Pith. sign in

REVIEW 3 major objections 5 minor 55 references

Towards explainable decision support using hybrid neural models for logistic terminal automation

T0 review · 3 major / 5 minor · reviewed 2026-08-04 · deepseek-v4-flash

Pith's one-line read The paper claims a three-stage INSD pipeline can give neural decision-support models for intermodal terminals human-meaningful concepts, causal grounding, and interpretable dynamic equations, replacing post-hoc explanations with by-design t

desk verdict A coherent roadmap paper that overclaims in the abstract: the proposed pipeline is plausible, but the load-bearing causal stage is entirely unvalidated. read the letter →

arxiv 2509.07577 v2 pith:55UIU3YM submitted 2025-09-09 cs.AI

classification cs.AI
keywords explainableAIsystemdynamicscausalmachinelearningintermodallogisticscognitivedigitaltwinstrustworthyautonomoussystemsinterpretabilitybydesignneuro-symbolic
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

The paper argues that the opacity of deep learning models can be overcome by construction rather than by post-hoc explanation. It proposes the Interpretable Neural System Dynamics (INSD) pipeline, in which raw terminal sensor data is first mapped to human-understandable concepts such as yard congestion and crane idleness, then to a causal graph, and finally to interpretable dynamic equations. If the pipeline works as claimed, logistics operators would get predictions that can be traced to meaningful variables, tested with 'what-if' interventions, and audited by regulators—properties that current black-box models lack. The paper is a research roadmap for an EU-funded intermodal-terminal automation project rather than a report of a working system, so its contribution is the architecture and the open challenges it identifies.

What carries the argument

The Interpretable Neural System Dynamics (INSD) pipeline—a three-stage hybrid architecture (concept learner → structural causal learner → neural equation learner) that converts raw sensor data into a transparent, causally grounded dynamic model. It carries the argument by replacing end-to-end black-box prediction with a chain of semantically meaningful representations, where each stage is auditable and feeds the next stage's constraints.

What would settle it

Apply the pipeline to a terminal dataset or simulator with a known ground-truth causal graph and recorded interventions; if the recovered graph omits the operative link (for example, rerouting trucks does not actually reduce train delays) or its counterfactual predictions disagree with observed post-intervention outcomes, the causal-validity claim is falsified.

Watch

Extended reading notes

Core claim

The paper's central claim is that semantic interpretability, mechanistic interpretability, and causal reliability can be designed into neural network models of dynamical systems instead of bolted on afterwards. The INSD pipeline does this in three stages: concept learning extracts semantically meaningful variables from raw operational data; causal learning produces a structural causal graph among those concepts; and equation learning fits interpretable dynamic equations constrained by that graph. The resulting model retains the transparency and causal grounding of traditional System Dynamics while gaining the scalability and flexibility of deep learning, enabling counterfactual simulation an

Load-bearing premise

The pipeline assumes causal discovery can recover the true cause-effect structure from the terminal's noisy, partially observed, non-stationary streaming data; if that assumption fails, the claimed causal reliability and counterfactual support collapse.

Editorial extensions

If this is right

  • Decision support for terminal operators can be traced to named operational concepts such as crane idleness or yard congestion, rather than to opaque features.
  • What-if planning becomes possible: operators can simulate interventions such as adding a crane or rerouting trucks and read off predicted effects from the structural equations.
  • Models become auditable along three dimensions—semantic alignment, mechanistic structure, and causal validity—supporting regulatory compliance and operator trust.
  • The same three-stage architecture can be generalized to other cyber-physical systems beyond logistics, such as transportation and industrial IoT settings.
  • The framework directly addresses the integration gap noted in XAI literature, where semantic, causal, and mechanistic interpretability are usually pursued in isolation.

