Abstract
Counterfactual explanations, and their associated algorithmic recourse, are typically leveraged to understand and explain predictions of individual instances coming from a black-box classifier. In this paper, we propose to extend the use of counterfactuals to evaluate progress in sequential decision making tasks. To this end, we introduce a model-agnostic modular framework, TraCE (Trajectory Counterfactual Explanation) scores, to distill and condense progress in highly complex scenarios into a single value. We demonstrate TraCE's utility by showcasing its main properties in two case studies spanning healthcare and climate change.
| Original language | English |
|---|---|
| Title of host publication | Proceedings of the Northern Lights Deep Learning Conference 2024 |
| Editors | Tetiana Lutchyn, Adín Ramírez Rivera, Benjamin Ricaud |
| Pages | 36-45 |
| Number of pages | 10 |
| Publication status | Published - 11 Jan 2024 |
| Event | 5th Northern Lights Deep Learning Conference, NLDL 2024 - UiT The Arctic University, Tromso, Norway Duration: 9 Jan 2024 → 11 Jan 2024 https://www.nldl.org/ |
Publication series
| Name | Proceedings of Machine Learning Research |
|---|---|
| Volume | 233 |
| ISSN (Print) | 2640-3498 |
Conference
| Conference | 5th Northern Lights Deep Learning Conference, NLDL 2024 |
|---|---|
| Country/Territory | Norway |
| City | Tromso |
| Period | 9/01/24 → 11/01/24 |
| Internet address |
Bibliographical note
Publisher Copyright:© 2024 Jeffrey N. Clark, Edward A. Small, Nawid Keshtmand, Michelle W.L. Wan, Elena Fillola Mayoral, Enrico Werner, Christopher P. Bourdeaux, and Raul Santos-Rodriguez.
