What Moved the Alarm Score? Causal Attribution with Exact Computation and Scarce-Incident Estimation

Masaaki Takada; Ryusei Shingaki

stat.MLOct 20, 2026

What Moved the Alarm Score? Causal Attribution with Exact Computation and Scarce-Incident Estimation

Authors: Masaaki Takada, Ryusei Shingaki

Organizations: Toshiba Corporation

Abstract

Alarms in industrial systems and scientific instruments often demand immediate diagnosis despite scarce incident data. To guide intervention, an explanation must trace how changes in the data-generating process propagate through the causal structure to alter the alarm score, rather than merely flagging shifts that may be downstream effects. Yet exact attribution across interacting causal components appears combinatorial, while existing methods often either assume an unchanged causal model or require enough incident data to estimate the changed system reliably. We address this setting by formulating the alarm-score change as a Shapley game over hybrid structural causal model (SCM) states. Treating structural functions and exogenous laws as distinct players yields a finer-grained, function/law-resolved report of how each type of change propagates to the alarm score. The same framework defines realized-event and population reports, with an event--population coherence identity linking their attributions. For ordered linear SCMs and quadratic scores, a causal first- and second-moment recursion combined with quadrature computes the full Shapley vector exactly in polynomial time, eliminating combinatorial computation. To estimate this target from scarce incident data, we adapt a well-observed reference SCM to the incident regime using cross-fitted estimates of mechanism changes and residual moments, and derive attribution-error bounds and top-k recovery guarantees. Experiments on controlled and measured shifts illustrate the diagnoses and assess scaling to hundreds of variables and estimation from small incident samples.

Figures & tables

Report	Causal question	Source	Target	Contrast	Players	Causal estimand
Population shift	Which regime changes moved the mean score?	$M(F^{(0)},Q^{(0)})$	$M(F^{(1)},Q^{(1)})$	$T(M^{(1)})-T(M^{(0)})$	$\mathcal{U}_{\mathtt{F}\cup\mathtt{Q}}$	$\Phi_{u}$ : hybrid expected-score game
Shifted event	Which target-state components explain this event?	$M(F^{(0)},Q^{(0)})$	$M(F^{(1)},\delta_{e^{(1)}})$	$\tau(x)-T(M^{(0)})$	$\mathcal{U}_{\mathtt{F}\cup\mathtt{Q}}$	$\Phi_{u}^{\mathrm{shift}}(e^{(1)})$ : abducted-event game
Fixed-SCM event	Which disturbances explain the event under $M^{(0)}$ ?	$M(F^{(0)},Q^{(0)})$	$M(F^{(0)},\delta_{e^{(0)}})$	$\tau(x)-T(M^{(0)})$	$\mathcal{U}_{\mathtt{Q}}$	$\Phi_{u}^{\mathrm{fix}}(e^{(0)})$ : fixed-function event game

Table 1: Three representative causal score-attribution questions in the source–target–player–score framework.

Figure 1: Same-oracle Shapley aggregation on the full

K=2d

function/law game (mean

\pm

SE over three non-timeout problems; crosses: first 1000-second timeout). Only exact S 4 -RCA reaches

K=1000

. Panel (b) reports relative

\ell_{2}

error; DW-Fast’s diagnosed

K=20

failure is omitted (DoWhy 0.14; Supplement Sec. D.3).

Figure 2: Scarce-incident recovery over 30 ordered

d=100

SCM pairs with balanced function/law oracle mass (mean

\pm

SE). Panels report

F

and

Q

Recall@5 and

F+Q

Recall@10; node rankings are lifted for

F+Q

. Beyond the smallest regime, S 4 -RCA has the highest mean, most clearly for

F

(broader benchmark: Supplement Sec. D.4).

Figure 3: Causal Chamber all-channel score. (a) Target rank (1 best; worst of five splits; m / s : severity). (b) Target-

Q

and downstream- ir_3 mass;

F

denotes the fitted linear mechanism.

Appendix figures & tables7 assets

Supplementary material from the paper’s appendix.

