Pith. sign in

REVIEW 4 major objections 5 minor 42 references

Searching for actual causes: Approximate algorithms with adjustable precision

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 ↗ Review JSON

Pith's one-line read This paper claims that Halpern-Pearl actual causes can be identified by a beam search over counterfactual interventions, using a user-supplied oracle, in polynomial time with adjustable precision.

desk verdict A useful, honest engineering contribution to approximate HP actual causes; the formal precision claim is weaker than advertised, but the empirical scope is new and the code is real. read the letter →

arxiv 2507.07857 v1 pith:BOTP6ZBQ submitted 2025-07-10 cs.AI

Samuel Reyd , Ada Diaconescu , Jean-Louis Dessalles This is my paper · ORCID
classification cs.AI
keywords actualcausationHalpern-PearlcausesbeamsearchcounterfactualreasoningstructuralcausalmodelsexplainableAIstochasticsystemsapproximatealgorithms
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 tackles the problem of identifying actual causes—facts that, if they had not occurred, would have prevented a given outcome—as opposed to general causal relationships. Its target is the Halpern-Pearl (HP) definition of actual cause, which has been computationally hard to use in practice. The authors propose a beam search over the space of counterfactual interventions that reports minimal causes, using a user-supplied answer function (oracle) to decide whether the outcome would still hold under each intervention. They claim the search runs in expected polynomial time in the number of variables, can be tuned by a beam-size parameter to trade runtime against precision and exhaustiveness, and works for non-Boolean, black-box, and stochastic systems. The paper demonstrates the approach on a Boolean scenario from the literature and on non-Boolean, black-box, and noisy variants of it.

What carries the argument

The central object is the space of interventions $E$: each element is a set of variable-value pairs, split into the counterfactual pairs $e_C$ and the actual-value contingency pairs $e_W$. An HP cause $C$ with contingency set $W$ is exactly the counterfactual part of some intervention $e$ that cancels the target predicate ($\phi(e)=0$) and is minimal under inclusion. The search itself is a beam search: at each depth it keeps only the $b$ most promising partial interventions according to a heuristic $\psi$, uses the oracle $\phi$ to decide whether the target still holds, and discards candidates that are supersets of causes already found. The ISI variant exploits the causal graph by replacing variables in a found cause with their parents and searching smaller sub-instances, and the LUCB variant replaces the deterministic oracle with confidence-bounded sampling when evaluations are stochastic.

What would settle it

Take a small Boolean SCM whose HP-cause set is known by exhaustive enumeration, run the base algorithm with an exact oracle and a beam wide enough to keep every candidate at every depth, and compare the output; if the output differs from the enumeration, the claimed intervention-space encoding or the minimality filter is wrong.

Watch

Extended reading notes

Core claim

The paper's central claim is that every HP cause of a target fact can be represented as an intervention $e$ in the space of variable-value assignments, where the variables set to counterfactual values form the cause $e_C$ and the variables kept at their actual values form the contingency set $e_W$. Finding the HP causes therefore reduces to finding interventions for which the target no longer holds ($\phi(e)=0$) and whose counterfactual part is minimal under inclusion. The base algorithm performs a beam search guided by a heuristic, evaluates candidates through the oracle $\phi$, and prunes any candidate that is a superset of an already found cause; the ISI algorithm repeats this search on sub-instances built from the causal graph; and the LUCB algorithm estimates $\phi(e)$ by confidence-bounded sampling when the system is stochastic. The authors claim this yields a polynomial-time approximation of the full HP-cause set, with complexity $O(|V|^2 \times |D_{\max}| \times b \times N_C \times |C_{\max}|)$, and that the beam size $b$ is an adjustable precision-exhaustiveness-runtime tradeoff.

Load-bearing premise

Correctness depends entirely on the user-supplied answer function: it must truthfully report whether the target outcome would still happen under every conceivable intervention, including counterfactual worlds that may be unrealistic or unobservable.

