Pith. sign in

REVIEW 4 major objections 6 minor 19 references

The Role of XAI in Transforming Aeronautics and Aerospace Systems

T0 review · 4 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read This paper surveys XAI and reports that post-hoc techniques have been proven to make AI models interpretable across aeronautics and aerospace applications.

desk verdict A readable but careless XAI survey: the standard definitions are fine, but the application section's misattributed citations undercut the paper's only contribution. read the letter →

arxiv 2412.17440 v1 pith:GGVH3FPB submitted 2024-12-23 cs.AI

Francisco Javier Cantero Zorita , Mikel Galafate , Javier M. Moguerza , Isaac Martín de Diego , M. Teresa Gonzalez , Gema Gutierrez Peña This is my paper · ORCID
classification cs.AI
keywords AeronauticsAerospaceExplainableAIMachineLearningSafety-CriticalApplicationsPost-hocexplanationLIMESHAP
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 paper surveys explainable artificial intelligence (XAI) as a bridge between humans and AI systems in aeronautics and aerospace. It argues that post-hoc techniques such as LIME and SHAP, along with transparent models, are already being used so professionals can understand and trust AI predictions in air traffic management, drone operations, predictive maintenance, spacecraft telemetry, and satellite imagery. The review organizes the field by defining XAI's goals, the properties that make models interpretable, and a taxonomy of transparent models versus post-hoc explanation methods, concluding that across these sectors the use of post-hoc techniques has been shown to ensure model interpretability.

What carries the argument

The paper's central organizing device is a two-part taxonomy of XAI methods: transparent models, which are inherently interpretable because of their algorithmic simplicity, and post-hoc techniques, which are applied after training to explain black-box models. Post-hoc methods are split into model internals, model surrogates (LIME, SHAP, Anchors), feature summaries (feature importance, partial dependence plots), and example-based methods (counterfactuals, influential observations, prototypes and criticisms). This taxonomy frames every application described in the survey and supplies the vocabulary the paper uses to claim that interpretability can be achieved.

What would settle it

Inspect reference [18] and the other application citations: if the cited sources do not actually contain the described case studies, such as post-disaster drone damage assessment or poverty-index prediction, the survey's characterization of the state of the art collapses. A reader could also verify whether each cited paper names the specific post-hoc technique the survey attributes to it, such as LIME in spacecraft telemetry or SHAP in predictive maintenance.

Watch

Extended reading notes

Core claim

The paper asserts that XAI has moved from a research concept to an applied tool in safety-critical aeronautics and aerospace settings. In particular, it contends that post-hoc explanation techniques, including local surrogate models like LIME, feature-attribution methods like SHAP, and example-based methods such as counterfactuals, can make the decisions of otherwise opaque neural networks intelligible to domain professionals. It catalogs applications in takeoff and landing time prediction, fuzzy-rule route adaptation for UAVs, post-disaster damage assessment, predictive maintenance, spacecraft telemetry anomaly detection, and satellite-image-based poverty mapping. The paper's conclusion is that post-hoc techniques have been proven to ensure interpretability in each of these areas.

Load-bearing premise

The survey's map of XAI applications is only as reliable as the primary sources behind its citations, and the text itself uses reference [18] for two unrelated applications that do not match the reference's title, so this reliability cannot be taken for granted.

Editorial extensions

If this is right

  • The paper claims that the performance–interpretability trade-off is a central constraint, so XAI adoption in safety-critical settings must balance accuracy against explainability.
  • It presents the taxonomy of transparent versus post-hoc methods as a complete map of available explanation approaches.
  • It asserts that post-hoc techniques such as LIME, SHAP, feature importance, and counterfactuals have been proven to ensure interpretability in ATM, UAVs, predictive maintenance, telemetry anomaly detection, and satellite imagery.
  • It states that explanations must be tailored to the user profile, which implies that practitioners and regulators may need different explanations from the same system.

