pith. sign in

arxiv: 2605.02603 · v1 · submitted 2026-05-04 · 💻 cs.AI

Counterfactual Reasoning in Automated Planning

Alberto Pozanco , Daniel Borrajo , Manuela Veloso This is my paper

Pith reviewed 2026-05-09 15:48 UTC · model grok-4.3

classification 💻 cs.AI
keywords counterfactual reasoningautomated planningsurveytaxonomydynamic planningflexible planningAI planning
0
0 comments X

The pith

Existing works on counterfactual reasoning in automated planning can be categorized by the elements changed, the timing of reasoning, and the reasons and methods for changes.

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

Automated planning has traditionally assumed fixed initial states, goals, and actions, which suits deterministic domains but struggles with real-world deviations and unforeseen events. This survey reviews how counterfactual reasoning enables planners to consider alternative scenarios by modifying those fixed aspects. The authors group the literature along dimensions of what elements are altered, when the reasoning is triggered, and the purposes and techniques for making changes. The work identifies patterns across existing approaches and raises open questions for advancing the field. Readers would care because this organization clarifies how planning systems might gain flexibility to respond to unexpected circumstances and improve outcomes.

Core claim

This paper surveys existing works on counterfactual reasoning in automated planning, categorizing them by what elements are changed, when the reasoning is triggered, and why and how these changes are made. We conclude by discussing key findings and outlining open research questions to guide future work in this area.

What carries the argument

A categorization framework using dimensions for changed elements (initial state, goals, or actions), timing of the reasoning trigger, and the reasons plus methods for the changes.

If this is right

  • Planning systems can be extended to handle dynamic deviations by applying counterfactual reasoning at the right moments for specific purposes.
  • Research can prioritize less explored combinations of the categorization dimensions to fill gaps in current methods.
  • More adaptable planners could produce better plans in domains where initial assumptions often fail or where improvements are possible through alternative scenarios.

Where Pith is reading between the lines

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

  • The taxonomy could serve as a starting point for developing new algorithms that systematically combine different types of changes and triggers.
  • Links to related concepts in other areas of AI, such as learning from hypothetical outcomes, might emerge if the dimensions are applied across fields.
  • Applying the same categorization to papers published after this survey would test whether the structure remains useful or needs updating.

Load-bearing premise

The body of existing literature can be meaningfully organized by the proposed dimensions of changed elements, timing, reasons, and methods without significant omissions or alternative taxonomies that would better capture the field's structure.

What would settle it

The discovery of many relevant papers that cannot be placed into the proposed categories or the proposal of a substantially different taxonomy that reorganizes the literature more effectively would show the categorization is incomplete or insufficient.

Figures

Figures reproduced from arXiv: 2605.02603 by Alberto Pozanco, Daniel Borrajo, Manuela Veloso.

Figure 1
Figure 1. Figure 1: LOGISTICS task where a truck must deliver two packages by moving them from their current (filled) locations to their goal (empty) destinations. of actions so that no plan exists to solve the planning task, thereby completely limiting agent behavior. We identify two interconnected gaps in the literature. First, existing research primarily focuses on either com￾pletely removing actions or modifying the preco… view at source ↗
read the original abstract

Automated planning traditionally assumes that all aspects of a planning task (initial state, goals, and available actions) are fully specified in advance, an approach well-suited to domains with fixed rules and deterministic execution. However, real-world planning often requires flexibility, allowing for deviations from the original task parameters in response to unforeseen circumstances or to improve outcomes. This paper surveys existing works on counterfactual reasoning in automated planning, categorizing them by what elements are changed, when the reasoning is triggered, and why and how these changes are made. We conclude by discussing key findings and outlining open research questions to guide future work in this area.

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

0 major / 2 minor

Summary. The paper surveys existing works on counterfactual reasoning in automated planning. It organizes the literature into a taxonomy based on four dimensions: the elements changed (e.g., initial state, goals, actions), the timing of the reasoning, the reasons for the changes, and the methods employed. The survey concludes with a discussion of key findings and open research questions to guide future work.

Significance. If the categorization is accurate and reasonably comprehensive, the paper offers a useful organizing framework for an emerging sub-area of AI planning that extends beyond traditional deterministic assumptions. This synthesis can help researchers locate relevant prior work and spot gaps, particularly in how counterfactuals support flexibility in real-world planning tasks. The contribution is primarily taxonomic rather than technical, so its value rests on the clarity and coverage of the dimensions rather than on new theorems or experiments.

minor comments (2)
  1. The abstract and introduction should explicitly name the four dimensions (changed elements, timing, reasons, methods) rather than describing them indirectly, to improve immediate readability for readers scanning the paper.
  2. A brief subsection or paragraph on the paper-selection methodology (search terms, databases, inclusion criteria) would help readers evaluate the scope of the survey and reduce concerns about potential omissions.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for their positive evaluation of our survey on counterfactual reasoning in automated planning. We appreciate the recognition that the taxonomic framework can help organize an emerging area and guide future work. The recommendation for minor revision is noted, but no specific major comments were provided in the report.