Editorial extensions

If this is right

  • Actual-cause identification no longer requires explicit structural equations: any system that can answer counterfactual queries, such as a simulator or a hand-built expert model, can be searched directly.
  • Non-Boolean discrete domains are handled natively, so causes can be reported as variable-value facts rather than only as boolean on/off events.
  • For stochastic systems, repeated sampling and confidence bounds turn the search into approximate identification with user-set tolerances and sample budgets.
  • The beam-size parameter is a practical control: larger beams increase precision and exhaustiveness at roughly linear runtime cost.
  • When the causal graph is known, the ISI algorithm generally improves both accuracy and runtime by decomposing the search into smaller sub-instances.

Reading between the lines

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

  • If the oracle is a learned counterfactual model, its errors propagate directly into the reported causes; a natural next experiment is measuring how oracle noise degrades precision and recall.
  • The intervention-space encoding connects minimal HP causes to $L_0$-distance counterfactual explanations, suggesting the same beam search could serve as an explanation engine for black-box classifiers.
  • The confidence-bounds machinery in LUCB points toward a formal anytime guarantee: as the sample budget grows, the estimated $\bar\phi(e)$ converges to $\phi^*(e)$, so a PAC-style statement of the search's precision is a testable extension.
  • Because the reference sets for the non-Boolean and black-box experiments are assembled from the algorithms' own outputs, an independent exact enumeration for those settings would be a natural follow-up.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

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 proposes three algorithms for identifying Halpern-Pearl (HP) actual causes in structural causal models: a beam-search base algorithm that explores interventions with an oracle for the target predicate, an iterative sub-instance identification (ISI) variant that uses the causal graph, and a LUCB-based stochastic evaluation method. The authors claim polynomial complexity and an adjustable precision-exhaustiveness tradeoff, and present experiments on Boolean, non-Boolean, black-box, and stochastic versions of the 'Steal Master Key' scenario. The paper is clearly written and includes code, but the central approximation guarantee is not established, and the evaluation for the non-Boolean and black-box cases relies on reference sets built from the algorithms' own outputs.

Significance. If the claims held, the work would fill a real gap: practical identification of HP actual causes in non-Boolean, black-box, and stochastic systems, beyond the Boolean logic-formula setting of ILP-based approaches. Strengths include the explicit oracle-based formulation, the concrete algorithmic descriptions, the availability of code, and the use of an exact exhaustive reference set for the base and noisy SCMs. The empirical comparison with ILP is useful. However, the central notion of 'adjustable precision' is not backed by any formal bound relating beam size to precision or recall, and the experiments for non-Boolean and black-box systems use circular reference sets. The polynomial complexity claim is output-sensitive and not a worst-case guarantee. These gaps are load-bearing for the paper's main contributions, but they are fixable within the manuscript's scope by either adding guarantees, weakening the claims, or providing independent ground truth for small instances.

