pith. sign in

arxiv: 2606.31080 · v1 · pith:PHBE43K6new · submitted 2026-06-30 · 💻 cs.LO · cs.AI

Beyond But-for Test: Counterfactual Explanation in Abstract Argumentation via Actual Causality (Extended Version)

Siyi Liu , Muyun Shao , Beishui Liao This is my paper

Pith reviewed 2026-07-01 03:47 UTC · model grok-4.3

classification 💻 cs.LO cs.AI
keywords counterfactual explanationabstract argumentationactual causalityHalpern-Pearl definitionbut-for testintervention operatorpreemptionoverdetermination
0
0 comments X

The pith

Encoding argument acceptance as equations and applying an intervention operator from actual causality identifies causes beyond the but-for test in abstract argumentation.

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

The paper introduces a framework for counterfactual explanations in abstract argumentation by asking what-if questions about changes to argument status. Prior methods rely only on the but-for test, which cannot handle complex cases such as one argument preempting another or multiple arguments overdetermining acceptance. The new approach encodes acceptance conditions as equations, then defines an intervention operator that can alter sets of arguments at once while fixing witness arguments to their observed labels. This follows the refined counterfactual condition from the Halpern-Pearl definition of actual causality, allowing correct cause identification in preemption and overdetermination structures. The authors show through comparison that the method exceeds earlier approaches in expressiveness and reliability.

Core claim

By representing the acceptance conditions of arguments as a system of equations and introducing an intervention operator that supports simultaneous changes to argument sets together with fixing of witness arguments, the framework applies the Halpern-Pearl refined counterfactual condition to abstract argumentation and thereby identifies actual causes correctly in preemption and overdetermination cases where the but-for test fails.

What carries the argument

The intervention operator defined on the equation encoding of the argumentation framework, which permits changing multiple arguments and fixing witnesses.

If this is right

  • Causes can be identified correctly in preemption structures where one argument blocks another.
  • Overdetermination cases with multiple sufficient causes are resolved by the refined counterfactual condition.
  • The method supports changing sets of arguments simultaneously rather than single arguments.
  • Comparison establishes greater expressiveness and reliability than prior but-for-only approaches.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same equation-and-intervention approach could be tested on concrete argumentation frameworks drawn from legal or debate scenarios to check whether the identified causes match human judgments.
  • If the encoding step generalizes, the framework might extend to other structured reasoning systems that already use causal or intervention-based models.
  • A direct comparison on a shared benchmark set of preemption and overdetermination examples would make the reliability claim testable by independent replication.

Load-bearing premise

Acceptance conditions of arguments can be encoded as equations and the Halpern-Pearl definition of actual causality applies directly without modification to abstract argumentation frameworks.

What would settle it

A concrete preemption or overdetermination argumentation framework where applying the intervention operator and Halpern-Pearl condition yields a cause identification that conflicts with the intuitive or accepted causal judgment for that structure.

Figures

Figures reproduced from arXiv: 2606.31080 by Beishui Liao, Muyun Shao, Siyi Liu.

Figure 1
Figure 1. Figure 1: Counterexample to the but-for test in AA. The status of each argument is visually encoded as: in=dark gray (accepted), out=cross-hatched (rejected), und=light gray (undecided). The dashed box encloses the newly introduced argument and the attack relation. These visual conventions are used throughout the paper. Example 1. Consider the AF F1 in Fig. 1a, where η is the accepted argument to be explained. α and… view at source ↗
Figure 2
Figure 2. Figure 2: F2 with Even-Cycle fα = (in),out , [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: For Ex. 5 (F3) Example 5. Consider the AF F3 and its co-labelling L in [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: F4: Overdetermination Neither in(α) nor in(β) passes the but-for test BFT∃. By contrast, the compound hypothesis ψ = in(α) ∧ in(β) satisfies AC2m ∃ and there￾fore belongs to Counter(F4,η), providing a non￾trivial explanation with Aψ \ {η} ̸= /0. 5. Graph Mutilations of Intervention Following Pearl’s graph mutilations for interventions in causal networks [17, p. 23], we define a corresponding graph operatio… view at source ↗
Figure 5
Figure 5. Figure 5: Atomic Mutilation of F2 in [PITH_FULL_IMAGE:figures/full_fig_p010_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: For Ex. 8 (F5) Example 8. Consider the AF F5 and the co-labelling L in [PITH_FULL_IMAGE:figures/full_fig_p012_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: F6: V-Shape Consider the AF F6 in [PITH_FULL_IMAGE:figures/full_fig_p021_7.png] view at source ↗
read the original abstract

