logic of appropriateness , booktitle =

  77. [91]

    Sghaier, M. O. and Hadzagic, M. and Shahbazian, E. , year =. Verifiable. doi:10.1109/CogSIMA61085.2024.10553806 , booktitle =

    2024

  78. [92]

    Patterns , author =

    A. Patterns , author =. doi:10.1016/j.patter.2021.100425 , number =

    2021

  79. [93]

    and Harer, J

    Smogavc Cestar, A. and Harer, J. , year =. Mastering. Medical

  80. [94]

    The Lancet Digital Health , author =

    Diagnosis and prognosis of abnormal cardiac scintigraphy uptake suggestive of cardiac amyloidosis using. The Lancet Digital Health , author =. 2024 , pages =. doi:10.1016/S2589-7500(23)00265-0 , number =

    2024 doi

  81. [95]

    Transfers , author =

    Learning by. Transfers , author =. 2024 , pages =

    2024

  82. [96]

    Future Internet , author =

    Modeling and. Future Internet , author =. doi:10.3390/fi14100284 , number =

    doi

  83. [97]

    Summers, B. L. , year =. Effective methods for software and systems integration , publisher =

  84. [98]

    Risk, regulation, and governance: evaluating

    Szadeczky, Tamas and Bederna, Zsolt , year =. Risk, regulation, and governance: evaluating. Security Journal , pages =

  85. [99]

    2018 , pages =

    Annals of internal medicine , author =. 2018 , pages =

    2018

  86. [100]

    and Luccioni, S

    Varoquaux, G. and Luccioni, S. and Whittaker, M. , year =. Hype,. Proceedings of the 2025

    2025

  87. [101]

    and Levy, K

    Vecchione, B. and Levy, K. and Barocas, S. , year =. Algorithmic auditing and social justice:

  88. [102]

    Journal of Air Law and Commerce , author =

    Flight. Journal of Air Law and Commerce , author =

  89. [103]

    Proceedings of the ACM on Human-Computer Interaction , author =

    Lower. Proceedings of the ACM on Human-Computer Interaction , author =. 2024 , pages =. doi:10.1145/3687046 , number =

    2024 doi

  90. [104]

    European Journal of Risk Regulation , author =

    On regulating downstream. European Journal of Risk Regulation , author =. 2025 , pages =

    2025

  91. [105]

    and Ghosh, A

    Wilson, C. and Ghosh, A. and Jiang, S. and Mislove, A. and Baker, L. and Szary, J. and Trindel, K. and Polli, F. , year =. Building and. doi:10.1145/3442188.3445928 , booktitle =

  92. [106]

    , year =

    Winter, J. , year =. System. 2020

    2020

  93. [107]

    Proceedings of the IEEE , author =

    Massively digitized power grid:. Proceedings of the IEEE , author =. 2022 , pages =

    2022

  94. [108]

    and Zhang, T

    Yang, Y.-T. and Zhang, T. and Zhu, Q. , editor =. A. 2023 , pages =. doi:10.1007/978-3-031-50670-3_18 , booktitle =

    2023 doi

  95. [109]

    GRUR International , author =

    General-purpose. GRUR International , author =. 2024 , pages =

    2024

  96. [110]

    IEEE Transactions on Engineering Management , author =

    The progress function in the aerospace industry—. IEEE Transactions on Engineering Management , author =. 1963 , pages =

    1963

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.