major comments (4)
  1. [4.3.1-4.3.2, Algorithm 1] The central claim that Algorithm 1 identifies HP causes with adjustable precision is not supported. In Section 4.3.2, filterMinimality only discards an element if a subset of its cause set has already been identified as a cause. Because beam search prunes nodes at each depth, a true HP cause can be pruned before being evaluated, and a superset of it can later be generated from another branch, evaluate to φ(e)=0, and be reported as a cause. This yields both false positives (AC3 violations) and false negatives even with a perfect oracle. No theorem or bound is given that relates precision or recall to the beam size b, the heuristic ψ, or the search depth. The experiments in Annex H quantify 'Overshoot' (non-minimal causes) and show it does not clearly decrease with beam size, confirming that the issue is real. Without such a bound, the phrase 'adjustable precision' is not justified.
  2. [5.6.2] The reference sets for the non-Boolean and black-box SCMs are constructed by running the algorithms themselves: the text states that for these models the reference set combines all identified causes across beam sizes and then filters non-minimal ones, and that precision and recall are 'an upper bound instead of being exact.' Annex H further notes that a recall value of 1 in the non-Boolean SCM is an artifact of this computation method. This circularity means the experimental results cannot support the abstract's claim that the algorithms identify causes for non-Boolean and black-box systems. The exact reference set used for the base and noisy SCMs (via unlimited beam) is a genuine benchmark and should be emphasized, but the broader claim needs either an independent ground truth for small instances or a clear statement that the non-Boolean and black-box results are illustrative only.
  3. [4.3.4 and Annex C] The claimed polynomial complexity is output-sensitive and not a worst-case guarantee. The stated bound O(|V|^2 × |D_max| × b × N_C × |C_max|) depends on N_C, the number of identified causes, which can be exponential in |V|. Section 4.3.4 acknowledges that 'N_C depends on the parameter of interest in an unclear way' but still concludes expect polynomial complexity. Annex C's empirical regressions are specific to the SMK scenario and show several regimes with low R² values (e.g., 0.099, 0.121, 0.283 in Fig. C.2), so they do not substantiate a general polynomial bound. The complexity claim should be restated as output-sensitive, or a bound on N_C should be provided under stated assumptions.
  4. [4.5 and Annex F] The LUCB algorithm is presented as a way to 'control the reliability' of stochastic cause identification, but no correctness or high-probability guarantee is proved. The stop conditions (1)-(3) are expressed in terms of sample averages and confidence bounds, yet there is no theorem showing that when the conditions are met, the true Bernoulli parameters satisfy the corresponding inequalities with the stated tolerance and confidence. The user-specified maximum number of samples in Algorithm 4 can terminate the procedure before the conditions are met, and the experimental results in Section 6.4 show LUCB sometimes performing worse than naive sampling. Without an analysis, the stochastic extension does not provide a formal reliability guarantee.
minor comments (5)
  1. [2.3] The name 'Halpren and Hitchcock' should be 'Halpern and Hitchcock'.
  2. [5.4] Equation (4) defines a distribution for X but does not explain the temporal dependence: the text says the value is flipped 'each time it is set,' which suggests a process over time, but the equation is given as a static conditional distribution. Please clarify the relationship between the flipping process and the equation.
  3. [Annex F, Algorithm 4] The functions updateUB and updateLB are referenced in Algorithms 5-7 but never defined; the confidence-bound formulas should be given explicitly so the reader can see how the bounds are computed and how they depend on the step t.
  4. [6.1] The label 'Base SMC' should read 'Base SCM'.
  5. [Figure H.7] The heuristic name 'okham' should be 'Occam' for consistency with the text.

Circularity Check

1 steps flagged · score 4.0 of 10

Core search algorithms are not circular, but the non-Boolean and black-box evaluation reference sets are built from the algorithms' own outputs, so the reported F1 scores for those systems are partly self-defined.

  1. self definitional [Section 5.6.2 and Annex H.1]
    "To produce a reference set of causes, we created a set of all identified causes for a given context, combining the sets from every beam size used, from the base and ISI algorithms (when both were used). We then filtered out the non-minimal causes from this combined set. ... The recall score in the non-boolean SCM reaching 1 is an artefact of the computation method. Since our reference set of causes is built from the identified causes, when we find no cause at all, we have no reference."

    The reference set used to measure precision and recall is not independently derived from the HP definition; it is defined as the union of causes identified by the algorithms themselves at various beam sizes. For the non-Boolean SCM, no independent minimality check (beam size -1) was performed, and for the black-box SCM the reference set is likewise generated from algorithm outputs. Therefore the reported F1, precision, and recall are partly measuring the algorithms against their own outputs; the 'expected causes' are by construction causes the algorithms already found. The paper explicitly acknowledges the recall value of 1 in the non-Boolean case is an artefact of this construction.