Circularity Check

0 steps flagged

No significant circularity

full rationale

This is a survey paper that reviews and taxonomizes external published works on counterfactual reasoning in automated planning. It contains no derivations, equations, fitted parameters, or self-referential claims. The central contribution is a descriptive categorization along dimensions (changed elements, timing, reasons, methods) drawn from the cited literature, with no reduction of any claim to inputs defined within the paper itself. The organization is self-contained as an act of literature synthesis and does not rely on any load-bearing self-citation chain or definitional loop.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

As a survey paper the work introduces no free parameters, axioms, or invented entities; it depends entirely on the prior literature it cites for its content and structure.

pith-pipeline@v0.9.0 · 5388 in / 1122 out tokens · 43682 ms · 2026-05-09T15:48:03.427350+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

44 extracted references · 44 canonical work pages · 1 internal anchor

  1. [1]

    Proceedings of the international conference on automated planning and scheduling , volume=

    Coming up with good excuses: What to do when no plan can be found , author=. Proceedings of the international conference on automated planning and scheduling , volume=

    work page

  2. [2]

    International Joint Conference on Artificial Intelligence , year=

    Why can't you do that hal? explaining unsolvability of planning tasks , author=. International Joint Conference on Artificial Intelligence , year=

    work page

  3. [3]

    Proceedings of the International Conference on Automated Planning and Scheduling , volume=

    Computing opportunities to augment plans for novel replanning during execution , author=. Proceedings of the International Conference on Automated Planning and Scheduling , volume=

    work page

  4. [4]

    2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , pages=

    Intelligent Execution through Plan Analysis , author=. 2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , pages=. 2021 , organization=

    work page 2021

  5. [5]

    ECAI 2023 , pages=

    Generating Replanning Goals Through Multi-Objective Optimization in Response to Execution Observation , author=. ECAI 2023 , pages=. 2023 , publisher=

    work page 2023

  6. [6]

    Proceedings of the International Conference on Automated Planning and Scheduling , volume=

    Finding Centroids and Minimum Covering States in Planning , author=. Proceedings of the International Conference on Automated Planning and Scheduling , volume=

    work page

  7. [7]

    Proceedings of the International Conference on Automated Planning and Scheduling , volume=

    A compilation based approach to finding centroids and minimum covering states in planning , author=. Proceedings of the International Conference on Automated Planning and Scheduling , volume=

    work page

  8. [8]

    Proceedings of the International Conference on Automated Planning and Scheduling , volume=

    Computing planning centroids and minimum covering states using symbolic bidirectional search , author=. Proceedings of the International Conference on Automated Planning and Scheduling , volume=

    work page

  9. [9]

    Proceedings of the International Conference on Automated Planning and Scheduling , volume=

    Anticipatory on-line planning , author=. Proceedings of the International Conference on Automated Planning and Scheduling , volume=

    work page

  10. [10]

    arXiv preprint arXiv:1906.12249 , year=

    Anticipatory thinking: A metacognitive capability , author=. arXiv preprint arXiv:1906.12249 , year=

    work page arXiv 1906

  11. [11]

    AI Communications , volume=

    Anticipation of goals in automated planning , author=. AI Communications , volume=. 2018 , publisher=

    work page 2018

  12. [12]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Choosing the Initial State for Online Replanning , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    work page

  13. [13]

    AI Communications , volume=

    Learning-driven goal generation , author=. AI Communications , volume=. 2018 , publisher=

    work page 2018

  14. [14]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Goal-driven autonomy in a Navy strategy simulation , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    work page

  15. [15]

    IJCAI Proceedings-International Joint Conference on Artificial Intelligence , volume=

    Integrated learning for goal-driven autonomy , author=. IJCAI Proceedings-International Joint Conference on Artificial Intelligence , volume=

    work page

  16. [16]

    Proceedings of the International Conference on Automated Planning and Scheduling , volume=

    Planning with object creation , author=. Proceedings of the International Conference on Automated Planning and Scheduling , volume=

    work page

  17. [17]

    Special case of creation planning , author=

    Compiling irrelevant objects to counters. Special case of creation planning , author=. AI Communications , volume=. 2016 , publisher=

    work page 2016

  18. [18]

    Artificial Intelligence , volume=

    Learning action models with minimal observability , author=. Artificial Intelligence , volume=. 2019 , publisher=

    work page 2019

  19. [19]

    Proceedings of the International Conference on Automated Planning and Scheduling , volume=

    A planning approach to repair domains with incomplete action effects , author=. Proceedings of the International Conference on Automated Planning and Scheduling , volume=

    work page

  20. [20]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Generalising planning environment redesign , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    work page