Counterfactual explanation in abstract argumentation calls for an answer to the what-if query: would the topic argument still be accepted if the status of certain other arguments were changed? Existing approaches are limited to the but-for test and fail to accommodate more refined counterfactual conditions. To overcome these limitations, we introduce an intervention-based counterfactual reasoning framework in abstract argumentation. Our approach encodes the acceptance conditions of arguments as equations, then defines an intervention operator that supports (1) changing sets of arguments simultaneously, and (2) fixing witness arguments to their actual labels. Guided by the refined counterfactual condition introduced in the Halpern-Pearl definition, our method goes beyond the but-for test, thereby correctly identifying causes in argumentation structures such as Preemption and Overdetermination. Through comparison, we show that our method surpasses prior methods in both expressiveness and reliability.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 1 minor

Summary. The manuscript introduces an intervention-based counterfactual reasoning framework for abstract argumentation. It encodes argument acceptance conditions as equations, defines an intervention operator supporting simultaneous changes and witness fixing, and applies the Halpern-Pearl actual causality definition to identify causes beyond the but-for test, correctly handling preemption and overdetermination, and demonstrating superiority over prior methods.

Significance. If the encoding faithfully preserves Dung semantics, this approach would offer a more expressive and reliable method for counterfactual explanations in argumentation frameworks, addressing limitations of but-for tests in complex scenarios. The explicit use of refined counterfactual conditions from actual causality theory is a notable strength, potentially advancing explainability in non-monotonic reasoning systems.

major comments (1)
  1. [Abstract (paragraph on encoding and intervention operator)] Abstract (paragraph on encoding and intervention operator): The central claim requires that AF acceptance conditions can be encoded as equations to which the Halpern-Pearl definition applies directly. However, Dung semantics are defined via global extension properties (conflict-freeness, admissibility, completeness) rather than per-argument functional equations; the manuscript must demonstrate that the encoding preserves acceptance under simultaneous interventions, as cycles and mutual attacks may otherwise produce non-equivalent results and undermine the handling of preemption/overdetermination.
minor comments (1)
  1. The abstract states that the method 'surpasses prior methods' via comparison but provides no details on the specific prior methods, evaluation metrics, or example frameworks used; this should be expanded in the introduction or a dedicated comparison section.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive comments and the positive assessment of the framework's potential. We address the single major comment below, focusing on the encoding's fidelity to Dung semantics under interventions.

read point-by-point responses

  1. Referee: Abstract (paragraph on encoding and intervention operator): The central claim requires that AF acceptance conditions can be encoded as equations to which the Halpern-Pearl definition applies directly. However, Dung semantics are defined via global extension properties (conflict-freeness, admissibility, completeness) rather than per-argument functional equations; the manuscript must demonstrate that the encoding preserves acceptance under simultaneous interventions, as cycles and mutual attacks may otherwise produce non-equivalent results and undermine the handling of preemption/overdetermination.

    Authors: We agree that Dung semantics are characterized by global extension properties rather than purely local functional equations. Our encoding defines per-argument equations that are derived directly from the chosen semantics' acceptance criteria (e.g., an argument is in if all its attackers are out, for admissible semantics), such that the solutions to the system of equations correspond exactly to the extensions under that semantics. The intervention operator is defined to modify the right-hand sides of the equations for the intervened arguments while leaving the structural equations intact, thereby preserving the global consistency conditions. We acknowledge, however, that the manuscript does not contain an explicit formal lemma or proof establishing equivalence of the intervened models to the original semantics under simultaneous changes, especially in cyclic frameworks. This is a valid point. We will add a dedicated subsection (or appendix) providing this demonstration for the semantics considered in the paper, including explicit verification on cyclic examples to confirm that preemption and overdetermination cases remain correctly handled. revision: yes

Circularity Check

0 steps flagged

No circularity: method applies external Halpern-Pearl definition to a novel encoding of AF acceptance conditions

full rationale

The derivation chain begins with an explicit encoding of argument acceptance conditions as equations (a modeling choice) followed by definition of an intervention operator, then directly invokes the external Halpern-Pearl actual-causality definition (with its refined counterfactuals) to identify causes beyond but-for. No step equates a claimed result to its own fitted parameters, renames a known pattern, or relies on a load-bearing self-citation whose content is unverified. The central claims about handling preemption and overdetermination are evaluated by direct comparison to the external semantics and prior methods, keeping the construction self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 1 invented entities

Framework rests on encoding acceptance conditions as equations and direct applicability of Halpern-Pearl causality; no free parameters or invented entities with independent evidence are described.