full rationale

The central identification algorithms are not circular in the sense of deriving their target from their inputs: Algorithm 1 is a beam search over interventions, evaluated by an oracle and filtered by minimality checks on the causes already found; Algorithms 2 and 3 build on this search using causal-graph structure and bandit confidence bounds, respectively. These constructions are independent of the HP definition's conclusion: the oracle supplies AC2 evidence, the algorithm checks minimality approximately, and the result is not built into the inputs. The beam-size pruning gap identified by the skeptical analysis is a correctness or approximation-guarantee issue, not a circularity: the absence of a bound relating beam size to false positives/negatives means the 'adjustable precision' claim is empirically supported but not proven, which is a validity concern but not a definitional reduction. There are self-citations (e.g., [34], [35], [36], [17]) but they are contextual (related work, relevance, and roadmap) and not load-bearing for the derivation; the formal load-bearing results are cited from external sources (Halpern, Lowerre, Kaufmann and Kalyanakrishnan). The one concrete circularity is in the evaluation methodology for the non-Boolean and black-box SCMs: the reference set is constructed from the union of the algorithms' own identified causes, so the measured agreement is partly self-referential. This affects the empirical claims but not the formal algorithm design, hence the score is moderate rather than high.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

The method introduces no new physical entities or fitted constants; it is an algorithmic procedure whose guarantees rest on the user-provided oracle, the discrete-domain assumption, and the heuristic quality. The main free knobs are beam size, epsilon, and tolerances, which are explicitly traded against runtime.

free parameters (5)
  • beam size b = varied from 1 to 50 (and 0 to 100 for smallest-cause task)
    Controls how many promising nodes are kept per step; the precision/exhaustiveness tradeoff depends entirely on it, and the paper gives no principled way to set it a priori (Sections 5.5 and 7.3).
  • cancellation threshold epsilon = 0.3
    In stochastic settings, an intervention is considered to cancel the consequence if estimated probability is below this threshold; chosen by hand (Section 5.5).
  • noise level epsilon_n = 0.01
    Noise added to structural equations in the stochastic SCM to construct a testbed; an experimental setup choice, not fitted.
  • LUCB tolerances tc, tnc, tb = 1%, 1%, 10%
    Stop conditions for the stochastic evaluator; selected by hand with no sensitivity analysis (Section 5.5).
  • batch size = 10
    LUCB sampling batch, chosen by hand; Figure 4 varies batch sizes 2, 5, and 10.
assumptions (5)
  • domain assumption The Halpern-Pearl definition (AC1-AC3) is the target notion of actual cause.
    The paper explicitly adopts HP causation as the baseline (Section 2.2) and does not defend it against alternative definitions.
  • domain assumption Oracle phi correctly evaluates the target predicate under any intervention.
    Input to Algorithm 1; the algorithm's correctness is conditional on this oracle being accurate for every counterfactual (Section 4.1), which the authors acknowledge is hard to obtain (Section 7.3).
  • domain assumption Every endogenous variable has a discrete and finite domain.
    Assumed in Section 4.1; continuous domains are left to future work (Section 7.3).
  • ad hoc to paper The heuristic function psi is informative enough that top-b nodes contain the relevant causes.
    Beam search quality depends on psi; the paper uses a hand-designed 'minimize positive variables' heuristic and one system-agnostic 'Occam' heuristic, with no guarantee for arbitrary systems (Section 4.1 and Annex H.3).
  • domain assumption For ISI, a superset of the true causal-parent relation is known.
    ISI requires the DAG or a superset of PA_X for all X (Section 4.4); using too loose a superset degrades performance (Annex G.1).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Searching for actual causes: Approximate algorithms with adjustable precision." pith.science (2026). https://pith.science/paper/BOTP6ZBQ

@misc{pith2026250707857,
  author       = {Pith},
  title        = {Pith review of: Searching for actual causes: Approximate algorithms with adjustable precision},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/BOTP6ZBQ}},
  note         = {Machine review of arXiv:2507.07857}
}
read the original abstract