  21. [21]

    Proceedings of the Twenty-Ninth International Conference on International Joint Conferences on Artificial Intelligence , pages=

    Goal recognition design-survey , author=. Proceedings of the Twenty-Ninth International Conference on International Joint Conferences on Artificial Intelligence , pages=

    work page

  22. [22]

    34th International Joint Conference on Artificial Intelligence , year=

    A Survey on Model Repair in AI Planning , author=. 34th International Joint Conference on Artificial Intelligence , year=

    work page

  23. [23]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Told You That Will Not Work: Optimal Corrections to Planning Domains Using Counter-Example Plans , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    work page

  24. [24]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Towards automated modeling assistance: An efficient approach for repairing flawed planning domains , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    work page

  25. [25]

    Proceedings of the International Conference on Automated Planning and Scheduling , volume=

    Multi-tier automated planning for adaptive behavior , author=. Proceedings of the International Conference on Automated Planning and Scheduling , volume=

    work page

  26. [26]

    Arizona State University CSE Department TR , year=

    Replanning as a Deliberative Re-selection of Objectives , author=. Arizona State University CSE Department TR , year=

    work page

  27. [27]

    Proceedings of the International Conference on Automated Planning and Scheduling , volume=

    A theoretical and algorithmic analysis of configurable MDPs , author=. Proceedings of the International Conference on Automated Planning and Scheduling , volume=

    work page

  28. [28]

    , author=

    Choosing Objectives in Over-Subscription Planning. , author=. ICAPS , volume=

    work page

  29. [29]

    Nau and Paolo Traverso , title =

    Malik Ghallab and Dana S. Nau and Paolo Traverso , title =. 2004 , isbn =

    work page 2004

  30. [30]

    Proceedings of the International Conference on Automated Planning and Scheduling , volume=

    Explaining plan quality differences , author=. Proceedings of the International Conference on Automated Planning and Scheduling , volume=

    work page

  31. [31]

    Proceedings of the aaai conference on artificial intelligence , volume=

    Classical planning in deep latent space: Bridging the subsymbolic-symbolic boundary , author=. Proceedings of the aaai conference on artificial intelligence , volume=

    work page

  32. [32]

    ECAI 2020 , pages=

    Learning First-Order Symbolic Representations for Planning from the Structure of the State Space , author=. ECAI 2020 , pages=. 2020 , publisher=

    work page 2020

  33. [33]

    The Knowledge Engineering Review , volume=

    A review of learning planning action models , author=. The Knowledge Engineering Review , volume=. 2018 , publisher=

    work page 2018

  34. [34]

    Artificial Intelligence , volume=

    Foundations of explanations as model reconciliation , author=. Artificial Intelligence , volume=. 2021 , publisher=

    work page 2021

  35. [35]

    arXiv preprint arXiv:2408.10889 , year=

    On Learning Action Costs from Input Plans , author=. arXiv preprint arXiv:2408.10889 , year=

    work page arXiv

  36. [36]

    International Conference on Machine Learning , pages=

    Configurable Markov decision processes , author=. International Conference on Machine Learning , pages=. 2018 , organization=

    work page 2018

  37. [37]

    2009 , publisher=

    A general approach to environment design with one agent , author=. 2009 , publisher=

    work page 2009

  38. [38]

    Cognitive science , volume=

    Structural counterfactuals: A brief introduction , author=. Cognitive science , volume=. 2013 , publisher=

    work page 2013

  39. [39]

    2019 , publisher=

    An introduction to the planning domain definition language , author=. 2019 , publisher=

    work page 2019

  40. [40]

    arXiv preprint arXiv:1701.08317 , year=

    Plan explanations as model reconciliation: Moving beyond explanation as soliloquy , author=. arXiv preprint arXiv:1701.08317 , year=

    work page arXiv

  41. [41]

    , author=

    Counterplanning using Goal Recognition and Landmarks. , author=. IJCAI , pages=

    work page

  42. [42]

    Proceedings of the International Conference on Automated Planning and Scheduling , volume=

    On the gains from using action observations in domain repair , author=. Proceedings of the International Conference on Automated Planning and Scheduling , volume=

    work page

  43. [43]

    European Conference on Artificial Intelligence (ECAI) , title =

    Morales, Marianela and Pozanco, Alberto and Canonaco, Giuseppe and Gopalakrishnan, Sriram and Borrajo, Daniel and Veloso, Manuela , year =. European Conference on Artificial Intelligence (ECAI) , title =

    work page

  44. [44]

    Planning Task Shielding: Detecting and Repairing Flaws in Planning Tasks through Turning them Unsolvable

    Planning Task Shielding: Detecting and Repairing Flaws in Planning Tasks through Turning them Unsolvable , author=. arXiv preprint arXiv:2604.07042 , year=

    work page internal anchor Pith review Pith/arXiv arXiv