Reading between the lines

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

  • If the pipeline matures, the equation-learning stage may let the causal graph be validated by prediction quality, allowing learned equations to be benchmarked against expert-built System Dynamics models on both forecast accuracy and intervention fidelity.
  • The recovered causal graph likely depends on how the concept vocabulary is fixed before causal discovery; an iterative human-in-the-loop concept refinement step may be necessary to keep the graph stable across terminal layouts and operational shifts.
  • A testable extension is to stress-test the causal stage with synthetic terminal simulators whose ground-truth causal graph is known, measuring how well the recovered graph survives sensor dropout, missing data, and non-stationarity.
  • The same hybrid architecture could migrate to other high-stakes operational domains—hospital logistics, energy grids, or port operations—where auditable, causally reliable decision support is equally critical.
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

3 major / 5 minor

Summary. The paper proposes an Interpretable Neural System Dynamics (INSD) pipeline for decision support in intermodal logistic terminals. The pipeline has three stages: Concept Learning (mapping raw sensor data to human-understandable variables), Causal Learning (discovering causal dependencies among concepts), and Equation Learning (deriving interpretable structural dynamic equations). The authors argue that this integration provides interpretability-by-design, causal reliability, and auditability, and they position it as a foundation for Cognitive Digital Twins in logistics. The paper includes a literature review, a schematic pipeline description, an illustrative terminal scenario, and a research agenda of open challenges. It explicitly frames the contribution as a roadmap rather than a fully implemented system.

Significance. If the INSD framework were realized as described, it would address a real gap: most existing XAI methods in logistics are post-hoc, statistically correlational, and not designed for interventional or counterfactual reasoning. The paper is valuable in synthesizing concept-based interpretability, causal machine learning, and mechanistic equation learning into a coherent research agenda, and it identifies concrete technical challenges (data gaps, cyclic feedback, non-stationarity, evaluation) that are often neglected in position papers. The strengths are the clear taxonomy of opacity, the explicit link to operational decision tasks in intermodal terminals, and the acknowledgment of open problems. However, the central claim that the pipeline 'enables' causally reliable, interpretable models is not demonstrated: there is no implementation, dataset experiment, or formal derivation. The contribution is therefore best read as a roadmap and research agenda, not as a validated method. As such, its significance is conditional on future work, and the paper must be revised to align its claims with the evidence it provides.

major comments (3)
  1. [Abstract and §3] The central claim that the INSD pipeline 'enables the construction of neural network models that operate on semantically meaningful and actionable variables, while retaining the causal grounding and transparency typical of traditional SD models' is not supported by any implementation, data analysis, or formal derivation. The paper itself characterizes the contribution as a 'roadmap' (Conclusion), and §3 presents a high-level diagram plus an illustrative scenario (§3.1), not a method specification. As stated, 'enables' is an assertion about a possible future system rather than a demonstrated property. Please either provide a proof-of-concept with data (real or synthetic), or reframe all such claims as research goals/hypotheses with explicit success criteria.
  2. [§4, 'Learning Causal Structures in Complex, Dynamic Environments'] The causal reliability claim depends on the Causal Learning stage recovering true cause-effect relations from observational data. The paper acknowledges that intermodal terminals involve 'partial data observation, ad-hoc interventions, and streaming constraints' and that terminals have cyclic feedback and non-stationarity, while the cited causal discovery methods 'typically assume dense observability and stationary processes.' This is a direct conflict: the assumptions needed for identifiability are conceded to be absent. No identifiability argument, sensitivity analysis, or experiment is offered to show that the causal graph can be recovered under these conditions, nor that subsequent counterfactual queries (do-calculus) are valid. Without this, the 'causal grounding' of the INSD pipeline is not established. Please specify which causal discovery method is intended, what assumptions it r
  3. [§3.1 and §3.3] The illustrative terminal example (§3.1) is used to argue that the pipeline 'reveals' causes and 'enables' intervention planning, but the example is hypothetical: no actual model outputs, learned equations, or validation are presented. The §3.3 novelty list asserts that the integration is novel without a systematic comparison to existing hybrid SD-ML or neuro-symbolic frameworks. Two of the cited references ([16] and [50]) are the authors' own and provide background concepts, not an analysis of the INSD pipeline's novelty relative to prior hybrid frameworks. Please either provide empirical evidence or clearly label the example as an intended scenario, and support the novelty claim by contrasting the INSD pipeline with the closest existing frameworks at the level of formalization or implementation.
minor comments (5)
  1. [§2.2] The term 'causal reliability' is load-bearing for the INSD claims, but it is only described informally. Please provide a definition that can be operationalized (e.g., what it means for a model to 'reflect underlying causal mechanisms') and tie it to the later pipeline stages.
  2. [Fig. 1] The schematic does not indicate the training objective, loss functions, or the interface between stages. A more detailed data-flow diagram would help readers understand how the three stages interact and what is fed forward from one stage to the next.
  3. [Abstract] The phrase 'We aim at showing how neuro-symbolic methods can bridge the gap...' is appropriate for a roadmap, but it conflicts with the stronger claim in the same abstract that the framework 'enables' the construction of such models. Please align the abstract with the paper's actual contribution.
  4. [§4, 'Towards an Integrated Evaluation Framework'] The evaluation framework is promising but only lists high-level dimensions. Please specify at least one concrete, measurable criterion per pipeline stage (e.g., concept alignment score, structural Hamming distance for the causal graph, equation recovery accuracy, or counterfactual validity test).
  5. [Global] Minor typographical and formatting issues: the 'Conclusion' section has no number; reference [51] contains 'Y .A.' with an extra space; reference [55] renders as 'SAHiN' in all caps. Please also consider consistent use of 'explainability' vs. 'interpretability' given the note in Section 1.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: self-contained conceptual roadmap with no fitted predictions or derivation-by-definition.