Reading between the lines

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

  • The survey's own citation trail is the fragile part: reference [18] is cited for both post-disaster drone damage assessment and poverty-index prediction in Uganda, while its listed title is a brief review of XAI in drones. If the primary sources are misattributed, the survey's map of the field would mischaracterize the evidence base even though the broad claim may be plausible.
  • An implicit consequence of the paper's framing is that XAI's value in these sectors depends on whether explanations actually change human decisions; the survey documents that explanations are generated but does not measure their operational impact.
  • A testable extension would be a quantitative comparison of explanation fidelity across the cited application domains, since the paper assumes that LIME, SHAP, and other surrogates faithfully represent the black-box models they explain.
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 / 6 minor

Summary. This manuscript is a narrative survey of explainable artificial intelligence (XAI) aimed at aeronautics and aerospace applications. It defines XAI and its objectives, reviews the properties associated with interpretable and transparent models, describes transparent versus black-box models and post-hoc explanation techniques, and surveys six application areas: air traffic management, UAV route adaptation and simulation, post-disaster damage assessment, predictive maintenance, spacecraft telemetry anomaly detection, and satellite image analysis. The paper's central claim is that post-hoc techniques such as LIME and SHAP have been applied to make AI models understandable to professionals in these safety-critical sectors, and the conclusion asserts that this has been 'proven' by the surveyed applications.

Significance. If the survey's characterizations are accurate, it would provide a concise entry point for practitioners seeking to understand XAI methods in aeronautics and aerospace. The taxonomy in Sections 2–4 is broadly faithful to the authoritative sources it cites, and the selection of application areas is relevant. However, the paper contains no original analysis, systematic methodology, or independent evaluation, so its value rests entirely on the correctness of its citation-to-claim mapping. The citation problems in Section 5 are therefore load-bearing. The paper does not ship code, data, or proofs; its contribution is a synthesis, and that synthesis must be reliable.

major comments (4)
  1. [Section 5, paragraphs 3–4] Reference [18] (Banimelhem and Al-khateeb, 'Explainable artificial intelligence in drones: A brief review') is cited for two unrelated empirical applications: post-natural-disaster damage assessment using actual and predicted values, and prediction of poverty indices in Uganda using decision trees and deep object detection. A brief review cannot be the primary source for both empirical studies. The manuscript's own reference list contains the likely intended sources—[14] (Cheng et al., uncertainty-aware CNN for disaster damage assessment) and [19] (Ayush et al., efficient poverty mapping from remote sensing)—but [14] is instead cited in Section 2 for the definition of 'comprehensibility,' where it also does not belong. As written, two of the six application claims in Section 5 cannot be verified from the cited literature, and Section 6's statement that 'it has been proven' that post-hoc techniques ensure interpretability is not supported. The authors must re-verify every application claim against its primary source and correct the in-text citations accordingly.
  2. [Section 5, paragraph 3] The sentence about drone simulations in three different modes validated with MSE, RMSE, and MAE cites [17], Mualla et al., 'The quest of parsimonious XAI: A human-agent architecture for explanation formulation.' The title and venue do not indicate that this paper reports drone simulations with the listed error metrics; please confirm that [17] is the correct source, replace it with the study that actually performed these simulations, or remove the quantitative detail. In addition, the following sentence about Grad-CAM attribution maps for daytime satellite images and nighttime light data in Sub-Saharan Africa cites [19], Ayush et al., whose abstract does not mention Grad-CAM; if a different paper is meant, it is missing from the reference list.
  3. [Section 6 (Conclusions)] The sentence 'it has been proven how the use of post-hoc techniques can ensure the interpretability of models' overstates what a survey can establish. The cited works are applications or proposals; they do not constitute a proof, and the citation problems in Section 5 further weaken the evidential basis. The authors should replace 'proven' with language such as 'reported' or 'demonstrated in the cited cases,' and they should qualify the claim as reflecting the selected literature rather than a general guarantee.
  4. [Section 5 (Applications)] The review does not state a literature search strategy, inclusion/exclusion criteria, or time window, and it covers six application areas with one or two citations each. If the intent is a comprehensive survey of XAI in aeronautics and aerospace, the methodology and coverage criteria should be described; if the intent is to present illustrative examples, the authors should say so explicitly and soften the language in the conclusions. Without this framing, the reader cannot judge whether the selected applications are representative or selective.
minor comments (6)
  1. [Abstract] The phrase 'this paper provides a review of the concept of XAI is carried out' is ungrammatical; consider 'this paper reviews the concept of XAI, defining the term and its objectives.'
  2. [Section 1] 'Aerial Traffic Management' should be 'Air Traffic Management', and the comma after 'DARPA' should be removed.
  3. [Section 4] The sentence beginning 'A distinction is drawn between transparent models...' is duplicated immediately after Figure 1; one copy should be removed.
  4. [Section 6] The statement that 'the properties that should be evaluated in AI systems and models to be considered black-box models have been defined' is inconsistent with Section 3, where the properties are introduced as criteria for comprehensibility and explainability, not for black-box status.
  5. [Section 3] Reference [15] is a predictive-maintenance survey; it is not an obvious source for the performance-interpretability trade-off, which is already discussed in [8] and [9]. Please cite the original source for the trade-off concept.
  6. [References] The reference list should be checked for consistency: [2] should be formatted as 'Nature News', and the citation numbering should be cross-checked against the in-text citations after the Section 5 corrections are made.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular reasoning: the paper is a survey that restates external definitions and cited applications, with no derivation chain that reduces to its own inputs.

