Abstract / Description of output
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.
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.
Original language | English |
---|---|
Title of host publication | Proceedings 40th International Conference on Logic Programming |
Publisher | Open Publishing Association |
Publication status | Accepted/In press - 24 Mar 2023 |
Event | The 39th International Conference on Logic Programming - Imperial College London, London, United Kingdom Duration: 9 Jul 2023 → 15 Jul 2023 Conference number: 39 https://iclp2023.imperial.ac.uk/home |
Publication series
Name | Electronic Proceedings in Theoretical Computer Science (EPTCS). |
---|---|
Publisher | Open Publishing Association |
ISSN (Electronic) | 2075-2180 |
Conference
Conference | The 39th International Conference on Logic Programming |
---|---|
Abbreviated title | ICLP 2023 |
Country/Territory | United Kingdom |
City | London |
Period | 9/07/23 → 15/07/23 |
Internet address |