full rationale

The paper is a position/roadmap paper proposing the INSD pipeline, which combines concept-based interpretability, causal machine learning/discovery, and equation learning. It contains no equations, no fitted parameters, and no empirical predictions, so there is no derivation chain that could reduce to its own inputs. The central claim—that such a pipeline would yield semantically meaningful, causally grounded, and interpretable models—is a design proposal rather than a derived result. The two self-citations involving co-author Termine (refs. [16] and [50]) provide a taxonomy of opacity and the notion of causal reliability, but they serve as conceptual framing rather than load-bearing evidence; the pipeline's components are also supported by independent references (e.g., [23], [39], [44], [10]), and the proposal does not hinge on the self-citations being correct in any formal sense. Section 4 explicitly acknowledges the hard open challenges for causal discovery in terminal environments (partial data observation, ad-hoc interventions, streaming constraints, cyclic feedback, non-stationarity), which is an honest limitation of the roadmap, not a circular step. No pattern of self-definition, fitted-input-called-prediction, uniqueness-imported-from-authors, or renaming-known-result is present.

Assumptions & free parameters 0 free parameters · 4 assumptions · 1 invented entities

The paper's central claim rests on several unvalidated domain assumptions about the feasibility of concept learning, causal discovery, and equation learning in real logistics terminals. These are explicitly identified as open challenges in Section 4, making them load-bearing but unverified premises.

assumptions (4)
  • domain assumption Causal discovery from observational data can recover true causal structures in intermodal terminal operations.
    Section 4 notes terminals have partial observability, cyclic feedback, and non-stationarity. The pipeline assumes these can be handled, but the paper itself lists this as an open challenge.
  • domain assumption Concept-based interpretability can extract stable, semantically aligned concepts from dynamic streaming sensor data.
    Section 4 states that extending concept methods to streaming data 'demands novel architectures.' The pipeline assumes this is feasible without providing evidence.
  • domain assumption Equation learning methods such as SINDy can recover interpretable dynamics for discontinuous logistics processes.
    Section 4 calls applying these methods to terminal operations 'an open frontier.' The pipeline assumes the necessary equation learners exist.
  • domain assumption System Dynamics models provide a correct causal ground truth for validating learned models.
    Section 2.1 presents SD as a reliable causal modeling tool, even while acknowledging its limitations. The INSD framework relies on this grounding.
invented entities (1)
  • INSD pipeline
    purpose: Unify concept learning, causal learning, and equation learning into one interpretable-by-design modeling framework.
    Proposed in Section 3, no implementation or empirical validation is provided, so no independent falsifiable handle exists.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Towards explainable decision support using hybrid neural models for logistic terminal automation." pith.science (2026). https://pith.science/paper/55UIU3YM

@misc{pith2026250907577,
  author       = {Pith},
  title        = {Pith review of: Towards explainable decision support using hybrid neural models for logistic terminal automation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/55UIU3YM}},
  note         = {Machine review of arXiv:2509.07577}
}
abstract

The integration of Deep Learning (DL) in System Dynamics (SD) modeling for transportation logistics offers significant advantages in scalability and predictive accuracy. However, these gains are often offset by the loss of explainability and causal reliability $-$ key requirements in critical decision-making systems. This paper presents a novel framework for interpretable-by-design neural system dynamics modeling that synergizes DL with techniques from Concept-Based Interpretability, Mechanistic Interpretability, and Causal Machine Learning. The proposed hybrid approach enables the construction of neural network models that operate on semantically meaningful and actionable variables, while retaining the causal grounding and transparency typical of traditional SD models. The framework is conceived to be applied to real-world case-studies from the EU-funded project AutoMoTIF, focusing on data-driven decision support, automation, and optimization of multimodal logistic terminals. We aim at showing how neuro-symbolic methods can bridge the gap between black-box predictive models and the need for critical decision support in complex dynamical environments within cyber-physical systems enabled by the industrial Internet-of-Things.