Appendix

Framework	$M^{(0)}$ : source	$M^{(1)}$ : target	Modules $\mathcal{U}$	Scalar $T$	Report in words
S 4 -RCA (ours)	$M(F^{(0)},Q^{(0)})$	$M(F^{(1)},Q^{(1)})$	$\mathcal{U}_{\mathtt{F}\cup\mathtt{Q}}$ : separate $\mathtt{F}_{j},\mathtt{Q}_{j}$	$T(M)=\mathbb{E}_{M}[\tau(X)]$	Allocates a population alarm-score change to structural functions and disturbance laws separately.
Population CCA/MRCCA ( Budhathoki et al. 2021 ; Quintas-Martinez et al. 2024 )	$M(F^{(0)},Q^{(0)})$	$M(F^{(1)},Q^{(1)})$	$\mathcal{U}_{(\mathtt{F},\mathtt{Q})}$ : joint node switches	$T(M)=\psi(P_{Y}^{M})$	Switches each node conditional jointly. MRCCA estimates $T(M)=\mathbb{E}_{M}[h(Y)]$ multiply robustly; with $h(Y)=\tau(X)$ , the game is paired-node S 4 -RCA.
Unit-level change attribution ( Budhathoki et al. 2022a )	$M(F_{Y}^{(0)},\delta_{x^{(0)}})$	$M(F_{Y}^{(1)},\delta_{x^{(1)}})$	$\{\mathtt{F}_{Y},\mathtt{X}_{1},\ldots,\mathtt{X}_{d}\}$ : function/value switches	$T(M)=\mathbb{E}_{M}[Y]=F_{Y}(x)$	Explains one unit’s output change by switching the prediction function and observed input coordinates (fine game), or all inputs jointly (coarse game).
Fixed-SCM outlier RCA ( Budhathoki et al. 2022b )	$M(F^{(0)},Q^{(0)})$	$M(F^{(0)},\delta_{e^{(0)}(x)})$	$\mathcal{U}_{\mathtt{Q}}$ : factual-noise fixes	$T_{x}(M)=\log P_{M}(A_{x})$	Fixes reference noises to their factual values and allocates the information-theoretic score $S^{\tau}_{M^{(0)}}(x)$ .
FANS ( Seo et al. 2026 )	$M(F^{(0)},Q^{(0)})$	$M(F^{(1)},Q^{(1)})$	—; returns $\mathtt{F}_{j}/\mathtt{Q}_{j}$ labels	—	From samples in two environments, detects changed nodes and labels function or noise shifts; it does not allocate a score contrast.
Exact do-Shapley ( Witter et al. 2026 )	$M^{(0)}=M$	$M^{(1)}=M_{\operatorname{do}(X=x)}$	$\{\mathtt{X}_{j}:j\in[d]\}$ : variable interventions	$T(M)=\mathbb{E}_{M}[Y]$	Allocates an interventional outcome to variable interventions, not function/law shifts.

Appendix

Table 1: Closest frameworks translated into this paper’s notation. CCA/MRCCA neither identifies nor separately switches the function/law representation of an induced node conditional. For outlier RCA,

A_{x}=\{\tau(X)\geq\tau(x)\}

and

S^{\tau}_{M^{(0)}}(x)=-\log P_{M^{(0)}}(A_{x})

. Unit-level attribution observes paired input vectors rather than estimating input laws. Its deterministic one-output SCM embedding uses

\delta_{x^{(r)}}=\bigotimes_{j=1}^{d}\delta_{x_{j}^{(r)}}

, so

\mathtt{X}_{j}

switches an observed value, not a population disturbance law. FANS has no attribution modules or scalar game.

Metric	Function/law-resolved	Paired node
Target-Q selected (%)	100.0	0.0
Selected score reduction	1.934 $\pm$ 0.029	0.838 $\pm$ 0.016
Post-repair score shift	-0.543 $\pm$ 0.018	0.552 $\pm$ 0.011
Alarm cleared (%)	100.0	0.0

Appendix

Table 2: Q-actionable repair localization over 100 independent source problems (mean

\pm

standard error). Score reduction is

T(M^{(1)})-T(M^{(1),a})

; post-repair shift is

T(M^{(1),a})-T(M^{(0)})

, so a nonpositive value clears the alarm.