Causality has gained popularity in recent years. It has helped improve the performance, reliability, and interpretability of machine learning models. However, recent literature on explainable artificial intelligence (XAI) has faced criticism. The classical XAI and causality literature focuses on understanding which factors contribute to which consequences. While such knowledge is valuable for researchers and engineers, it is not what non-expert users expect as explanations. Instead, these users often await facts that cause the target consequences, i.e., actual causes. Formalizing this notion is still an open problem. Additionally, identifying actual causes is reportedly an NP-complete problem, and there are too few practical solutions to approximate formal definitions. We propose a set of algorithms to identify actual causes with a polynomial complexity and an adjustable level of precision and exhaustiveness. Our experiments indicate that the algorithms (1) identify causes for different categories of systems that are not handled by existing approaches (i.e., non-boolean, black-box, and stochastic systems), (2) can be adjusted to gain more precision and exhaustiveness with more computation time.

Figures

Figures reproduced from arXiv: 2507.07857 by the authors.

Figure 1
Figure 1. Diagram for the original SMK scenario with 3 attackers. [PITH_FULL_IMAGE:figures/full_fig_p010_1.png] view at source ↗
Figure 2
Figure 2. F1 score and runtime for full causes identification in our 4 versions of SMK [PITH_FULL_IMAGE:figures/full_fig_p012_2.png] view at source ↗
Figure 3
Figure 3. Time and accuracy vs. beam size and num [PITH_FULL_IMAGE:figures/full_fig_p012_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: F1 score and runtime for various numbers [PITH_FULL_IMAGE:figures/full_fig_p013_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

42 extracted references · 41 canonical work pages

  1. [1]

    Albantakis, W

    L. Albantakis, W. Marshall, E. Hoel, and G. Tononi. What Caused What? A Quantitative Account of Actual Causation Using Dynamical Causal Networks. Entropy, 21(5):459, May 2019

    work page 2019

  2. [2]

    Aleksandrowicz, H

    G. Aleksandrowicz, H. Chockler, J. Y. Halpern, and A. Ivrii. The Computational Complexity of Structure-Based Causality. Journal of Artificial Intelligence Research, 58:431–451

    work page

  3. [3]

    Alzubaidi, J

    L. Alzubaidi, J. Bai, A. Al-Sabaawi, J. San- tamar ´ ıa, A. S. Albahri, B. S. N. Al-dabbagh, M. A. Fadhel, M. Manoufali, J. Zhang, A. H. Al-Timemy, Y. Duan, A. Abdullah, L. Farhan, Y. Lu, A. Gupta, F. Albu, A. Abbosh, and Y. Gu. A survey on deep learning tools deal- ing with data scarcity: Definitions, challenges, solutions, tips, and applications. Journa...

    work page 2023

  4. [4]

    C. K. Assaad, E. Devijver, and E. Gaussier. Sur- vey and Evaluation of Causal Discovery Methods for Time Series. Journal of Artificial Intelligence Research, 73:767–819, Feb. 2022

    work page 2022

  5. [5]

    S. Beckers. Causal Sufficiency and Actual Causation. Journal of Philosophical Logic, 50(6):1341–1374, Dec. 2021

    work page 2021

  6. [6]

    D. C. Castro, I. Walker, and B. Glocker. Causality matters in medical imaging. Nature Communications, 11(1):3673, July 2020

    work page 2020

  7. [7]

    Chockler and J

    H. Chockler and J. Y. Halpern. Responsi- bility and Blame: A Structural-Model Ap- proach. Journal of Artificial Intelligence Research, 22:93–115, Oct. 2004

    work page 2004

  8. [8]

    Y.-L. Chou, C. Moreira, P. Bruza, C. Ouyang, and J. Jorge. Counterfactuals and causability in explainable artificial intelligence: Theory, al- gorithms, and applications. Information Fusion, 81:59–83, May 2022

    work page 2022

Show all 42 references
  1. [9]

    Chuck, S

    C. Chuck, S. Vaidyanathan, S. Giguere, A. Zhang, D. Jensen, and S. Niekum. Automated Discovery of Functional Actual 15 Causes in Complex Environments, Apr. 2024. arXiv:2404.10883

    2024 arXiv

  2. [10]

