Explainable Deep Reinforcement Learning: State of the Art and Challenges
Open accessThe DOI matches its Crossref record; 13 of the 13 references it cites were found in a registry or an index.
Abstract
Interpretability, explainability, and transparency are key issues to introducing artificial intelligence methods in many critical domains. This is important due to ethical concerns and trust issues strongly connected to reliability, robustness, auditability, and fairness, and has important consequences toward keeping the human in the loop in high levels of automation, especially in critical cases for decision making, where both (human and the machine) play important roles. Although the research community has given much attention to explainability of closed (or black) prediction boxes, there are tremendous needs for explainability of closed-box methods that support agents to act autonomously in the real world. Reinforcement learning methods, and especially their deep versions, are such closed-box methods. In this article, we aim to provide a review of state-of-the-art methods for explainable deep reinforcement learning methods, taking also into account the needs of human operators—that is, of those who make the actual and critical decisions in solving real-world problems. We provide a formal specification of the deep reinforcement learning explainability problems, and we identify the necessary components of a general explainable reinforcement learning framework. Based on these, we provide a comprehensive review of state-of-the-art methods, categorizing them into classes according to the paradigm they follow, the interpretable models they use, and the surface representation of explanations provided. The article concludes by identifying open questions and important challenges.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
A Unified Approach to Interpreting Model Predictions
Lundberg, S., & Lee, S.-I. (2017) · arXiv · doi:10.48550/arxiv.1705.07874 · Cited by 23,707
Distilling the Knowledge in a Neural Network
Hinton, G., Vinyals, O., & Dean, J. (2015) · arXiv · doi:10.48550/arxiv.1503.02531 · Cited by 14,204
“Why Should I Trust You?”: Explaining the Predictions of Any Classifier
Ribeiro, M., Singh, S., & Guestrin, C. (2016) · Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Demonstrations · doi:10.18653/v1/n16-3020 · Cited by 6,682
Lipton, Z. C. (2018) · Queue · doi:10.1145/3236386.3241340 · Cited by 5,427
Explanation in artificial intelligence: Insights from the social sciences
Miller, T. (2019) · Artificial Intelligence · doi:10.1016/j.artint.2018.07.007 · Cited by 5,302
On Pixel-Wise Explanations for Non-Linear Classifier Decisions by Layer-Wise Relevance Propagation
Bach, S., Binder, A., Montavon, G., Klauschen, F., Müller, K.-R., & Samek, W. (2015) · PLOS ONE · doi:10.1371/journal.pone.0130140 · Cited by 4,524 · cited 15×
“Other methods, such as for instance Layer-wise Relevance Propagation (LRP) [34] and Neighborhood Compo516 nent Analysis (NCA) [8] can be used for addressing the model inspection problem for any of the constituent DRL 517 models, also providing (a) importance of features inM , and (b) reduction of predictors’ dimension…”
BackgroundMethodInfluentialDefinitions, methods, and applications in interpretable machine learning
Murdoch, W. J., Singh, C., Kumbier, K., Abbasi-Asl, R., & Yu, B. (2019) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.1900654116 · Cited by 2,167 · cited 1×
“Relevance, of information for relationships contained in the data or learned by a machine learning model, for a particular audience into a chosen problem, is an important aspect of interpretable models in [55].”
BackgroundAnchors: High-Precision Model-Agnostic Explanations
Ribeiro, M. T., Singh, S., & Guestrin, C. (2018) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v32i1.11491 · Cited by 2,162 · cited 3×
“478 Extensions of LIME include Anchor [31] that uses decision rules as local interpretable classiier cl , and LORE 479 [32] that learns a local interpretable predictor cl on a synthetic neighborhood using a genetic algorithm.”
MethodInfluentialA Multidisciplinary Survey and Framework for Design and Evaluation of Explainable AI Systems
Mohseni, S., Zarei, N., & Ragan, E. D. (2018) · arXiv · doi:10.48550/arxiv.1811.11839 · Cited by 783
Distilling a neural network into a soft decision tree
Frosst, N., & Hinton, G. (2017) · arXiv · doi:10.48550/arxiv.1711.09784 · Cited by 407
Cited by
Karras, A., Giannaros, A., Amasiadi, N., & Karras, C. (2026) · Future Internet · doi:10.3390/fi18020083 · Cited by 10
Dizon-Paradis, O., Wormald, S., Capecci, D., Bhandarkar, A., & Woodard, D. (2023) · doi:10.36227/techrxiv.23739099 · Cited by 1
Metadata
- Authors
- Vouros, G. A.
- Venue
- ACM Computing Surveys
- Published
- 2022-12-03
- Volume
- 55
- Issue
- 5
- Pages
- 1-39
- ISSN
- 0360-0300 (ISSN-L)
- 1557-7341
- OpenAlex
- W4220970489
- Cited by
- 160
- References
- 13
- FWCI
- 9.21
- Type
- journal article
- Language
- en
- Open access
- green
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.