Figures

Figures reproduced from arXiv: 2509.07577 by the authors.

Figure 1
Figure 1. Overview of the INSD pipeline, from concept learning to causal and equation learning, ensuring interpretability and causal reliability in the resulting model. 3.1 Application of INSD Pipeline to Intermodal Terminals To illustrate how the proposed pipeline operates in practice, let us consider a digital twin simulation of an intermodal freight terminal. Here, a deep learning model fore￾casts recurring train delays du… view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

55 extracted references · 36 canonical work pages

  1. [16]

    In: Müller, V .C

    Facchini, A., Termine, A.: Towards a Taxonomy for the Opacity of AI Systems. In: Müller, V .C. (ed.) Philosophy and Theory of Artificial Intelligence 2021, vol. 63, pp. 73–89. Springer International Publishing, Cham (2022). https://doi.org/10.1007/978-3-031-09153-7_7, se- ries Title: Studies in Applied Philosophy, Epistemology and Rational Ethics

  2. [50]

    In: The Rout- ledge Handbook of Causality and Causal Methods, pp

    Termine, A., Primiero, G.: Causality Problems in Machine Learning Systems. In: The Rout- ledge Handbook of Causality and Causal Methods, pp. 325–341. Routledge, New York, 1 edn. (Dec 2024). https://doi.org/10.4324/9781003528937-37

  3. [1]

    Trans- portation Research Part E: Logistics and Transportation Review186, 103563 (Jun 2024)

    Abdulrashid, I., Zanjirani Farahani, R., Mammadov, S., Khalafalla, M., Chiang, W.C.: Ex- plainable artificial intelligence in transport Logistics: Risk analysis for road accidents. Trans- portation Research Part E: Logistics and Transportation Review186, 103563 (Jun 2024). https://doi.org/10.1016/j.tre.2024.103563

  4. [2]

    Management Research Review 38(5), 505–539 (May 2015)

    Aschauer, G., Gronalt, M., Mandl, C.: Modelling interrelationships between logistics and transportation operations – a system dynamics approach. Management Research Review 38(5), 505–539 (May 2015). https://doi.org/10.1108/MRR-11-2013-0271

  5. [3]

    Oper Res Int J24(2), 23 (Jun 2024)

    Ashraf, M., Eltawil, A., Ali, I.: Disruption detection for a cognitive digital supply chain twin using hybrid deep learning. Oper Res Int J24(2), 23 (Jun 2024). https://doi.org/10.1007/ s12351-024-00831-y

  6. [4]

    Algorithms17(8), 333 (Aug 2024)

    Awasthi, A., Krpalkova, L., Walsh, J.: Deep Learning-Based Boolean, Time Series, Error Detection, and Predictive Analysis in Container Crane Operations. Algorithms17(8), 333 (Aug 2024). https://doi.org/10.3390/a17080333

  7. [5]

    Infor- mation and Software Technology159, 107197 (Jul 2023)

    Balasubramaniam, N., Kauppinen, M., Rannisto, A., Hiekkanen, K., Kujala, S.: Trans- parency and explainability of AI systems: From ethical guidelines to requirements. Infor- mation and Software Technology159, 107197 (Jul 2023). https://doi.org/10.1016/j.infsof. 2023.107197

  8. [6]

    Bellogín, A., Grau, O., Larsson, S., Schimpf, G., Sengupta, B., Solmaz, G.: The EU AI Act and the Wager on Trustworthy AI. Commun. ACM67(12), 58–65 (Dec 2024). https: //doi.org/10.1145/3665322

Show all 55 references
  1. [7]

    https://doi.org/10.48550/arXiv.2404.14082, arXiv:2404.14082 [cs]