    Dessalles

    J.-L. Dessalles. Algorithmic Simplicity and Relevance, pages 119–130. Springer Berlin Hei- delberg, Berlin, Heidelberg, 2013

    2013

  3. [11]

    R. Guo, L. Cheng, J. Li, P. R. Hahn, and H. Liu. A Survey of Learning Causality with Data: Problems and Methods. ACM Computing Surveys, 53(4):75:1–75:37, July 2020

    2020

  4. [12]

    J. Y. Halpern. A modification of the Halpern- Pearl definition of causality. In Proceedings of the 24th International Conference on Artificial Intelligence, IJCAI’15, pages 3022–

  5. [13]

    J. Y. Halpern and C. Hitchcock. Graded Cau- sation and Defaults. The British Journal for the Philosophy of Science, 66(2):413–457, June 2015

    2015

  6. [14]

    J. Y. Halpern and J. Pearl. Causes and Expla- nations: A Structural-Model Approach. Part I: Causes. The British Journal for the Philosophy of Science, 56(4):843–887

  7. [15]

    J. Y. Halpern and J. Pearl. Causes and explanations: A structural-model approach: Part i: Causes. In Proceedings of the Seventeenth Conference on Uncertainty in Artificial Intelligence, UAI’01, pages 194–202. Morgan Kaufmann Publishers Inc

  8. [16]

    M. Hopkins. Strategies for determining causes of events. In Eighteenth National Conference on Artificial Intelligence, pages 546–552. American Association for Artificial Intelligence, 2002

    2002

  9. [17]

    Houze, J.-L

    E. Houze, J.-L. Dessalles, A. Diaconescu, and D. Menga. What Should I Notice? Using Algorithmic Information Theory to Evaluate the Memorability of Events in Smart Homes. Entropy, 24(3):346

  10. [18]

    Houz´ e, J.-L

    ´E. Houz´ e, J.-L. Dessalles, A. Diaconescu, D. Menga, and M. Schumann. A Decentral- ized Explanatory System for Intelligent Cyber- Physical Systems. In K. Arai, editor, Intelligent Systems and Applications, Lecture Notes in Net- works and Systems, pages 719–738, Cham, 2022. S...

    2022

  11. [19]

    D. Hume. A Treatise of Human Nature. In E. S. Radcliffe, R. McCarty, F. Allhoff, and A. Vaidya, editors, Late Modern Philosophy: Essential Readings with Commentary. Wiley- Blackwell, 2007

    2007

  12. [20]

    Ibrahim, T

    A. Ibrahim, T. Klesel, E. Zibaei, S. Kacianka, and A. Pretschner. Actual Causality Canvas: A General Framework for Explanation-Based Socio-Technical Constructs. In ECAI 2020, pages 2978–2985. IOS Press

    2020

  13. [21]

    Ibrahim and A

    A. Ibrahim and A. Pretschner. From Check- ing to Inference: Actual Causality Computa- tions as Optimization Problems. In D. V. Hung and O. Sokolsky, editors,Automated Technology for Verification and Analysis, pages 343–359. Springer International Publishing, 2020

    2020

  14. [22]

    Ibrahim, S

    A. Ibrahim, S. Rehwald, and A. Pretschner. Efficient Checking of Actual Causality with SAT Solving. In Engineering Secure and Dependable Software Systems, pages 241–255. IOS Press, 2019

    2019

  15. [23]

    T. F. Icard, J. F. Kominsky, and J. Knobe. Nor- mality and actual causal strength. Cognition, 161:80–93, Apr. 2017

    2017

  16. [24]

    Kahneman and D

    D. Kahneman and D. T. Miller. Norm the- ory: Comparing reality to its alternatives. Psychological Review, 93(2):136–153, 1986

    1986

  17. [25]

    Kaufmann and S

    E. Kaufmann and S. Kalyanakrishnan. Informa- tion Complexity in Bandit Subset Selection. In Proceedings of the 26th Annual Conference on Learning Theory, pages 228–251. PMLR, June 2013

    2013

  18. [26]