Figure 1: Broad comparison on the same 30 scarce-incident SCMs as Main Figure 2 (mean

\pm

SE). The main figure emphasizes transfer, resolution, and propagation ablations; this figure adds fixed-SCM event attribution, CCA/MRCCA, DCI, iSCAN, and feature-shift SHAP. Single-ranking outputs are copied to both

F

and

Q

labels only for a common localization diagnostic and do not thereby identify change type. Fixed-

F

is the expected-score event mean, not the native information-theoretic tail game. DCI has no

n_{1}=100

result because every such cell reached the 300-second cap.

$n_{1}$	$F$ R@5	$Q$ R@5	$F{+}Q$ R@10
20	0.21 (0.03)	0.18 (0.03)	0.20 (0.02)
40	0.37 (0.03)	0.33 (0.03)	0.30 (0.02)
60	0.46 (0.04)	0.39 (0.05)	0.38 (0.03)
80	0.59 (0.02)	0.45 (0.04)	0.43 (0.02)
100	0.69 (0.03)	0.52 (0.04)	0.47 (0.02)

Appendix

Table 3: Calibrated

d=100

cross-fitted diagnostics over 30 SCMs; entries are mean (SE).

F

and

Q

R@5 are family-specific changed-support recalls, and

F+Q

R@10 ranks the 200 typed players jointly. Relative

\ell_{1}

errors compare the family attribution vectors with their exact oracles. Endpoint errors first average the five folds within each SCM.

Reference-only order estimation
$n_{0}$	Oracle $\ell_{1}$	E2E $\ell_{1}$	E2E R@2
100	0.14 (0.09)	2.11 (0.19)	0.40 (0.12)
200	0.11 (0.11)	1.63 (0.22)	0.55 (0.09)
300	0.16 (0.16)	1.67 (0.20)	0.60 (0.07)
400	0.00 (0.00)	1.12 (0.15)	0.75 (0.08)

Appendix

Table 4: Core-assumption diagnostics over ten

d=5

SCMs (mean (SE)). The three blocks vary reference size, nonlinear-score quartic ratio, and nonlinear-mechanism strength. Oracle

\ell_{1}

isolates population distortion from the learned order or linear approximation; E2E

\ell_{1}

additionally includes endpoint estimation and cross-fitting. R@

k

is changed-player recall.

Polarizer coefficients
	$\to$ ir_3		$\to$ vis_3
Regime	P1	P2	P1	P2
Baseline	$-0.05$	$+0.05$	$-0.05$	$+0.04$
P1-mid	$-0.38$	$+0.35$	$-0.36$	$+0.34$
P1-strong	$-0.10$	$+0.10$	$-0.08$	$+0.09$
P2-mid	$+0.37$	$-0.38$	$+0.35$	$-0.36$

Appendix

Table 5: Reference-standardized polarization projections on the fixed official subgraph. P1/P2 are pol_1 / pol_2 ; coefficients and

R^{2}

use all rows and are descriptive rather than held-out evidence.

(a) Wind Tunnel
Method	H@1	H@3	Med.
S 4 -RCA (function/law)	0.808	1.000	1.0
Fixed- $F$ control	0.769	1.000	1.0
Delta shift	1.000	1.000	1.0
Observed PCA	0.731	0.923	1.0
Residual local KL	1.000	1.000	1.0

Appendix

Table 6: All-channel external interventions. H@

k

requires a target to rank within the top

k

in all five splits; Med. is the median worst rank across 26 Wind or seven Sachs conditions. Primary rows use official-edge support, and order-only S 4 -RCA is a sensitivity. Bold marks the within-dataset best, including ties.

Causal Attribution

Causal attribution identifies which features, actions, or mechanisms make a difference to an outcome using causal models and interventions.

See all 52 papers

Sep 3, 2026

A Computationally Feasible Framework for Causal Probabilistic Explanation

Explaining why a specific outcome occurred, and which inputs deserve the blame or credit, is central to philosophical, scientific, and policy analysis. Existing tools split into two camps. The theory of actual causality (AC) gives principled verdicts, but only for toy-sized models, because computing them requires enumerating counterfactual scenarios. Scalable attribution methods like SHAP (or even causal SHAP) at least partially ignore the causal structure that generated the data, and can give answers that conflict with a careful causal analysis. We close this gap with Probabilistic Causal Impact (PCI). PCI builds on actual causality and on Pearl's notions of probability of necessity and sufficiency, but recasts the question of explainability as an estimation problem on a probabilistic causal model that is easily approximated via Monte Carlo. By specifying a distribution over "candidate explanations," a distribution over counterfactual values, and a scoring function, PCI provides tractable, causally grounded, graded explanations, generalizing AC and Pearl's probability of causation as degenerate cases. We evaluate PCI in synthetic and real-world examples, spanning consistency checks with AC, scaling experiments, complex continuous-valued dynamical systems, and a real-world deployed causal machine learning model trained on millions of datapoints.