    Bereska, L., Gavves, E.: Mechanistic Interpretability for AI Safety – A Review (Aug 2024). https://doi.org/10.48550/arXiv.2404.14082, arXiv:2404.14082 [cs]

  2. [8]

    Neural Comput & Applic36(21), 12809–12844 (Jul 2024)

    Bhuyan, B.P., Ramdane-Cherif, A., Tomar, R., Singh, T.P.: Neuro-symbolic artificial in- telligence: a survey. Neural Comput & Applic36(21), 12809–12844 (Jul 2024). https: //doi.org/10.1007/s00521-024-09960-z

  3. [9]

    In: Merk- ert, R., Hoberg, K

    Boute, R.N., Udenio, M.: AI in Logistics and Supply Chain Management. In: Merk- ert, R., Hoberg, K. (eds.) Global Logistics and Supply Chain Strategies for the 2020s, pp. 49–65. Springer International Publishing, Cham (2023). https://doi.org/10.1007/ 978-3-030-95764-3_3

  4. [10]

    Brunton, S.L., Proctor, J.L., Kutz, J.N.: Discovering governing equations from data by sparse identification of nonlinear dynamical systems. Proc. Natl. Acad. Sci. U.S.A.113(15), 3932– 3937 (Apr 2016). https://doi.org/10.1073/pnas.1517384113

  5. [11]

    Logistics5(2), 25 (Apr 2021)

    Busse, A., Gerlach, B., Lengeling, J.C., Poschmann, P., Werner, J., Zarnitz, S.: Towards Digital Twins of Multimodal Supply Chains. Logistics5(2), 25 (Apr 2021). https://doi.org/ 10.3390/logistics5020025 Title Suppressed Due to Excessive Length 13

  6. [12]

    https://doi.org/10.48550/arXiv.2309.09901, arXiv:2309.09901 [cs]

    Carloni, G., Berti, A., Colantonio, S.: The role of causality in explainable artificial intelli- gence (Sep 2023). https://doi.org/10.48550/arXiv.2309.09901, arXiv:2309.09901 [cs]

  7. [13]

    https://doi.org/10.48550/arXiv.2108.03012, arXiv:2108.03012 [cs]

    Chazette, L., Brunotte, W., Speith, T.: Exploring Explainability: A Definition, a Model, and a Knowledge Catalogue (Aug 2021). https://doi.org/10.48550/arXiv.2108.03012, arXiv:2108.03012 [cs]

  8. [14]

    IEEE Transactions on Industrial Informatics19(12), 11553–11563 (2023)

    De Benedictis, A., Flammini, F., Mazzocca, N., Somma, A., Vitale, F.: Digital twins for anomaly detection in the industrial internet of things: Conceptual architecture and proof-of- concept. IEEE Transactions on Industrial Informatics19(12), 11553–11563 (2023). https: //doi.or...

  9. [15]

    IEEE Trans

    Elkhawaga, G., Elzeki, O.M., Abu-Elkheir, M., Reichert, M.: Why Should I Trust Your Ex- planation? An Evaluation Approach for XAI Methods Applied to Predictive Process Mon- itoring Results. IEEE Trans. Artif. Intell.5(4), 1458–1472 (Apr 2024). https://doi.org/10. 1109/TAI.2024.3357041

  10. [17]

    Sci Eng Ethics26(6), 3333–3361 (Dec 2020)

    Felzmann, H., Fosch-Villaronga, E., Lutz, C., Tamò-Larrieux, A.: Towards Transparency by Design for Artificial Intelligence. Sci Eng Ethics26(6), 3333–3361 (Dec 2020). https: //doi.org/10.1007/s11948-020-00276-4

  11. [18]

    https://doi.org/10.48550/arXiv

    Fernsel, L., Kalff, Y ., Simbeck, K.: Assessing the Auditability of AI-integrating Systems: A Framework and Learning Analytics Case Study (Oct 2024). https://doi.org/10.48550/arXiv. 2411.08906, arXiv:2411.08906 [cs]

  12. [19]

    Flammini, F.: Digital twins as run-time predictive models for the resilience of cyber-physical systems: a conceptual framework. Phil. Trans. R. Soc. A.379(2207), 20200369 (Oct 2021). https://doi.org/10.1098/rsta.2020.0369

  13. [20]