    D. Lewis. Causation. The Journal of Philosophy, 70(17):556–567, May 1974

    1974

  19. [27]

    B. T. Lowerre. The harpy speech recognition system. PhD thesis, USA, 1976. AAI7619331. 16

    1976

  20. [28]

    Madumal, T

    P. Madumal, T. Miller, L. Sonenberg, and F. Vetere. Explainable Reinforcement Learn- ing through a Causal Lens. Proceedings of the AAAI Conference on Artificial Intelligence, 34(03):2493–2500

  21. [29]

    T. Miller. Explanation in artificial intelligence: Insights from the social sciences. Artificial Intelligence, 267:1–38, Feb. 2019

    2019

  22. [30]

    T. Miller. Contrastive explanation: A structural-model approach. The Knowledge Engineering Review, 36:e14, Jan. 2021

    2021

  23. [31]

    Moraffah, M

    R. Moraffah, M. Karami, R. Guo, A. Raglin, and H. Liu. Causal Interpretability for Ma- chine Learning - Problems, Methods and Evalu- ation. ACM SIGKDD Explorations Newsletter, 22(1):18–33, May 2020

    2020

  24. [32]

    J. Pearl. Causality. Cambridge University Press, Cambridge, 2 edition, 2009

    2009

  25. [33]

    Rafieioskouei and B

    A. Rafieioskouei and B. Bonakdarpour. Effi- cient Discovery of Actual Causality Using Ab- straction Refinement. IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, 43(11):4274–4285, Nov. 2024

    2024

  26. [34]

    S. Reyd, A. Diaconescu, and J.-L. Dessalles. CIRCE: A Scalable Methodology for Causal Explanations in Cyber-Physical Systems. In 2024 IEEE International Conference on Autonomic Computing and Self-Organizing Systems (ACSOS), pages 81–90, Sept. 2024

    2024

  27. [35]

    S. Reyd, A. Diaconescu, and J.-L. Dessalles. Finding relevant causes in complex systems: A generic method adaptable to users and contexts,

  28. [36]

    S. Reyd, A. Diaconescu, J.-L. Dessalles, and L. Esterle. A Roadmap for Causality Research in Complex Adaptive Systems. In 2024 IEEE International Conference on Autonomic Computing and Self-Organizing Systems Companion (ACSOS-C), pages 35–40, Sept. 2024

    2024

  29. [37]

    Scholkopf, F

    B. Scholkopf, F. Locatello, S. Bauer, N. R. Ke, N. Kalchbrenner, A. Goyal, and Y. Ben- gio. Toward Causal Representation Learning. Proceedings of the IEEE, 109(5):612–634, May 2021

    2021

  30. [38]

    Schwab and W

    P. Schwab and W. Karlen. CXPlain: Causal Explanations for Model Interpre- tation under Uncertainty. In Advances in Neural Information Processing Systems, vol- ume 32. Curran Associates, Inc

  31. [39]

    Sch¨ olkopf, D

    B. Sch¨ olkopf, D. Janzing, J. Peters, E. Sgouritsa, K. Zhang, and J. Mooij. On causal and anticausal learning. In Proceedings of the 29th International Coference on International Conference on Machine Learning, ICML’12, pages 459–466. Omnipress

  32. [40]

    ST, ¬SH

    S. Triantafyllou, A. Singla, and G. Radanovic. Actual Causality and Responsibility Attribution in Decentralized Partially Observable Markov Decision Processes. In Proceedings of the 2022 AAAI/ACM Conference on AI, Ethics, and Society, AIES ’22, pages 739–752, New York, NY, USA...

    2022

  33. [42]

    candidate causes

    We first sample one batch of each element in the beam. We then update the confidences until they are all under the corresponding tolerance threshold, or we reach the maximum number of samples set by the user. When we update the confidence for the beam (see Algorithm 5), we fir...

  34. [2025]

    In IEEE International Conference on Au- tonomic Computing and Self-Organizing Sys- tems (ACSOS) (accepted)

Pith tools

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