axioms (2)
  • domain assumption Argument acceptance conditions in abstract argumentation frameworks can be represented as equations
    Explicitly stated as the first step of the approach in the abstract.
  • domain assumption Halpern-Pearl definition of actual causality can be applied to define refined counterfactual conditions in argumentation
    Guiding principle for the intervention operator and beyond but-for test.
invented entities (1)
  • Intervention operator no independent evidence
    purpose: Supports simultaneous changes to sets of arguments and fixing witness arguments to actual labels
    New operator introduced to implement the framework; no independent evidence outside the paper.

pith-pipeline@v0.9.1-grok · 5677 in / 1291 out tokens · 41188 ms · 2026-07-01T03:47:52.787713+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

31 extracted references

  1. [1]

    Argumentative XAI: A Survey

    ˇCyras K, Rago A, Albini E, Baroni P, Toni F. Argumentative XAI: A Survey. In: Proceedings of the 13th International Joint Conference on Artificial Intelligence (IJCAI); 2021. p. 4392-9

    2021

  2. [2]

    Argumentation as a Method for Explainable AI : A Systematic Literature Review

    Engelmann D, Damasio J, Panisson AR, Mascardi V , Bordini RH. Argumentation as a Method for Explainable AI : A Systematic Literature Review. In: 2022 17th Iberian Conference on Information Systems and Technologies (CISTI); 2022. p. 1-6

    2022

  3. [3]

    Related Explanations in Formal Argumentation, an Empirical Study

    Scheffers R, Bex F, Borg A. Related Explanations in Formal Argumentation, an Empirical Study. In: Proceedings of COMMA; 2024. p. 265-76

    2024

  4. [4]

    On the Acceptability of Arguments and Its Fundamental Role in Nonmonotonic Reasoning, Logic Programming and n-Person Games

    Dung PM. On the Acceptability of Arguments and Its Fundamental Role in Nonmonotonic Reasoning, Logic Programming and n-Person Games. Artificial Intelligence. 1995;77(2):321-57

    1995

  5. [5]

    Post-hoc explanation of extension semantics

    Amgoud L. Post-hoc explanation of extension semantics. In: 27th European Conference on Artificial Intelligence (ECAI). vol. 392 of Frontiers in Artificial Intelligence and Applications; 2024. p. 3276-83

    2024

  6. [6]

    On Computing Explanations in Argumentation

    Fan X, Toni F. On Computing Explanations in Argumentation. In: Proceedings of the 29th AAAI Conference on Artificial Intelligence. vol. 29 of 15; 2015. p. 1496-502

    2015

  7. [7]

    Explanation Semantics for Abstract Argumentation

    Liao B, Van Der Torre L. Explanation Semantics for Abstract Argumentation. In: Proceedings of COMMA; 2020. p. 271-82

    2020

  8. [8]

    Strong Explanations in Abstract Argumentation

    Ulbricht M, Wallner JP. Strong Explanations in Abstract Argumentation. In: Proceedings of the 35th AAAI Conference on Artificial Intelligence. vol. 35 of 7; 2021. p. 6496-504

    2021

  9. [9]

    Minimality, Necessity and Sufficiency for Argumentation and Explanation

    Borg A, Bex F. Minimality, Necessity and Sufficiency for Argumentation and Explanation. International Journal of Approximate Reasoning. 2024;168:109143

    2024

  10. [10]

    A Basic Framework for Explanations in Argumentation

    Borg A, Bex F. A Basic Framework for Explanations in Argumentation. IEEE Intelligent Systems. 2021;36(2):25-35

    2021

  11. [11]

    Necessary and Sufficient Explanations for Argumentation-Based Conclusions

    Borg A, Bex F. Necessary and Sufficient Explanations for Argumentation-Based Conclusions. In: Symbolic and Quantitative Approaches to Reasoning with Uncertainty, ECSQARU 2021. vol. 12897. Cham: Springer; 2021. p. 45-58

    2021

  12. [12]

    Explanation in Artificial Intelligence: Insights from the Social Sciences

    Miller T. Explanation in Artificial Intelligence: Insights from the Social Sciences. Artificial Intelligence. 2019;267:1-38

    2019

  13. [13]

    Counterfactuals in explainable artificial intelligence (XAI): Evidence from human reason- ing

    Byrne RMJ. Counterfactuals in explainable artificial intelligence (XAI): Evidence from human reason- ing. In: Proceedings of the 28th International Joint Conference on Artificial Intelligence (IJCAI); 2019. p. 6276-82

    2019

  14. [14]

    Counterfactual Reasoning in Argumentation Frameworks

    Sakama C. Counterfactual Reasoning in Argumentation Frameworks. In: Proceedings of COMMA

  15. [15]

    Argumentation In Flux (Modelling Change in the Theory of Argumentation) [Ph.D

    Rienstra T. Argumentation In Flux (Modelling Change in the Theory of Argumentation) [Ph.D. thesis]. Universit´e Montpellier II-Sciences et Techniques du Languedoc; Universit´e du Luxembourg; 2014