    IEEE Trans

    Flammini, F., Alcaraz, C., Bellini, E., Marrone, S., Lopez, J., Bondavalli, A.: Towards Trust- worthy Autonomous Systems: Taxonomies and Future Perspectives. IEEE Trans. Emerg. Topics Comput.12(2), 601–614 (Apr 2024). https://doi.org/10.1109/TETC.2022.3227113

  14. [21]

    https://doi.org/10.48550/arXiv.1907.07165, arXiv:1907.07165 [cs]

    Goyal, Y ., Feder, A., Shalit, U., Kim, B.: Explaining Classifiers with Causal Concept Effect (CaCE) (Feb 2020). https://doi.org/10.48550/arXiv.1907.07165, arXiv:1907.07165 [cs]

  15. [22]

    Big Data & Society11(1), 20539517241234298 (Mar 2024)

    Högberg, C.: Stabilizing translucencies: Governing AI transparency by standardization. Big Data & Society11(1), 20539517241234298 (Mar 2024). https://doi.org/10.1177/ 20539517241234298

  16. [23]

    https://doi.org/10.48550/ARXIV .2206.15475, version 2

    Kaddour, J., Lynch, A., Liu, Q., Kusner, M.J., Silva, R.: Causal Machine Learning: A Survey and Open Problems (2022). https://doi.org/10.48550/ARXIV .2206.15475, version 2

  17. [24]

    https://doi.org/10.48550/arXiv.1711.11279, arXiv:1711.11279 [stat]

    Kim, B., Wattenberg, M., Gilmer, J., Cai, C., Wexler, J., Viegas, F., Sayres, R.: Inter- pretability Beyond Feature Attribution: Quantitative Testing with Concept Activation Vectors (TCA V) (Jun 2018). https://doi.org/10.48550/arXiv.1711.11279, arXiv:1711.11279 [stat]

  18. [25]

    Engineering Applications of Artifi- cial Intelligence129, 107620 (Mar 2024)

    Kobayashi, K., Alam, S.B.: Explainable, Interpretable & Trustworthy AI for Intelligent Digital Twin: Case Study on Remaining Useful Life. Engineering Applications of Artifi- cial Intelligence129, 107620 (Mar 2024). https://doi.org/10.1016/j.engappai.2023.107620, arXiv:2301.06676 [cs]

  19. [26]

    Interna- tional Journal of Production Research62(4), 1510–1540 (Feb 2024)

    Kosasih, E.E., Papadakis, E., Baryannis, G., Brintrup, A.: A review of explainable arti- ficial intelligence in supply chain management using neurosymbolic approaches. Interna- tional Journal of Production Research62(4), 1510–1540 (Feb 2024). https://doi.org/10.1080/ 00207543....

  20. [27]

    https://doi.org/10.48550/ arXiv.2305.02251, arXiv:2305.02251 [cs] 14 R

    Kramer, S., Cerrato, M., Džeroski, S., King, R.: Automated Scientific Discovery: From Equa- tion Discovery to Autonomous Discovery Systems (May 2023). https://doi.org/10.48550/ arXiv.2305.02251, arXiv:2305.02251 [cs] 14 R. D’Elia et al

  21. [28]

    Euro Jnl Phil Sci 14(4), 52 (Dec 2024)

    Kästner, L., Crook, B.: Explaining AI through mechanistic interpretability. Euro Jnl Phil Sci 14(4), 52 (Dec 2024). https://doi.org/10.1007/s13194-024-00614-4

  22. [29]

    Information & Management61(5), 103969 (Jul 2024)

    Laine, J., Minkkinen, M., Mäntymäki, M.: Ethics-based AI auditing: A systematic literature review on conceptualizations of ethical principles and knowledge contributions to stakehold- ers. Information & Management61(5), 103969 (Jul 2024). https://doi.org/10.1016/j.im.2024. 103969

  23. [30]

    Computers & Industrial Engineering187, 109768 (Jan 2024)

    Le, T.V ., Fan, R.: Digital twins for logistics and supply chain systems: Literature review, conceptual framework, research potential, and practical challenges. Computers & Industrial Engineering187, 109768 (Jan 2024). https://doi.org/10.1016/j.cie.2023.109768

  24. [31]

    https://doi.org/10.48550/arXiv.1705.07874, arXiv:1705.07874 [cs]