full rationale

This paper is a literature review, not a derivation or prediction pipeline. Its sections define XAI terms by citing external sources (Gunning et al. [3], Arrieta et al. [8], Molnar [9], Lipton [10], etc.), enumerate model properties and post-hoc technique categories from those same external taxonomies, and then survey applications in aeronautics and aerospace by citing external primary sources. There is no fitted parameter, no mathematical derivation, no 'prediction' computed from the paper's own assumptions, and no uniqueness theorem or load-bearing self-citation invoked to force a conclusion. The paper's broad claims, including Section 6's statement that 'it has been proven how the use of post-hoc techniques can ensure the interpretability of models,' are empirical claims about the cited literature, not consequences of the paper's own definitions. The reader-flagged concern that reference [18] may be misattributed for two application claims is a citation-accuracy or evidence-quality issue: if the cited sources do not support those sentences, the survey's reliability weakens, but this does not make the argument circular because the survey's content is not the source of the evidence it cites. No circular step can be exhibited by quoting a reduction of Eq. X to Eq. Y or a fitted input renamed as a prediction, so the honest finding is no significant circularity, score 0.

Assumptions & free parameters 0 free parameters · 2 assumptions · 0 invented entities

No free parameters or invented entities are present. The paper rests entirely on external literature, so the ledger consists of fidelity-to-source assumptions rather than fitted quantities or postulated mechanisms.

assumptions (2)
  • domain assumption The definitions and taxonomy in Sections 2 to 4 faithfully represent the cited XAI literature (Arrieta et al. [8], Molnar [9], Lipton [10]).
    The survey's conceptual content is not independently derived; it is taken from these sources, and any misreading propagates through the paper.
  • domain assumption The application examples in Section 5 accurately reflect what the cited papers actually did.
    The claims about ATM, drones, predictive maintenance, anomaly detection, and satellite imaging rest on the fidelity of the citations; the text appears to cite [18] for two different applications, so this assumption is not fully met.

how reviews work

0 comments
Cite this review

Pith. "Pith review of The Role of XAI in Transforming Aeronautics and Aerospace Systems." pith.science (2026). https://pith.science/paper/GGVH3FPB