    2014

  16. [16]

    Abduction in Argumentation Frameworks

    Sakama C. Abduction in Argumentation Frameworks. Journal of Applied Non-Classical Logics. 2018;28(2-3):218-39

    2018

  17. [17]

    Causality: Models, Reasoning, and Inference

    Pearl J. Causality: Models, Reasoning, and Inference. Cambridge University Press; 2009

    2009

  18. [18]

    Causes and Explanations: A Structural-Model Approach

    Halpern JY , Pearl J. Causes and Explanations: A Structural-Model Approach. Part II: Explanations. The British Journal for the Philosophy of Science. 2005;56(4):889-911

    2005

  19. [19]

    A modification of the Halpern-Pearl definition of causality

    Halpern JY . A modification of the Halpern-Pearl definition of causality. In: Proceedings of the 24th International Joint Conference on Artificial Intelligence (IJCAI); 2015. p. 3022-33

    2015

  20. [20]

    Actual Causality

    Halpern JY . Actual Causality. MIT Press; 2016

    2016

  21. [21]

    Abstract Argumentation Frameworks and Their Semantics

    Baroni P, Caminada M, Giacomin M. Abstract Argumentation Frameworks and Their Semantics. In: Handbook of Formal Argumentation. 1st ed. London, England: College Publications; 2018. p. 159-236

    2018

  22. [22]

    A Logical Account of Formal Argumentation

    Caminada M, Gabbay DM. A Logical Account of Formal Argumentation. Studia Logica. 2009;93(2- 3):109-45

    2009

  23. [23]

    Attack-Defense Semantics of Argumentation

    Liao B, Van Der Torre L. Attack-Defense Semantics of Argumentation. In: Proceedings of COMMA

  24. [24]

    Counterfactual and Semifactual Explanations in Abstract Argumentation: Formal Foundations, Complexity and Computation

    Alfano G, Greco S, Parisi F, Trubitsyna I. Counterfactual and Semifactual Explanations in Abstract Argumentation: Formal Foundations, Complexity and Computation. In: Proceedings of the 21st Inter- national Conference on Principles of Knowledge Representation and Reasoning (KR); 2024. p. 14-26

    2024

  25. [25]

    Counterfactuals

    Lewis D. Counterfactuals. Blackwell Publishing; 1973

    1973

  26. [26]

    A principled approach to defining actual causation

    Beckers S, Vennekens J. A principled approach to defining actual causation. Synthese. 2018;195(2):835- 62

    2018

  27. [27]

    Explainable Acceptance in Probabilistic Abstract Argumentation: Complexity and Approximation

    Alfano G, Calautti M, Greco S, Parisi F, Trubitsyna I. Explainable Acceptance in Probabilistic Abstract Argumentation: Complexity and Approximation. In: Proceedings of the 17th International Conference on Principles of Knowledge Representation and Reasoning (KR); 2020. p. 33-43

    2020

  28. [28]

    Explaining Argument Acceptance in ADFs

    Rienstra T, Heyninck JLA, Kern-Isberner G, Skiba K, Thimm M. Explaining Argument Acceptance in ADFs. In: 1st International Workshop on Argumentation for eXplainable AI (ArgXAI, co-located with COMMA 22). vol. 3209 of CEUR Workshop Proceedings; 2022

    2022

  29. [29]

    Argumentation-based Causal and Counterfactual Reason- ing

    Bengel L, Bl ¨umel L, Rienstra T, Thimm M. Argumentation-based Causal and Counterfactual Reason- ing. In: 1st International Workshop on Argumentation for eXplainable AI (ArgXAI, co-located with COMMA 22). vol. 3209 of CEUR Workshop Proceedings; 2022

    2022

  30. [30]

    Modelling cause-in-fact in legal cases through defeasible argumentation

    Pisano G, Prakken H, Sartor G, Liepina R. Modelling cause-in-fact in legal cases through defeasible argumentation. In: Proceedings of the 20th International Conference on Artificial Intelligence and Law (ICAIL); 2025. p. 268-77

    2025

  31. [31]

    Causation and Argumentation

    Bochman A, Cerutti F, Rienstra T. Causation and Argumentation. Journal of Applied Logics. 2025;12(3):713-86. Appendix: Proofs Proof of Theorem 1. Proof.(⇒)Sol(M F)⊆L co(F):LetL∈Sol(M F)be a solution of the acceptability modelM F. Take an arbitrary argumentα∈A. SinceLsatisfies every equation inM F, we haveL(α) =f α (L(β) β∈α −) , wheref α is the acceptabil...

    2025