    Lundberg, S., Lee, S.I.: A Unified Approach to Interpreting Model Predictions (Nov 2017). https://doi.org/10.48550/arXiv.1705.07874, arXiv:1705.07874 [cs]

  25. [32]

    Sci Rep14(1), 18219 (Aug 2024)

    Luo, C., Li, A.J., Xiao, J., Li, M., Li, Y .: Explainable and generalizable AI-driven multiscale informatics for dynamic system modelling. Sci Rep14(1), 18219 (Aug 2024). https://doi. org/10.1038/s41598-024-67259-4

  26. [33]

    ACM Trans

    Mohseni, S., Zarei, N., Ragan, E.D.: A Multidisciplinary Survey and Framework for Design and Evaluation of Explainable AI Systems. ACM Trans. Interact. Intell. Syst.11(3-4), 1–45 (Dec 2021). https://doi.org/10.1145/3387166

  27. [34]

    Minds & Machines31(2), 323–327 (Jun 2021)

    Mokander, J., Floridi, L.: Ethics-Based Auditing to Develop Trustworthy AI. Minds & Machines31(2), 323–327 (Jun 2021). https://doi.org/10.1007/s11023-021-09557-8, arXiv:2105.00002 [cs]

  28. [35]

    In: Dolgui, A., Bernard, A., Lemoine, D., V on Cieminski, G., Romero, D

    Mugurusi, G., Oluka, P.N.: Towards Explainable Artificial Intelligence (XAI) in Sup- ply Chain Management: A Typology and Research Agenda. In: Dolgui, A., Bernard, A., Lemoine, D., V on Cieminski, G., Romero, D. (eds.) Advances in Production Manage- ment Systems. Artificial In...

  29. [36]

    Cambridge University Press, 2 edn

    Pearl, J.: Causality: Models, Reasoning, and Inference. Cambridge University Press, 2 edn. (Sep 2009). https://doi.org/10.1017/CBO9780511803161

  30. [37]

    Logistics6(2), 35 (May 2022)

    Phan, T.L.J., Gehrhardt, I., Heik, D., Bahrpeyma, F., Reichelt, D.: A Systematic Mapping Study on Machine Learning Techniques Applied for Condition Monitoring and Predictive Maintenance in the Manufacturing Sector. Logistics6(2), 35 (May 2022). https://doi.org/ 10.3390/logisti...

  31. [38]

    Sustainable Futures9, 100526 (Jun 2025)

    Pirouzrahi, Z., Vanelslander, T., Nassiri Aghdam, A.: Applying system dynamics modelling to modal shift: A systematic review. Sustainable Futures9, 100526 (Jun 2025). https://doi. org/10.1016/j.sftr.2025.100526

  32. [39]

    https://doi.org/10.48550/arXiv.2312.12936, arXiv:2312.12936 [cs]

    Poeta, E., Ciravegna, G., Pastor, E., Cerquitelli, T., Baralis, E.: Concept-based Explainable Artificial Intelligence: A Survey (Dec 2023). https://doi.org/10.48550/arXiv.2312.12936, arXiv:2312.12936 [cs]

  33. [40]

    JAIBD1(1), 139–150 (Dec 2021)

    Puli, V .O.R., Yasmeen, Z.: Predictive Analytics and Deep Learning for Logistics Optimiza- tion in Supply Chain Management. JAIBD1(1), 139–150 (Dec 2021). https://doi.org/10. 31586/jaibd.2021.1187

  34. [41]

    Nonlinear Dyn99(2), 1709–1761 (Jan 2020)

    Quaranta, G., Lacarbonara, W., Masri, S.F.: A review on computational intelligence for iden- tification of nonlinear dynamical systems. Nonlinear Dyn99(2), 1709–1761 (Jan 2020). https://doi.org/10.1007/s11071-019-05430-7

  35. [42]

    Why Should I Trust You?

    Ribeiro, M.T., Singh, S., Guestrin, C.: "Why Should I Trust You?": Explaining the Predictions of Any Classifier (Aug 2016). https://doi.org/10.48550/arXiv.1602.04938, arXiv:1602.04938 [cs]

  36. [43]

    J of Business Logistics44(4), 532–549 (Oct 2023)

    Richey, R.G., Chowdhury, S., Davis-Sramek, B., Giannakis, M., Dwivedi, Y .K.: Artificial intelligence in logistics and supply chain management: A primer and roadmap for research. J of Business Logistics44(4), 532–549 (Oct 2023). https://doi.org/10.1111/jbl.12364 Title Suppress...