Rafal Urbaniak, Sam Witty, Daniel Waxman +7

Basis Research Institute · Sorbus AI · Massachusetts Institute of Technology +3

Structural Causal Models Actual Causality Causal Attribution Causal Inference Causal Explanation

Ask PDF

Jun 23, 2026

Beyond Shapley: Efficient Computation of Asymmetric Shapley Values

We address the problem of explainability in machine learning models through feature attribution methods. In particular, we consider a variant of Shapley values known as Asymmetric Shapley Values (ASV), which enables the incorporation of causal knowledge into model-agnostic explanations through the use of a causal graph. We show that in certain contexts in which the computation of SHAP is

\#P

-hard, the exact computation of ASV can be done in polynomial time. To extend this algorithmic result, we introduce a notion of equivalence classes over the topological orderings of the underlying causal graph, which is useful to reduce the time to compute ASV. In particular, we present a polynomial-time algorithm (in the number of equivalence classes) to compute it whenever the causal graph is a rooted directed tree. Finally, we develop an algorithm for approximating ASV in arbitrary causal DAGs which relies on a procedure to sample topological orderings uniformly at random. To implement this sampling mechanism we leverage known algorithms as well as simpler alternatives. Our experimental results demonstrate the practical viability of the proposed approach in realistic causal structures.

Ezequiel Companeetz, Santiago Cifuentes, Sergio Abriola

Departamento de Computación, Facultad de Ciencias Exactas y Naturales, UBA · Instituto de Ciencias de la Computación (ICC), CONICET

Cooperative Game Theory Feature Attribution Causal Attribution Shapley Value Attribution Causal Interpretability

Ask PDF

Aug 12, 2026

CAS: A Causal Attribution Score for Local and Global Explainable Artificial Intelligence

Predictive explanation methods attribute a model output; they do not, by themselves, attribute an intervention effect on the real-world outcome. We introduce the Causal Attribution Score (CAS), a compact score architecture for causal explanation. CAS starts from an identified interventional coalition game, allocates the joint intervention contrast with causal Shapley contributions, and converts those raw outcome-scale effects into Local CAS, Signed Local CAS, and two complementary Global CAS summaries. The innovation is not a new Shapley formula, but a local-to-global causal reporting layer with an explicit intervention target. In the known-truth benchmark, eight repeated primary-interaction simulations (n = 2,200 each, three actions) gave mean Local CAS MAE of 0.107 for coalition-aware CAS, compared with 0.173 for one-at-a-time normalisation and 0.213 for a global normalised absolute ATE vector. The paired advantage over one-at-a-time normalisation increased from -0.003 under additivity to 0.091 under strong interactions. On both empirical DoubleML datasets, 401(k) eligibility/net financial assets (n = 9,915) and Pennsylvania reemployment bonus/unemployment duration (n = 5,099), predictive SHAP/TreeSHAP rankings differed materially from Feature-CAS rankings of treatment-effect modifiers. In Pennsylvania, dep1 (exactly one dependent) moved from predictive global rank 13 to Feature-CAS rank 2 and was the leading local Feature-CAS modifier. These results isolate the added value of separating what predicts the outcome from what explains heterogeneity in an estimated causal effect.

Michael Georgiades, Charalambia Varnava

Department of Computer Science Neapolis University Pafos Cyprus · CaSToRC The Cyprus Institute Nicosia, Cyprus

Feature Attribution Causal Attribution Causal Explanation Causal Interpretability Explainability Evaluation

Ask PDF

Abstract

Causal Attribution

Explore similar work

A Computationally Feasible Framework for Causal Probabilistic Explanation

Beyond Shapley: Efficient Computation of Asymmetric Shapley Values

CAS: A Causal Attribution Score for Local and Global Explainable Artificial Intelligence