# Counterfactual Explanations as Plans

arXiv: https://arxiv.org/abs/2502.09205 · HTML: https://mlsanity.com/paper/2502.09205/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 9 of its 44 references that are in the map.
- First posted on arXiv 13 Feb 2025.

## Builds on (most cited first, 9 of 9)

- [Counterfactual Explanations without Opening the Black Box: Automated Decisions and the GDPR](https://mlsanity.com/paper/1711.00399/) (2017)
- [Explaining Machine Learning Classifiers through Diverse Counterfactual Explanations](https://mlsanity.com/paper/1905.07697/) (2019)
- [Explainable Planning](https://mlsanity.com/paper/1709.10256/) (2017)
- [Towards Explainable AI Planning as a Service](https://mlsanity.com/paper/1908.05059/) (2019)
- [Declarative Approaches to Counterfactual Explanations for Classification](https://mlsanity.com/paper/2011.07423/) (2020)
- [Challenges of Human-Aware AI Systems](https://mlsanity.com/paper/1910.07089/) (2019)
- [Towards Providing Explanations for AI Planner Decisions](https://mlsanity.com/paper/1810.06338/) (2018)
- [Explanation Generation for Multi-Modal Multi-Agent Path Finding with Optimal Resource Utilization using Answer Set Programming](https://mlsanity.com/paper/2008.03573/) (2020)
- [Thirty years of Epistemic Specifications](https://mlsanity.com/paper/2108.07669/) (2021)

## Abstract (arXiv)

There has been considerable recent interest in explainability in AI, especially with black-box machine learning models. As correctly observed by the planning community, when the application at hand is not a single-shot decision or prediction, but a sequence of actions that depend on observations, a richer notion of explanations are desirable. In this paper, we look to provide a formal account of ``counterfactual explanations," based in terms of action sequences. We then show that this naturally leads to an account of model reconciliation, which might take the form of the user correcting the agent's model, or suggesting actions to the agent's plan. For this, we will need to articulate what is true versus what is known, and we appeal to a modal fragment of the situation calculus to formalise these intuitions. We consider various settings: the agent knowing partial truths, weakened truths and having false beliefs, and show that our definitions easily generalize to these different settings.