@misc{pith2026241217440,
  author       = {Pith},
  title        = {Pith review of: The Role of XAI in Transforming Aeronautics and Aerospace Systems},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GGVH3FPB}},
  note         = {Machine review of arXiv:2412.17440}
}
read the original abstract

Recent advancements in Artificial Intelligence (AI) have transformed decision-making in aeronautics and aerospace. These advancements in AI have brought with them the need to understand the reasons behind the predictions generated by AI systems and models, particularly by professionals in these sectors. In this context, the emergence of eXplainable Artificial Intelligence (XAI) has helped bridge the gap between professionals in the aeronautical and aerospace sectors and the AI systems and models they work with. For this reason, this paper provides a review of the concept of XAI is carried out defining the term and the objectives it aims to achieve. Additionally, the paper discusses the types of models defined within it and the properties these models must fulfill to be considered transparent, as well as the post-hoc techniques used to understand AI systems and models after their training. Finally, various application areas within the aeronautical and aerospace sectors will be presented, highlighting how XAI is used in these fields to help professionals understand the functioning of AI systems and models.

Figures

Figures reproduced from arXiv: 2412.17440 by the authors.

Figure 1
Figure 1. Taxonomy of XAI methods. the simplicity of the algorithms, and the post-hoc techniques applied after the model’s training to better understand its workings and predictions. A distinction is drawn between transparent models, which are inherently interpretable due to their simplicity, and post-hoc techniques, applied after model training to elucidate the internal workings and predictions of black-box models. Furthermo… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

19 extracted references · 14 canonical work pages

  1. [18]

    Explainable artificial intelligence in drones: A brief review,

    O. Banimelhem and B. Al-khateeb, “Explainable artificial intelligence in drones: A brief review,” in2023 14th International Conference on Information and Communication Systems (ICICS). IEEE, 2023, pp. 1–5

    work page 2023

  2. [14]

    Uncertainty-aware convolutional neural network for explainable artificial intelligence-assisted disaster damage assessment,

    C.-S. Cheng, A. H. Behzadan, and A. Noshadravan, “Uncertainty-aware convolutional neural network for explainable artificial intelligence-assisted disaster damage assessment,”Structural Control and Health Monitoring, vol. 29, no. 10, p. e3019, 2022

    work page 2022

  3. [19]

    Efficient poverty mapping from high resolution remote sensing images,

    K. Ayush, B. Uzkent, K. Tanmay, M. Burke, D. Lobell, and S. Ermon, “Efficient poverty mapping from high resolution remote sensing images,” inProceedings of the AAAI Conference on Artificial Intelligence, vol. 35, no. 1, 2021, pp. 12–20. 7

    work page 2021

  4. [17]

    The quest of parsimonious xai: A human-agent architecture for explanation formulation,

    Y. Mualla, I. Tchappi, T. Kampik, A. Najjar, D. Calvaresi, A. Abbas-Turki, S. Galland, and C. Nicolle, “The quest of parsimonious xai: A human-agent architecture for explanation formulation,”Artificial intelligence, vol. 302, p. 103573, 2022

    work page 2022

  5. [1]

    Eu regulations on algorithmic decision-making and a “right to explanation

    B. Goodman and S. Flaxman, “Eu regulations on algorithmic decision-making and a “right to explanation”,” vol. 38, 2016

    work page 2016

  6. [2]

    Can we open the black box of ai?

    D. Castelvecchi, “Can we open the black box of ai?”Nature News, vol. 538, no. 7623, p. 20, 2016

    2016

  7. [3]

    Xai—explainable artificial intelligence,

    D. Gunning, M. Stefik, J. Choi, T. Miller, S. Stumpf, and G.-Z. Yang, “Xai—explainable artificial intelligence,” Science robotics, vol. 4, no. 37, p. eaay7120, 2019

    work page 2019

  8. [4]

    A survey on artificial intelligence (ai) and explainable ai in air traffic management: Current trends and development with future research trajectory,