  37. [44]

    Scholkopf, B., Locatello, F., Bauer, S., Ke, N.R., Kalchbrenner, N., Goyal, A., Bengio, Y .: Toward Causal Representation Learning. Proc. IEEE109(5), 612–634 (May 2021). https: //doi.org/10.1109/JPROC.2021.3058954

  38. [45]

    ENG18(5), e170423215860 (Jul 2024)

    Sharma, J., Lal Mittal, M., Soni, G., Keprate, A.: Explainable Artificial Intelligence (XAI) Approaches in PredictiveMaintenance: A Review. ENG18(5), e170423215860 (Jul 2024). https://doi.org/10.2174/1872212118666230417084231

  39. [46]

    European Journal of Operational Research310(3), 1201–1217 (Nov 2023)

    Sobrie, L., Verschelde, M., Hennebel, V ., Roets, B.: Capturing complexity over space and time via deep learning: An application to real-time delay prediction in railways. European Journal of Operational Research310(3), 1201–1217 (Nov 2023). https://doi.org/10.1016/j. ejor.2023.03.040

  40. [47]

    https://doi.org/10.48550/arXiv.2112

    Sokol, K., Flach, P.: Explainability Is in the Mind of the Beholder: Establishing the Founda- tions of Explainable Artificial Intelligence (Sep 2022). https://doi.org/10.48550/arXiv.2112. 14466, arXiv:2112.14466 [cs]

  41. [48]

    Soumpenioti, V ., Panagopoulos, A.: AI Technology in the Field of Logistics. In: 2023 18th International Workshop on Semantic and Social Media Adaptation & Personalization (SMAP)18th International Workshop on Semantic and Social Media Adaptation & Personal- ization (SMAP 2023)...

  42. [49]

    Future Generation Computer Systems145, 1–11 (Aug 2023)

    Tao, M.: Semantic ontology enabled modeling, retrieval and inference for incomplete mobile trajectory data. Future Generation Computer Systems145, 1–11 (Aug 2023). https://doi.org/ 10.1016/j.future.2023.03.012

  43. [51]

    https://doi.org/10.48550/arXiv

    Wang, K., Varambally, S., Watson-Parris, D., Ma, Y .A., Yu, R.: Discovering Latent Struc- tural Causal Models from Spatio-Temporal Data (Nov 2024). https://doi.org/10.48550/arXiv. 2411.05331, arXiv:2411.05331 [cs]

  44. [52]

    https://doi.org/10.48550/arXiv.2504.03692, arXiv:2504.03692 [cs]

    Wasi, A.T., Anik, M.A., Rahman, A., Hoque, M.I., Islam, M.S., Ahsan, M.M.: A Theoretical Framework for Graph-based Digital Twins for Supply Chain Management and Optimization (Mar 2025). https://doi.org/10.48550/arXiv.2504.03692, arXiv:2504.03692 [cs]

  45. [53]

    Chaos, Solitons & Fractals165, 112866 (Dec 2022)

    Wei, B.: Sparse dynamical system identification with simultaneous structural parameters and initial condition estimation. Chaos, Solitons & Fractals165, 112866 (Dec 2022). https://doi. org/10.1016/j.chaos.2022.112866

  46. [54]

    Hum-Cent Intell Syst3(3), 161–188 (Aug 2023)

    Yang, W., Wei, Y ., Wei, H., Chen, Y ., Huang, G., Li, X., Li, R., Yao, N., Wang, X., Gu, X., Amin, M.B., Kang, B.: Survey on Explainable AI: From Approaches, Limitations and Applications Aspects. Hum-Cent Intell Syst3(3), 161–188 (Aug 2023). https://doi.org/10. 1007/s44230-02...

  47. [55]

    Neural Comput & Applic 37(2), 859–965 (Jan 2025)

    ¸ SAHiN, E., Arslan, N.N., Özdemir, D.: Unlocking the black box: an in-depth review on interpretability, explainability, and reliability in deep learning. Neural Comput & Applic 37(2), 859–965 (Jan 2025). https://doi.org/10.1007/s00521-024-10437-2

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.