    A. Degas, M. R. Islam, C. Hurter, S. Barua, H. Rahman, M. Poudel, D. Ruscio, M. U. Ahmed, S. Begum, M. A. Rahman et al., “A survey on artificial intelligence (ai) and explainable ai in air traffic management: Current trends and development with future research trajectory,”Applied Sciences, vol. 12, no. 3, p. 1295, 2022

    work page 2022

Show all 19 references
  1. [5]

    (explainable) artificial intelligence in aerospace safety-critical systems,

    S. Sutthithatip, S. Perinpanayagam, and S. Aslam, “(explainable) artificial intelligence in aerospace safety-critical systems,” in2022 IEEE Aerospace Conference (AERO). IEEE, 2022, pp. 1–12

    2022

  2. [6]

    Opportunities for explainable artificial intelligence in aerospace predictive maintenance,

    B. Shukla, I.-S. Fan, and I. Jennions, “Opportunities for explainable artificial intelligence in aerospace predictive maintenance,” inPHM Society European Conference, vol. 5, no. 1, 2020, pp. 11–11

    2020

  3. [7]

    Explainable anomaly detection in spacecraft telemetry,

    S. Cuéllar, M. Santos, F. Alonso, E. Fabregas, and G. Farias, “Explainable anomaly detection in spacecraft telemetry,”Engineering Applications of Artificial Intelligence, vol. 133, p. 108083, 2024

    2024

  4. [8]

    Explainable artificial intelligence (xai): Concepts, taxonomies, opportunities and challenges toward responsible ai,

    A.B.Arrieta,N.Díaz-Rodríguez,J.DelSer,A.Bennetot,S.Tabik,A.Barbado,S.García,S.Gil-López,D.Molina, R. Benjaminset al., “Explainable artificial intelligence (xai): Concepts, taxonomies, opportunities and challenges toward responsible ai,”Information fusion, vol. 58, pp. 82–115, 2020

    2020

  5. [9]

    Molnar,Interpretable machine learning

    C. Molnar,Interpretable machine learning. Lulu. com, 2020

    2020

  6. [10]

    The mythos of model interpretability: In machine learning, the concept of interpretability is both important and slippery

    Z. C. Lipton, “The mythos of model interpretability: In machine learning, the concept of interpretability is both important and slippery.”Queue, vol. 16, no. 3, pp. 31–57, 2018

    2018

  7. [11]

    Explanation in artificial intelligence: Insights from the social sciences,

    T. Miller, “Explanation in artificial intelligence: Insights from the social sciences,”Artificial intelligence, vol. 267, pp. 1–38, 2019

    2019

  8. [12]

    Interpretable to whom? a role-based model for analyzing interpretable machine learning systems,

    R. Tomsett, D. Braines, D. Harborne, A. Preece, and S. Chakraborty, “Interpretable to whom? a role-based model for analyzing interpretable machine learning systems,”arXiv preprint arXiv:1806.07552, 2018

    2018 arXiv

  9. [13]

    Methods for interpreting and understanding deep neural networks,

    G. Montavon, W. Samek, and K.-R. Müller, “Methods for interpreting and understanding deep neural networks,” Digital signal processing, vol. 73, pp. 1–15, 2018

    2018

  10. [15]

    A survey of predictive maintenance: Systems, purposes and approaches,

    Y. Ran, X. Zhou, P. Lin, Y. Wen, and R. Deng, “A survey of predictive maintenance: Systems, purposes and approaches,”arXiv preprint arXiv:1912.07383, pp. 1–36, 2019

    1912 arXiv

  11. [16]

    Evolving rule-based explainable artificial intelligence for unmanned aerial vehicles,

    B. M. Keneni, D. Kaur, A. Al Bataineh, V. K. Devabhaktuni, A. Y. Javaid, J. D. Zaientz, and R. P. Marinier, “Evolving rule-based explainable artificial intelligence for unmanned aerial vehicles,”IEEE Access, vol. 7, pp. 17001–17016, 2019

    2019

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.